PanDrugs: a novel method to prioritize anticancer drug treatments according to individual genomic data

PanDrugs: a novel method to prioritize anticancer drug treatments according to individual genomic... Background: Large-sequencing cancer genome projects have shown that tumors have thousands of molecular alterations and their frequency is highly heterogeneous. In such scenarios, physicians and oncologists routinely face lists of cancer genomic alterations where only a minority of them are relevant biomarkers to drive clinical decision- making. For this reason, the medical community agrees on the urgent need of methodologies to establish the relevance of tumor alterations, assisting in genomic profile interpretation, and, more importantly, to prioritize those that could be clinically actionable for cancer therapy. Results: We present PanDrugs, a new computational methodology to guide the selection of personalized treatments in cancer patients using the variant lists provided by genome-wide sequencing analyses. PanDrugs offers the largest database of drug-target associations available from well-known targeted therapies to preclinical drugs. Scoring data- driven gene cancer relevance and drug feasibility PanDrugs interprets genomic alterations and provides a prioritized evidence-based list of anticancer therapies. Our tool represents the first drug prescription strategy applying a rational based on pathway context, multi-gene markers impact and information provided by functional experiments. Our approach has been systematically applied to TCGA patients and successfully validated in a cancer case study with a xenograft mouse model demonstrating its utility. Conclusions: PanDrugs is a feasible method to identify potentially druggable molecular alterations and prioritize drugs to facilitate the interpretation of genomic landscape and clinical decision-making in cancer patients. Our approach expands the search of druggable genomic alterations from the concept of cancer driver genes to the druggable pathway context extending anticancer therapeutic options beyond already known cancer genes. The methodology is public and easily integratable with custom pipelines through its programmatic API or its docker image. The PanDrugs webtool is freely accessible at http://www.pandrugs.org. Keywords: Precision oncology, Personalized medicine, Translational bioinformatics, Cancer genomics,Insilicoprescription, Targeted therapy, Druggable genome * Correspondence: falshahrour@cnio.es Elena Piñeiro-Yáñez and Miguel Reboiro-Jato contributed equally to this work. Spanish National Cancer Research Centre (CNIO), 3rd Melchor Fernandez Almagro st., E-28029 Madrid, Spain Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 2 of 11 Background atreatment, thepathwaycontext [17], and the pharmaco- Identifying the most appropriate therapies from cancer logical evidence reported in large-scale experiments [18, 19]. genome data is a major challenge in personalized cancer medicine. Standard of care treatments are commonly se- Implementation lected following criteria such as: cancer type; stage; pa- PanDrugs database tient status; and/or the presence of prognostic and The PanDrugs database (PanDrugsdb) has been imple- predictive biomarkers. However, cancer treatment could mented to store gene–drug relationships. PanDrugs be revolutionized if the information contained in large methodology mines PanDrugsdb to provide a catalogue genomic datasets were to be systematically deconvoluted of prioritized candidate drugs and targetable genes esti- in terms of potential treatments [1]. Here, the identifica- mated from a list of variants (or genes) provided by a tion and evaluation of somatic alterations and their col- user (Fig. 1a). lective impact on tumor progression pose considerable Pharmacological data and drug annotations available in challenges to their clinical application [2, 3]. More spe- PanDrugsdb were collected from 24 databases. These in- cifically, physicians and researchers are challenged with cluded 18 sources with information curated by experts long lists of tumor-specific genomic variants where most and gene–drug associations obtained from experimental variants are either clinically “unactionable,” their bio- drug screenings: The Cancer Therapeutics Response Por- logical role unknown, or they are irrelevant for tumor tal [19] and GDSC [20] (Additional file 1: Table S1). Since biology [4]. In addition, the current list of cancer driver different sources employ a variety of non-standardized genes [5] has clinical limitations since genomic alter- identifiers to mention the same compound, drug names ations in these genes may be essential for oncogenesis, were standardized in order to be consistently integrated tumor cell growth, and survival; but the same genes may within PanDrugsdb (Additional file 1: Materials and not be targetable by current therapies [6]. Moreover, Methods). Following this, drug annotations were enriched very recent studies have revealed that cancer gene lists with additional information regarding drug families, drug are still incomplete and that there are many more cancer indication status, cancer type, and therapy description genes yet to be discovered [7–9]. In this scenario, it is (Additional file 1: Figure S1 and S2). Gene–drug relation- essential to develop new methodologies to analyze gen- ships were also annotated with the type of gene–drug rela- etic alterations in terms of treatment options, helping to tion (i.e. drug target or biomarker), drug sensitivity or prioritize those that could be useful for the management resistance response, and the type of genomic alteration as- of cancer patients. sociated to the drug response. Several remarkable efforts have addressed the The current version of PanDrugsdb includes 9092 prioritization of genomic alterations [10–13]. These drugs, 4804 unique genes, and 43,909 direct and methods exploit the extensive literature and knowledge non-redundant gene–drug interactions. The database is available in public repositories to catalogue cancer gen- built in MySQL RDBMS. Full details regarding Pan- omic variants and their impact on biological functions, Drugsdb implementations and PanDrugs software are although none of these methodologies directly link gen- described in Additional file 1: Materials and Methods. omic alterations to potential therapies. Tools such DGIdb [14], OncoKB [15], and the Cancer Genome In- PanDrugs nomenclature terpreter (CGI) [16] have been developed to identify PanDrugs categorizes druggable genes as: (1) direct tar- clinically actionable genomic alterations in tumors. Al- gets; (2) biomarkers; and (3) pathway members. though these tools demonstrate the potential of targeted The term “direct targets” includes those genes that con- therapies and provide drug repurposing strategies, they tribute to disease phenotype and can be directly targeted present some limitations. They only consider known by a drug (i.e. small molecule or monoclonal antibody). cancer driver genes for drug prescription, they are based For instance, BRAF is a direct target for vemurafenib [21]. exclusively on somatic DNA alterations, the therapeutic op- “Biomarkers” refers to genes that have a genetic status as- tions are restricted to “one target - one drug” ignoring mul- sociated with drug response (according to clinical or tiple targetable mutations and the protein pathway-specific pre-clinical evidence) but the protein product is not the activity, and they do not provide a prioritized list of treat- drug target itself. For example, BRCA-mutated cancers ments based on clinical, biological, and pharmacological that respond to poly-ADP-ribose polymerase (PARP)in- evidence. hibitors [22], PTEN loss that is associated with decreased Here we introduce PanDrugs, a new computational sensitivity of colorectal cancer tumors to anti-EGFR resource to propose drug therapies from genome-wide antibodies [23], or mutations in TSC1/2 as clinically ap- experimental results, including variant and gene lists. Pan- proved biomarkers of PI3K/Akt/mTOR inhibitor response Drugs expands cancer therapeutic options by taking into [24, 25]. PanDrugs “biomarkers” information was obtained account multiple genomic events potentially responsive to from manually curated databases (see Additional file 1: Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 3 of 11 Fig. 1 PanDrugs score calculation. a Overview of the DScore and GScore calculation and their corresponding annotation sources. PanDrugs considers drug indication and status, gene–drug associations and number of hits to calculate the DScore. GScore is estimated according to gene essentiality and tumoral vulnerability, gene relevance in cancer, the biological impact of mutations, the frequency of gene alterations, and their clinical implications. b PanDrugs considers the “Best therapeutic candidates” based on the accumulated and weighted scoring of GScore and DScore Materials and Methods for details) and from experimental Finally, PanDrugs analyzes the “collective gene impact” assays in cancer cell lines (GDSC and CTRP). defined as the number of druggable genes (direct targets, Targeted therapies may target cell signals that are biomarkers, and pathway members) in the input list that needed for cancer cells to develop, proliferate, and in- points to a particular drug. Those drugs capable of target- vade. Drugs targeting the activity of the surrounding ing the largest number of druggable genes are prioritized. interactors in the biological pathway of a mutated gene could: (1) produce the same downstream effect as target- PanDrugs uses two scores to prioritize cancer treatments ing the mutated gene itself; (2) enhance response by syn- PanDrugs calculates two scores integrating a variety of ergistic effects; and (3) be used in combination to avoid clinical, biological, and pharmacological sources and possible compensatory drug resistance mechanisms databases to suggest tailored anticancer therapies based [26–29]. Following this paradigm, PanDrugs includes on user supplied variant lists and PanDrugsdb (Fig. 1a). “pathway member” referring to all those downstream Gene Score (GScore) is in the range of 0–1 based on the druggable targets taking advantage of the pathway back- level of evidence supporting gene clinical implication ground underlying the user’s gene list. Interestingly, this and its biological relevance in cancer (Additional file 1: paradigm unlocks alternative therapeutic ways for untar- Figure S3A). Drug Score (DScore) estimates drug re- getable genes. sponse and treatment suitability (Additional file 1: Figure Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 4 of 11 S3B). A larger number of supporting databases, curated Although it is known that only few genomic alterations annotation, and clinical impact enhance the weight in will follow these stringent criteria, the emerging “drug- both GScore and DScore calculation. Full descriptions gable genome” concept opens the whole genome to of GScore and DScore calculations are available in therapeutic intervention. In other words, both a given Additional file 1: Materials and Methods. mutated gene and its interactions are putative drug tar- GScore has been implemented to consider: (1) gen- gets [37, 38]. Following this paradigm, PanDrugs offers a omic feature evidence by mutation consequence, func- systems biology framework to propose drugs that arise tional impact, and population allele frequency; (2) as rational therapeutic candidates. For example, MET relevance in cancer estimated by Cancer Gene Census amplification plays a role in acquired resistance to EGFR (CGC) of COSMIC v84 [30], TumorPortal resource [31], inhibitors of patients with EGFR-mutated tumors by ac- Tamborero et al. [32], and OncoScape [33]; (3) essential- tivating MAPK and PI3K/AKT signaling pathways [39, ity from RNA interference (RNAi) experiments in cancer 40]. Combination therapy of EGFR and MET inhibitors cell lines from Achilles project [34, 35] and; (4) clinical is used to block both MET and EGFR signaling pathways implications based on its pathogenicity supporting evi- [41]. In this scenario, PanDrugs proposes the following dence (taken from COSMIC and ClinVar). GScore as therapeutic options: (1) avoiding EGFR inhibitors weight assignation for non-ranked gene lists and for alone due to the known lack of sensitivity; (2) the usage VCF files are described in Additional file 1: Tables S2 of MET inhibitors that can overcome resistance of and S3, respectively. EGFR-TKIs; and (3) targeting downstream druggable DScore is calculated using PanDrugsdb to evaluate the genes (i.e. RAF, MEK) with available drugs (i.e. Sorafenib, therapeutic implications of those altered genes previ- Trametinib) blocking the oncogenic consequences of the ously employed for GScore calculation. DScore takes pathway overstimulation [42, 43] (Fig. 2). into account: (1) drug-cancer type indication (from the FDA and clinicaltrials.gov); (2) drug clinical status (ap- PanDrugs web tool and application programming interface proved by the FDA, clinical trials, or preclinical); (3) PanDrugs is available as a user-friendly web tool with pre- gene–drug relationship (i.e. direct target, biomarker, or loaded queries and demo examples at http://www.pandrug- pathway member); (4) number of curated databases sup- s.org. The detailed user manual is accessible online at porting that relationship (i.e. database factor); and (5) https://www.pandrugs.org/#!/help. The server supports three collective gene impact (Additional file 1: Figure S3C). alternative types of input: (1) single and multiple queries DScore has values from − 1 to 1 where negative values (gene lists); (2) standard VCF files; and (3) a ranked list of correspond to drug unresponsiveness and positive values genes, where ranking is obtained from experimental obser- to drug sensitivity (Additional file 1: Figure S3B). vations (i.e. RNA-sequencing experiments). Regular analysis PanDrugs provides a prioritized list of candidate drugs of a 500-line VCF file takes an average time of ~ 6 min in considering GScore and DScore values. Those drug ther- thePanDrugsserver. Theresults page provides a panel with apies supported by scores nearest to 1 in both GScore basic statistics of the query, pie-charts depicting clinical and DScore will have more evidence for their effective- status distribution and families for the drugs proposed, and ness in cancer treatment and will be considered “Best abubbleplotrepresentingGScoreand DScore together therapeutic candidates” by PanDrugs (Fig. 1b). with the best candidate therapies suggested by PanDrugs. Moreover, the tool generates a ranked summary table of Exploiting pathway information to increase therapeutic the treatments with raw scores and links to external data- options bases (Additional file 1:FigureS4).InadditiontoGScore PanDrugs expands the anticancer therapeutic arsenal and DScore, the summary table displays comprehensive suggesting drugs to target genes located downstream to and sortable information about genes, drugs, type of gene– the altered gene(s). To this end, PanDrugs integrates drug interaction, drug family, drug clinical status, type of previously modelled biological circuit information (e.g. therapy available (only for the approved drugs), and sources signaling pathways) [36], the interaction types between of annotation employed. PanDrugs uses KEGG pathways to nodes (activation or inhibition), and the gene functional map the relationship between input genes and pathway role (oncogene or tumor suppressor). Ideally, a perfect members suggested as candidate drug targets. Our tool also gene–drug(s) solution will meet the following criteria: supports drug queries to explore the gene–drug interac- (1) the gene is affected by activating point mutations tions available in PanDrugsdb providing the subsequent (predicted by functional impact algorithms or confirmed summary table. All results generated by PanDrugs are easily by databases/literature); (2) the gene is essential in syn- downloadable in standard formats (i.e. CSV, PNG, PDF, thetic lethal screenings; (3) the gene is sensitive to spe- etc.). PanDrugs provides a REST-based Application Pro- cific drugs included in PanDrugs; and (4) an FDA drug gramming Interface (API) allowing developers to make is approved or under clinical trial that targets the gene. queries directly to PanDrugsdb, to incorporate output from Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 5 of 11 Fig. 2 Possible scenarios for PanDrugs therapeutic candidates. PanDrugs proposes three potential types of druggable candidates. This includes: (1) direct targets, a gene that contributes to a disease phenotype and can be directly targeted by a drug; (2) drug-resistance biomarkers, a gene which genetic status is associated with a drug response from clinical or pre-clinical evidence but its protein product is not the direct target of the drug; and (3) pathway members, a targetable gene located downstream to the altered one. To illustrate this, tumors mutated in EGFR carrying MET amplifications will not respond to EGFR inhibitors (red). PanDrugs proposes as therapeutic strategy MET inhibitors and targeting MET downstream proteins (green) to drive tumor cell death PanDrugs within their own algorithms, and to combine the drops drastically at GScore = 0.4. In agreement with previ- tool as a novel module in NGS analysis pipelines integrat- ous studies, this observation suggests that most genomic al- ing genetic data and therapeutic alternatives. PanDrugs is terationsinTCGApatientshavelittleevidenceof being also available as a docker image at https://github.com/ associated with cancer and, in consequence, are poorly an- sing-group/pandrugs-docker. notated in public databases [31, 44](Additional file 1:Figure S6A). Genes with a GScore > 0.4 and carrying at least Results one mutational and/or CNV event were used to identify PanDrugs in The Cancer Genome Atlas (TCGA) data potential therapies and were found present in > 6000 of PanDrugs has been systematically applied to a cohort of the 7069 TCGA samples (Additional file 1: Figure S6B, 7069 samples from the TCGA project that correspond to S6C; Table S6). We decided to use this threshold for 20 different tumor types (Additional file 1: Figure S5). TCGA analysis to establish a compromise solution be- File sources employed for the TCGA analysis are listed tween gene annotation quality and retaining the largest in Additional file 1: Table S4. Databases used in the number of patients. For instance, EGFR mutations across study and their corresponding versions are detailed in different TCGA tumor types exhibit a 0.23 < GScore Additional file 1: Table S5. Full results may be inter- < 0.82 while mutations in KRAS have GScore values in actively accessed at the PanDrugs website (http:// the range of 0.36–0.97. These GScores underline the bio- www.pandrugs.org/). logical relevance and clinical utility of both KRAS and ThePanDrugsanalysisofTCGAtumorsshowedthatthe EGFR genomic alterations in cancer. Differences in GScore distribution of genes affected by genomic alteration GScore values reflect the higher frequency of KRAS hot- events (SNVs + indels + Copy Number Variations (CNVs)) spot mutations and their well-known pathogenicity in Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 6 of 11 contrast to the EGFR mutational heterogeneity and its Gene–drug associations per database are compared in broader functional impact and pathological diversity Fig. 3b. PanDrugs supports single and multiple queries (Additional file 1: Figure S7). In particular, EGFR (i.e. gene lists) as well as standard VCF files, while GScore ~ 0.8 reveals those well-known mutations with clinical DGIdb, CR, and PCT do not support variants and significance related to drug response (L858R) while OncoKB, PCT, and CR do not accept multiple gene EGFR ~ 0.4 corresponds to mutations residues that queries. To make tool comparisons viable, we selected as GScore are not located in the protein kinase domain and that input only those altered genes annotated in cBioPortal. are less frequently found in cancer patients but that may The comparison was carried out by selecting an have a deleterious functional effect (i.e. EGFR p.L62R). EGFR-mutant lung adenocarcinoma patient with known KRAS ~ 0.9 represents well-described and very fre- drug-resistance mechanisms to EGFR inhibitors via GScore quent driver mutations located in exon 2 codons 12 and MET amplifications from the TCGA cohort (NSCLC, 13 with clinical significance as diagnostic, prognostic, TCGA-38-4629). Our results show that only PanDrugs and predictive biomarkers. and CGI alert of the risk of a possible resistance mech- Using a GScore threshold our observations showed that anism to EGFR inhibitors. By contrast, the other tools an in silico prescription of approved drugs for direct tar- offer EGFR inhibitors as main therapeutic options since gets plus biomarkers offered treatments for 65% of pa- they do not support co-occurring alterations among tients when point mutations, indels, and CNVs are their functionalities. considered simultaneously. (Additional file 1:Figure S8A). Remarkably, only PanDrugs suggests clinically ap- Notably, PanDrugs is able to extend drug prescription for proved treatments and drugs in clinical trials for genes 86% of TCGA patients by exploiting pathway member– not considered by the other tools using biomarkers and drug associations (Additional file 1:Figure S8B). pathway members. For instance, PanDrugs prescribes The PanDrugs TCGA analysis was also used to evaluate Palbociclib (Additional file 1: Figure S10A), a selective the potential of non-driver genes as effective targets for inhibitor of the cyclin-dependent kinases CDK4 and cancer treatment by selecting the most frequently altered CDK6, to treat this NSCLC patient as a result of the fol- genes in TCGA patients annotated in PanDrugsdb. This lowing evidence: (1) CDK6 is a direct target; (2) CCND1, selection was carried out for every TCGA tumor type by CDKN2A, and CDKN2B are biomarkers; and (iii) CDK4 considering the top five genes according to frequency in is a downstream pathway member gene. Another clear small variants (point mutations and indels) and CNVs sep- example is Navitoclax, a BCL2 family inhibitor currently arately. Following these criteria, we obtained 200 alter- tested in clinical trials for NSCLC (NCT02520778). Pan- ations located in 100 genes (Additional file 1: Table S7). Drugs suggests Navitoclax since CDK6 is a biomarker of Interestingly, 54% of these most frequently altered genes Navitoclax response and BCL2 as pathway member be- were found druggable but are currently labelled as cause it is downstream to the TP53 and ERBB2 genes non-driver genes [45, 32]. Our results strongly suggest which are altered in this particular NSCLC patient. that the extension of genomic events detection beyond Interestingly, PanDrugs is also capable of expanding the known cancer driver genes can help in finding add- drug prescription beyond known cancer genes. To illus- itional effective therapies for cancer treatment. trate this, we use the list of novel candidate cancer genes We compared PanDrugs’ performance by applying our provided by Martincorena et al. PanDrugs analysis re- methodology to the TCGA cohort previously used in vealed 436 gene–drug associations not reported by the other studies [46, 47] This comparative TCGA analysis other tools (i.e. MAP2K7-Lenalidomide, BMPR2-Serdeme- showed that the PanDrugs pathway member approach tan, or ZFP36L2-Embelin) (Additional file 1: Figure S10B). expands therapeutic options for FDA-approved drugs to As expected, all these associations have low GScores due an average percentage of 93.41% in TCGA patients to the limited clinical and biological gene annotations; (Additional file 1: Figure S9). This result shows that the however, 32 associations corresponding to 13 non-driver PanDrugs pathway member strategy might be useful to genes showed DScore > 0.7, suggesting their viability as complement current in silico prescription tools. potential targets for cancer treatment. PanDrugs has been integrated within an online resource Comparison with other tools for PanCancer Analysis of Whole Genomes (PCAWG) We compared PanDrugs’ performance to DGIdb, covering 2658 donors from 48 cancer types [50]. Among OncoKB, CGI, CancerResource (CR) [48], Personalized these donors, we chose three patients without druggable Cancer Therapy (PCT, https://pct.mdanderson.org/), cancer driver-altered genes from colorectal, breast, and JAX-Clinical Knowledgebase [49], and Precision Medi- prostate cancer to evaluate PanDrugs therapeutic proposals. cine Knowledgebase (https://pmkb.weill.cornell.edu/). The colorectal cancer patient (DO10486) showed 32 pre- The tools included in the comparison and the descrip- dicted damaged genes. None of the six altered drivers tion of their main functionalities are shown in Fig. 3a. (STAT3, SOX9, ARID1A, TGFBR2, RTN4, PPP2R1A)are Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 7 of 11 Fig. 3 a Comparison of current in silico drug prescription tools based on genomic data. b Venn diagram for drug–gene associations available in DGIdb v3.0.2, Cancer Genome Interpreter, OncoKB, and PanDrugs. Global data for associations from CancerResource and Personalized Cancer Therapy is not accessible. Total numbers for non-redundant drug–gene interactions after drug standardization using PubChem to compare the resources are 29,197 (DGidb), 349 (CGI), 129 (OncoKB), and 43,909 (PanDrugs) currently targeted with approved drugs although STAT3 in- PanDrugs application in a cancer case study hibitors are under clinical trial. PanDrugs was queried with Unfortunately, detailed clinical annotations for patients the 32 damaged genes and proposed, among others, Dabra- in cancer genomics international consortiums are not fenib (LIMK1 as the direct target) and Paclitaxel (CDK5R1 publicly available to validate PanDrugs results. To over- as the pathway member). come this limitation, we have experimentally validated In the TP53-deficient breast cancer patient (DO5375) PanDrugs results using a patient-derived cancer mouse with 15 damaged genes detected (none of them known xenograft (PDX) model on which several therapeutic drivers), Vismodegib, an inhibitor of the Hedgehog sig- strategies have been tested as part of a personalized naling pathway, is proposed by PanDrugs as best thera- medicine protocol. The protocol includes whole exome peutic candidate driven by LRP2, a damaged gene that sequencing analysis and the development of PDX models belongs to this pathway. Paclitaxel and Doxorubicin are as described elsewhere [51]. also proposed by PanDrugs as the best therapeutic can- For this validation, our case study was a 58-year-old didates. Interestingly, the combination of Vismodegib man diagnosed with advanced squamous cell lung carcin- plus Paclitaxel and Epirubicin (an analog of Doxorubi- oma (SCLC; stage IV with brain metastasis). After surgery cin) is currently under clinical trial as neoadjuvant (R0), he received a first-line chemotherapy with carbopla- chemotherapy in triple negative breast cancer patients tin/Paclitaxel. Pemetrexed/Erlotinib was administered as a [https://clinicaltrials.gov/ct2/show/NCT02694224]. Eight second-line therapy to treat the progression of the disease. non-driver damaged genes were found in the prostate Tumor and normal samples of this patient were se- cancer case (DO46813). Currently these genes have no quenced to identify tumor-specific sequence alterations. drugs available to directly target them. Here, PanDrugs as- We found 965 somatic mutations and 501 somatic copy signs the best DScores to approved MEK inhibitors and number alterations (389 gain regions and 112 loss regions) Vinblastine, an antitumoral alkaloid, to target pathway (Additional file 1: Figure S11A). We detected 318 genes members downstream to the damaged TRAF2 gene. that would have proteins classified as damaged by mutation These examples highlight PanDrugs’ capability for propos- consequences such as stop gains, frameshifts, and deleteri- ing drugs used in clinical practice in those cases with no ous missense variants. Additional file 1: Table S8 summa- known driver mutations and limited molecular evidence. rizes the 46 gene mutations predicted as deleterious. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 8 of 11 The patient’s variant list (e.g. the complete VCF file) pathway member) for the genes under assessment. In was evaluated by PanDrugs to identify druggable gen- addition, the DScore calculation integrates in vitro omic alterations. MAPK pathway inhibitors were sug- drug-screening data from GDSC and CTRP. GScore gested as the best candidates. (Additional file 1: Figure and DScore are finally evaluated together to create a S11B). Indeed, likely underlying this suggestion, examin- gene–drug ranking offering personalized candidate ation of the genomic events in this patient revealed dele- treatments for the variant input list. This approach esti- terious somatic mutations in HRAS (G13 V), NF1 mates the treatment adequacy based on the gene–drug (K297*), and RAF1 (M562I) proteins. These mutations associations covering more biomedical sources than are predicted as damaging and may produce an activa- any other current in silico prescription tool. tion of MAPK/ERK pathway. Constitutive activation of Unfortunately, current anticancer therapies are based this pathway has been associated with cancers of the on single biomarkers that do not consider the mutational lung, colon, melanoma, lung, thyroid, leukemia, and landscape of the tumor and intratumoral clonal hetero- pancreas [52] what makes it a suitable target to treat geneity [55]. Additionally, cancer genomics studies have these tumours. MAPK inhibitors include compounds clearly revealed that tumoral survival and progression are targeting MAP2K1 (MEK). Also, MAPK activity can still mainly activated by accumulation of genetic alterations in occur as a result of PI3K activation through RAS. Dual crucial molecular pathways rather than driven by single activation of these two pathways is observed in a num- gene alterations [56, 57]. Thus, an accurate assessment of ber of cancer types including melanoma, prostate, and tumor fingerprints is essential for the development of ef- colorectal cancer, and provides the rationale for combin- fective therapies taking into account the collective gene ing therapeutic agents [53]. impact and pathway context of genomic alterations from We then performed an in vivo evaluation of the effi- cancer patients [58, 59]. Our methodology evaluates the cacy of several targeted antitumor agents—PI3K inhibi- collective gene impact by assigning a higher DScore to tor (PI3Ki), MEK inhibitor (MEKi), rapamycin, dasatinib, that drug capable to target the highest number of genes and lapatinib—in a low passage PDX mouse model for found in the input list. PanDrugs also provides prioritized this SCLC patient. Statistically significant (p < 0.05) treatments beyond single direct targets and biomarkers tumor growth inhibition was reported for MEKi and found in variant lists by exploiting the context of pathway PI3Ki treatments compared with the control group at members. Following the druggable genome paradigm Pan- the time point considered. Overall, benefit was reported Drugs offers a systems biology knowledge-based layer that with the combination of MEKi and PI3Ki towards the automatically inspects biological circuits. Interestingly, avatar model tested (Additional file 1: Figure S11C). this expands cancer candidate therapies from beyond lim- ited cancer-related gene lists to the whole druggable path- Discussion way. To our knowledge, there is no other current tool Precision oncology requires novel resources and tools to with similar characteristics. translate cancer genomic landscapes to clinical utility in PanDrugs-assisted therapeutic strategies have been order to prescribe rational, efficient, and tailored treatments systematically applied to large patient cohorts using to individual cancer patients [54]. The PanDrugs method TCGA patients. The feasibility of our candidate treat- has been implemented to address the interpretation gap be- ment proposals has been also tested in PDX experimen- tween raw genomic data and clinical usefulness. To this tal models. In these analyses, we found that the pathway end, our methodology relies on PanDrugsdb, the largest member paradigm is able to expand in silico drug pre- catalogue of drug-target associations currently available. scriptions for already approved drugs. This might have a This database is publicly accessible and relates druggable direct impact on improving clinical decision-making by genes to already approved treatments, well-known targeted extending treatment opportunities to those patients therapies, and preclinical drugs. without a clear approved pharmacogenomics biomarker. Starting from user-supplied gene or variant lists, Pan- The TCGA analysis was carried out establishing a Drugs identifies and prioritizes both direct or indirect compromise GScore threshold to retrieve highly reliable targetable genomic alterations in tumors using a novel candidate treatments for well-known genes. This strat- approach based on two scores: GScore and DScore. The egy avoids handling huge lists of results by preserving GScore calculates target suitability for each variant (or best candidates, but also discards drug–gene associations gene) by considering its essentiality using RNAi experimen- found in poorly annotated genes. However, Martincor- tal data from the Achilles project, gene relevance in cancer, ena et al. have recently reported that half of the coding tumor frequency, and the biological and clinical impacts. driver mutations occur outside of known cancer driver The DScore evaluates drug applicability by considering genes [7]. If this is true, precision oncology will demand its clinical indications, drug status, collective gene im- the implementation of novel methodologies capable of pact, and druggability (e.g. direct target, biomarker, or prescribing therapies beyond known cancer genes. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 9 of 11 PanDrugs offers novel therapeutic strategies for such tumors. Indeed, PanDrugs represents the first drug pre- genes; lowering the GScore threshold while keeping the scription tool that proposes cancer therapies with a ra- default DScore cut-off is enough to uncover reliable tionale based on pathway context, collective gene therapeutic options that can target poorly annotated impact, and information provided by functional experi- genes. This would allow the discovery of novel clinically ments. PanDrugs has demonstrated its adaptability by significant and actionable mutations that could become being systematically applied to large cohorts of patients new genetic predictive and prognostic biomarkers. and by providing candidate treatments directed to drug- It is important to remark that PanDrugs is more than gable genes beyond cancer driver genes. Overall, our an organized catalogue of known gene–drug relation- method highlights new areas of opportunity for advan- ships. PanDrugs is the first method to systematically cing precision cancer medicine, providing a novel and infer novel targeted treatments following a rational fully accessible method that could be useful in decreas- framework supported by multi-gene markers, molecular ing the complexity of the interpretation of genomic data pathway context, and pharmacological evidence. Our re- and clinical decision-making. PanDrugs is freely available sults show that in silico prescription approaches focused at http://www.pandrugs.org. uniquely on known cancer genes should be complemen- ted by incorporating drug information associated to gen- Availability and requirements omic alterations located in non-cancer genes. Our Project name: PanDrugs. approach extends the treatment opportunities of cancer Project home page: http://www.pandrugs.org patients by enriching the therapeutic arsenal against tu- Operating system(s): Platform independent. mors and opens new avenues for personalized medicine. Programming language: Java, MySQL RDBMS, Perl. Although PanDrugs offers a valuable methodology for Other requirements: Chrome, Firefox, Safari. in silico prescription, further efforts are required to im- License: GPLv3. prove cancer treatment by proposing more effective drugs and anticipating drug resistance. Since drug effi- Additional file cacy substantially depends on tissue-, cell-, and molecular-specific context [60], precision oncology tools Additional file 1: Supplementary materials and methods, Figures S1–S11 and Tables S1–S8. (PDF 14845 kb) should integrate data beyond pure genomics [61] includ- ing the combination of high-throughput drug screenings Abbreviations and functional experiments to unravel heterogeneous CGI: Cancer Genome Interpreter; CR: CancerResource; NSCLC: Non-small cell multi-omic dependencies influencing response to ther- lung cancer; PCT: Personalized cancer therapy; PDX: Patient-derived apy [62, 63]. In addition, the integration of additional xenograft; PMKB: Precision Medicine Knowledgebase; RNAi: RNA interference; SCLC: Small-cell lung cancer; TCGA: The Cancer Genome Atlas; TKI: Tyrosine biological relationship layers such as protein interaction kinase inhibitor; VCF: Variant call format networks [64], transcriptional regulatory modules [65], or pathway activity footprints [66] should improve drug Acknowledgements The authors thank Joaquín Dopazo, Patricia León, and José Carbonell for prioritization and will help to propose alternative thera- kindly providing the modelled pathways employed in PanDrugs peutic strategies. It is also crucial to have a comprehen- implementation; and Michael Tress for his helpful comments and sive and well-structured drug ontology available that suggestions in the earlier version of the manuscript. provides drug annotations (i.e. drug indication, mecha- Funding nisms of action, chemical structure, side effects, Part of our own work has been supported by a Marie-Curie Career drug-target associations, and drug families) for a more Integration Grant (CIG) CIG334361. KT and JP-P are supported by a Severo accurate drug prescription [67]. Ochoa FPI grant doctoral fellowship by the Spanish Ministry of Economy and Competitiveness. MR-J and DG-P are supported by the Biomedical Research Finally, it should be emphasized that current in silico Centre (“Centro Singular de Galicia,” www.cinbio.es) funded by “Consellería drug prescription tools are limited by the lack of large de Cultura, Educación e Ordenación Universitaria,”“Xunta de Galicia,” and longitudinal precision medicine studies with accessible FEDER (European Union) and by the CITI (Centro de Investigación Transferencia e Innovación) from the University of Vigo. JMR was supported by the Spanish clinical records. Such studies are crucial to assess and National Institute of Bioinformatics (www.inab.org), a platform of the “Instituto validate drug proposals and refine in silico prescription de Salud Carlos III” [INB-ISCIII, PRB2]. algorithms to consider additional factors such as mode Availability of data and materials of drug administration, combinatorial therapies, drug re- The webtool is freely accessible at http://www.pandrugs.org and through its positioning, and side effects. programmatic API or docker image. Conclusions Authors’ contributions EP-Y, MR-J, DG-P, KT, and JMR designed PanDrugsdb, developed the PanDrugs provides a feasible method to guide computational framework, and continue to maintain PanDrugs; EP-Y and MR-J genomic-hypothesis therapies as well as to prioritize led and executed the benchmarking analyses with input from JP-P, HT, JC, GG-L, multiple druggable alterations in genomically complex and MH and supervision by DG-P and FA; PPL-C and MH carried out the Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 10 of 11 experimental validation in PDX models; TS and JC provide fruitful discussion 13. Eilbeck K, Quinlan A, Yandell M. Settling the score: variant prioritization and and valuable suggestions to improve the methodology; GG-L and FA designed Mendelian disease. Nat Rev Genet. 2017;18(10):599–612. https://doi.org/10. the study and wrote the manuscript. FA initiated and led the study. All authors 1038/nrg.2017.52 read and approved the final manuscript. 14. Griffith M, Griffith OL, Coffman AC, Weible JV, McMichael JF, Spies NC, et al. DGIdb 2.0: mining clinically relevant drug–gene interactions. Nat Methods. Ethics approval and consent to participate 2013;10:1209–10. Animals were maintained at the Spanish National Cancer Research Centre 15. Chakravarty D, Gao J, Phillips SM, Kundra R, Zhang H, Wang J, et al. OncoKB: (CNIO) Animal Facility (awarded with the AAALAC accreditation) in A Precision Oncology Knowledge Base. JCO Precis Oncol. 2017; https://doi. accordance with the guidelines stated in the International Guiding Principles org/10.1200/PO.17.00011 for Biomedical Research Involving Animals, developed by the Council for 16. Tamborero D, Rubio-Perez C, Deu-Pons J, Schroeder MP, Vivancos A, Rovira International Organizations of Medical Sciences (CIOMS). Four- to six-week- A, et al. Cancer Genome Interpreter annotates the biological and clinical old female athymic nude-Foxn1nu (nu/nu) mice were purchased from Envigo relevance of tumor alterations. Genome Med. 2018;10:25. https://doi.org/10. (Barcelona, Spain). All animal experiments were approved by the Competent 1101/140475 Authority of Comunidad de Madrid (project PROEX 104/16). Patients pro- 17. Amadoz A, Sebastian-Leon P, Vidal E, Salavert F, Dopazo J. Using activation vided written informed consent before study enrollment. The inform consent status of signaling pathways as mechanism-based biomarkers to predict form was reviewed and approved by the IRB of Hospital de Madrid in ac- drug sensitivity. Sci Rep. 2015;5:18494. https://doi.org/10.1038/srep18494 cordance with the guiding principles of the Declaration of Helsinki. 18. Garnett MJ, Edelman EJ, Heidorn SJ, Greenman CD, Dastur A, Lau KW, et al. Systematic identification of genomic markers of drug sensitivity in cancer Competing interests cells. Nature. 2012;483(7391):570–5. The authors declare that they have no competing interests. 19. Basu A, Bodycombe NE, Cheah JH, Price EV, Liu K, Schaefer GI, et al. An interactive resource to identify cancer genetic and lineage dependencies targeted by small molecules. Cell. 2013;154:1151–61. Publisher’sNote 20. Iorio F,Knijnenburg TA, Vis DJ,Bignell GR,Menden MP,Schubert M, et al.A Springer Nature remains neutral with regard to jurisdictional claims in published landscape of pharmacogenomic interactions in cancer. Cell. 2016;166(3):740–54. maps and institutional affiliations. 21. Bollag G, Tsai J, Zhang J, Zhang C, Ibrahim P, Nolop K, et al. Vemurafenib: the first drug approved for BRAF-mutant cancer. Nat Rev Drug Discov. 2012; Author details 11(11):873–86. Spanish National Cancer Research Centre (CNIO), 3rd Melchor Fernandez 22. Bryant HE, Schultz N, Thomas HD, Parker KM, Flower D, Lopez E, et al. Almagro st., E-28029 Madrid, Spain. Computer Science Department - Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP- University of Vigo, Vigo, Spain. Biomedical Research Centre (CINBIO), Vigo, ribose) polymerase. Nature. 2005;434(7035):913–7. Spain. Spanish National Bioinformatics Institute (INB), Madrid, Spain. 5 6 23. De Roock W, De Vriendt V, Normanno N, Ciardiello F, Tejpar S. KRAS, BRAF, Department of Physiology - University of Valencia, Valencia, Spain. Loyola 7 PIK3CA, and PTEN mutations: implications for targeted therapies in University Chicago Stritch School of Medicine, Maywood, IL, USA. Beth Israel metastatic colorectal cancer. Lancet Oncol. 2011;12(6):594–603. Deaconess Medical Center, Boston, USA. 24. Iyer G, Hanrahan AJ, Milowsky MI, Al-Ahmadie H, Scott SN, Janakiraman M. Genome sequencing identifies a basis for everolimus sensitivity. Science. Received: 28 December 2017 Accepted: 4 May 2018 2012;338(6104):221. 25. Perini GF, Campregher PV, Ross JS, Ali S, Hamerschlak N, Santos FPS. Clinical response to everolimus in a patient with Hodgkin's lymphoma harboring a TSC2 References mutation. Blood Cancer J. 2016;6(e420) https://doi.org/10.1038/bcj.2016.25. 1. Kamalakaran S, Varadan V, Janevski A, Banerjee N, Tuck D, McCombie WR, 26. Faber AC, Li D, Song Y, Liang MC, Yeap BY, Bronson RT, et al. Differential et al. Translating next generation sequencing to practice: opportunities and induction of apoptosis in HER2 and EGFR addicted cancers following PI3K necessary steps. Mol Oncol. 2013;7(4):743–55. inhibition. Proc Natl Acad Sci U S A. 2009;106(46):19503–8. 2. Lyon GJ, Wang K. Identifying disease mutations in genomic medicine 27. Zhu Z, Aref AR, Cohoon TJ, Barbie TU, Imamura Y, Yang S, et al. Inhibition of settings: current challenges and how to accelerate progress. Genome Med. KRAS-driven tumorigenicity by interruption of an autocrine cytokine circuit. 2012;4:7–58. Cancer Discov. 2014;4(4):452–65. 3. Andre F, Mardis E, Salm M, Soria JC, Siu LL, Swanton C. Prioritizing targets 28. Butler DE, Marlein C, Walker HF, Frame FM, Mann VM, Simms MS, et al. for precision cancer medicine. Ann Oncol. 2014;25:2295–303. Inhibition of the PI3K/AKT/mTOR pathway activates autophagy and 4. Hyman DM, Taylor BS, Baselga J. Implementing genome-driven oncology. compensatory Ras/Raf/MEK/ERK signalling in prostate cancer. Oncotarget. Cell. 2017;168(4):584–99. 2017;8(34):56698–713. 5. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA Jr, Kinzler KW. 29. Lee HJ, Zhuang G, Cao Y, Du P, Kim HJ, Settleman J. Drug resistance via Cancer genome landscapes. Science. 2013;339(6127):1546–58. feedback activation of Stat3 in oncogene-addicted cancer cells. Cancer Cell. 6. Van Allen EM, Wagle N, Stojanov P, Perrin DL, Cibulskis K, Marlow S, et al. 2014;26(2):207–21. Whole-exome sequencing and clinical interpretation of formalin-fixed, 30. Forbes SA, Beare D, Boutselakis H, Bamford S, Bindal N, Tate J, et al. COSMIC: paraffin-embedded tumor samples to guide precision cancer medicine. Nat somatic cancer genetics at high-resolution. Nucleic Acids Res. 2017;45(D1): Med. 2014;20(6):682–8. D777–83. 7. Martincorena I, Raine KM, Gerstung M, Dawson KJ, Haase K, Van Loo P, et al. 31. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, Golub TR, Universal patterns of selection in cancer and somatic tissues. Cell. 2017; et al. Discovery and saturation analysis of cancer genes across 21 tumour 171(5):1029–41. types. Nature. 2014;505(7484):495–501. https://doi.org/10.1038/nature12912 8. Cheng J, Demeulemeester J, Wedge DC, Vollan HKM, Pitt JJ, Russnes HG, 32. Tamborero D, Gonzalez-Perez A, Perez-Llamas C, Deu-Pons J, Kandoth C, et al. Pan-cancer analysis of homozygous deletions in primary tumours Reimand J, et al. Comprehensive identification of mutational cancer driver uncovers rare tumour suppressors. Nat Commun. 2017;1:1221. genes across 12 tumor types. Sci Rep. 2013;3:2650. 9. Peterson TA, Gauran IIM, Park J, Park D, Kann MG. Oncodomains: a protein 33. Schlicker A, Michaut M, Rahman R, Wessels LF. OncoScape: Exploring domain-centric framework for analyzing rare variants in tumor samples. PLoS the cancer aberration landscape by genomic data fusion. Sci Rep. Comput Biol. 2017;4:e1005428. https://doi.org/10.1371/journal.pcbi.1005428 2016;6:28103. 10. Wei CH, Phan L, Feltz J, Maiti R, Hefferon T, Lu Z. tmVar 2.0: integrating genomic variant information from literature with dbSNP and ClinVar for 34. Cowley GS, Weir BA, Vazquez F, Tamayo P, Scott JA, Rusin S, et al. Parallel precision medicine. Bioinformatics. 2018;34:80–7. https://doi.org/10.1093/ genome-scale loss of function screens in 216 cancer cell lines for the bioinformatics/btx541 identification of context-specific genetic dependencies. Sci Data. 2014;1: 11. Yang H, Robinson PN, Wang K. Phenolyzer: phenotype-based prioritization 140035. of candidate genes for human diseases. Nat Methods. 2015;12(9):841–3. 35. Shao DD, Tsherniak A, Gopal S, Weir BA, Tamayo P, Stransky N, et al. ATARiS: 12. Ghosh R, Oak N, Plon SE. Evaluation of in silico algorithms for use with ACMG/ computational quantification of gene suppression phenotypes from AMP clinical variant interpretation guidelines. Genome Biol. 2017;18:225. multisample RNAi screens. Genome Res. 2013;23(4):665–78. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 11 of 11 36. Salavert F, Hidalgo MR, Amadoz A, Çubuk C, Medina I, Crespo D, et al. 62. Tsherniak A, Vazquez F, Montgomery PG, Weir BA, Kryukov G, Cowley GS, Actionable pathways: interactive discovery of therapeutic targets using et al. Defining a cancer dependency map. Cell. 2017;170(3):564–76. signaling pathway models. Nucleic Acids Res. 2016;44(W1):W212–6. 63. McDonald ER, de Weck A, Schlabach MR, Billy E, Mavrakis KJ, Hoffman GR, 37. Workman P, Al-Lazikani B. Drugging cancer genomes. Nat Rev Drug Discov. et al. Project DRIVE: a compendium of cancer dependencies and synthetic 2013;12(12):889–90. lethal relationships uncovered by large-scale, deep RNAi screening. Cell. 38. Editorial. It's all druggable. Nat Genet. 2017;49(2):169. https://doi.org/10. 2017;170(3):577–92. 1038/ng.3788. 64. Ivanov AA, Revennaugh B, Rusnak L, Gonzalez-Pecchi V, Mo X, Johns MA, et al. The OncoPPi Portal: an integrative resource to explore and prioritize 39. Engelman JA, Zejnullahu K, Mitsudomi T, Song Y, Hyland C, Park J, et al. protein-protein interactions for cancer target discovery. Bioinformatics. 2018; MET amplification leads to gefitinib resistance in lung cancer by activating 34:1183–91. https://doi.org/10.1093/bioinformatics/btx743 ERBB3 signaling. Science. 2007;316(5827):1039–43. 65. Garcia-Alonso L, Iorio F, Matchan A, Fonseca N, Jaaks P, Peat G, et al. 40. Turke AB, Zejnullahu K, Wu YL, Song Y, Dias-Santagata D, Lifshits E, et al. Transcription factor activities enhance markers of drug sensitivity in cancer. Preexistence and clonal selection of MET amplification in EGFR mutant Cancer Res. 2018;78:3,769–80. NSCLC. Cancer Cell. 2010;17(1):77–88. 66. Schubert M, Klinger B, Klünemann M, Sieber A, Uhlitz F, Sauer S, et al. 41. Gainor JF, Niederst MJ, Lennerz JK, Dagogo-Jack I, Stevens S, Shaw AT, et al. Perturbation-response genes reveal signaling footprints in cancer gene Dramatic response to combination erlotinib and crizotinib in a patient with expression. Nat Commun. 2018;9(1):20. advanced, EGFR-mutant lung cancer harboring De Novo MET amplification. 67. Sharp ME. Toward a comprehensive drug ontology: extraction of drug- J Thorac Oncol. 2016;11(7):83–5. indication relations from diverse information sources. J Biomed Semantics. 42. Tricker EM, Xu C, Uddin S, Capelletti M, Ercan D, Ogino A, et al. Combined 2017;8(1):2. EGFR/MEK inhibition prevents the emergence of resistance in EGFR-Mutant lung cancer. Cancer Discov. 2015;5(9):960–71. 43. Ortiz-Cuaran S, Scheffler M, Plenker D, Dahmen L, Scheel AH, Fernandez-Cuesta L, et al. Heterogeneous mechanisms of primary and acquired resistance to third-generation EGFR inhibitors. Clin Cancer Res. 2016;22(19):4837–47. 44. Kandoth C, McLellan MD, Vandin F, Ye K, Niu B, Lu C, et al. Mutational landscape and significance across 12 major cancer types. Nature. 2013; 502(7471):333–9. https://doi.org/10.1038/nature12634 45. Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R, et al. A census of human cancer genes. Nat Rev Cancer. 2004;4(3):177–83. 46. Dienstmann R, Jang IS, Bot B, Friend S, Guinney J. Database of genomic biomarkers for cancer drugs and clinical targetability in solid tumors. Cancer Discov. 2015;5(2):118–23. 47. Rubio-Perez C, Tamborero D, Schroeder MP, Antolín AA, Deu-Pons J, Perez- Llamas C, et al. In silico prescription of anticancer drugs to cohorts of 28 tumor types reveals targeting opportunities. Cancer Cell. 2015;27(3):382–96. 48. Gohlke BO, Nickel J, Otto R, Dunkel M, Preissner R. CancerResource–updated database of cancer-relevant proteins, mutations and interacting drugs. Nucleic Acids Res. 2016;44(D1):D932–7. 49. Patterson SE, Liu R, Statz CM, Durkin D, Lakshminarayana A, Mockus SM. The clinical trial landscape in oncology and connectivity of somatic mutational profiles to targeted therapies. Hum Genomics. 2016;10:4. 50. Goldman M, Zhang J, Fonseca NA, Xiang Q, Craft B, Piñeiro-Yáñez E, et al. Online resources for PCAWG data exploration, visualization, and discovery. https://www.biorxiv.org/content/early/2017/10/20/163907 51. Garralda E, Paz K, López-Casas PP, Jones S, Katz A, Kann LM, et al. Integrated next-generation sequencing and avatar mouse models for personalized cancer treatment. Clin Cancer Res. 2014;20(9):2476–84. 52. Dhillon AS, Hagan S, Rath O, Kolch W. MAP kinase signalling pathways in cancer. Oncogene. 2007;26(22):3279–90. 53. Mendoza MC, Er EE, Blenis J. The Ras-ERK and PI3K-mTOR pathways: cross- talk and compensation. Trends Biochem Sci. 2011;36(6):320–8. 54. Gómez-López G, Dopazo J, Cigudosa JC, Valencia A, Al-Shahrour F. Precision medicine needs pioneering clinical bioinformaticians. Brief Bioinform. 2017; https://doi.org/10.1093/bib/bbx144. 55. Dagogo-Jack I, Shaw AT. Tumour heterogeneity and resistance to cancer therapies. Nat Rev Clin Oncol. 2018:15, 81–94. https://doi.org/10.1038/ nrclinonc.2017.166 56. Garraway LA, Verweij J, Ballman KV. Precision oncology: an overview. J Clin Oncol. 2013;31(15):1803–5. 57. Califano A, Alvarez MJ. The recurrent architecture of tumour initiation, progression and drug sensitivity. Nat Rev Cancer. 2017;17(2):116–30. 58. Naulaerts S, Dang CC, Ballester PJ. Precision and recall oncology: combining multiple gene mutations for improved identification of drug-sensitive tumours. Oncotarget. 2017;8:97025–40. 59. Schneider G, Schmidt-Supprian M, Rad R, Saur D. Tissue-specific tumorigenesis: context matters. Nat Rev Cancer. 2017;17(4):239–53. https://doi.org/10.1038/nrc.2017.5. 60. Mina M, Raynaud F, Tavernari D, Battistello E, Sungalee S, Saghafinia S, et al. Conditional selection of genomic alterations dictates cancer evolution and oncogenic dependencies. Cancer Cell. 2017;32(2):155–68. 61. Letai A. Functional precision cancer medicine-moving beyond pure genomics. Nat Med. 2017;23(9):1028–35. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Genome Medicine Springer Journals

Loading next page...
 
/lp/springer_journal/pandrugs-a-novel-method-to-prioritize-anticancer-drug-treatments-rxm7kbw61v
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s).
Subject
Biomedicine; Human Genetics; Metabolomics; Bioinformatics; Medicine/Public Health, general; Cancer Research; Systems Biology
eISSN
1756-994X
D.O.I.
10.1186/s13073-018-0546-1
Publisher site
See Article on Publisher Site

Abstract

Background: Large-sequencing cancer genome projects have shown that tumors have thousands of molecular alterations and their frequency is highly heterogeneous. In such scenarios, physicians and oncologists routinely face lists of cancer genomic alterations where only a minority of them are relevant biomarkers to drive clinical decision- making. For this reason, the medical community agrees on the urgent need of methodologies to establish the relevance of tumor alterations, assisting in genomic profile interpretation, and, more importantly, to prioritize those that could be clinically actionable for cancer therapy. Results: We present PanDrugs, a new computational methodology to guide the selection of personalized treatments in cancer patients using the variant lists provided by genome-wide sequencing analyses. PanDrugs offers the largest database of drug-target associations available from well-known targeted therapies to preclinical drugs. Scoring data- driven gene cancer relevance and drug feasibility PanDrugs interprets genomic alterations and provides a prioritized evidence-based list of anticancer therapies. Our tool represents the first drug prescription strategy applying a rational based on pathway context, multi-gene markers impact and information provided by functional experiments. Our approach has been systematically applied to TCGA patients and successfully validated in a cancer case study with a xenograft mouse model demonstrating its utility. Conclusions: PanDrugs is a feasible method to identify potentially druggable molecular alterations and prioritize drugs to facilitate the interpretation of genomic landscape and clinical decision-making in cancer patients. Our approach expands the search of druggable genomic alterations from the concept of cancer driver genes to the druggable pathway context extending anticancer therapeutic options beyond already known cancer genes. The methodology is public and easily integratable with custom pipelines through its programmatic API or its docker image. The PanDrugs webtool is freely accessible at http://www.pandrugs.org. Keywords: Precision oncology, Personalized medicine, Translational bioinformatics, Cancer genomics,Insilicoprescription, Targeted therapy, Druggable genome * Correspondence: falshahrour@cnio.es Elena Piñeiro-Yáñez and Miguel Reboiro-Jato contributed equally to this work. Spanish National Cancer Research Centre (CNIO), 3rd Melchor Fernandez Almagro st., E-28029 Madrid, Spain Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 2 of 11 Background atreatment, thepathwaycontext [17], and the pharmaco- Identifying the most appropriate therapies from cancer logical evidence reported in large-scale experiments [18, 19]. genome data is a major challenge in personalized cancer medicine. Standard of care treatments are commonly se- Implementation lected following criteria such as: cancer type; stage; pa- PanDrugs database tient status; and/or the presence of prognostic and The PanDrugs database (PanDrugsdb) has been imple- predictive biomarkers. However, cancer treatment could mented to store gene–drug relationships. PanDrugs be revolutionized if the information contained in large methodology mines PanDrugsdb to provide a catalogue genomic datasets were to be systematically deconvoluted of prioritized candidate drugs and targetable genes esti- in terms of potential treatments [1]. Here, the identifica- mated from a list of variants (or genes) provided by a tion and evaluation of somatic alterations and their col- user (Fig. 1a). lective impact on tumor progression pose considerable Pharmacological data and drug annotations available in challenges to their clinical application [2, 3]. More spe- PanDrugsdb were collected from 24 databases. These in- cifically, physicians and researchers are challenged with cluded 18 sources with information curated by experts long lists of tumor-specific genomic variants where most and gene–drug associations obtained from experimental variants are either clinically “unactionable,” their bio- drug screenings: The Cancer Therapeutics Response Por- logical role unknown, or they are irrelevant for tumor tal [19] and GDSC [20] (Additional file 1: Table S1). Since biology [4]. In addition, the current list of cancer driver different sources employ a variety of non-standardized genes [5] has clinical limitations since genomic alter- identifiers to mention the same compound, drug names ations in these genes may be essential for oncogenesis, were standardized in order to be consistently integrated tumor cell growth, and survival; but the same genes may within PanDrugsdb (Additional file 1: Materials and not be targetable by current therapies [6]. Moreover, Methods). Following this, drug annotations were enriched very recent studies have revealed that cancer gene lists with additional information regarding drug families, drug are still incomplete and that there are many more cancer indication status, cancer type, and therapy description genes yet to be discovered [7–9]. In this scenario, it is (Additional file 1: Figure S1 and S2). Gene–drug relation- essential to develop new methodologies to analyze gen- ships were also annotated with the type of gene–drug rela- etic alterations in terms of treatment options, helping to tion (i.e. drug target or biomarker), drug sensitivity or prioritize those that could be useful for the management resistance response, and the type of genomic alteration as- of cancer patients. sociated to the drug response. Several remarkable efforts have addressed the The current version of PanDrugsdb includes 9092 prioritization of genomic alterations [10–13]. These drugs, 4804 unique genes, and 43,909 direct and methods exploit the extensive literature and knowledge non-redundant gene–drug interactions. The database is available in public repositories to catalogue cancer gen- built in MySQL RDBMS. Full details regarding Pan- omic variants and their impact on biological functions, Drugsdb implementations and PanDrugs software are although none of these methodologies directly link gen- described in Additional file 1: Materials and Methods. omic alterations to potential therapies. Tools such DGIdb [14], OncoKB [15], and the Cancer Genome In- PanDrugs nomenclature terpreter (CGI) [16] have been developed to identify PanDrugs categorizes druggable genes as: (1) direct tar- clinically actionable genomic alterations in tumors. Al- gets; (2) biomarkers; and (3) pathway members. though these tools demonstrate the potential of targeted The term “direct targets” includes those genes that con- therapies and provide drug repurposing strategies, they tribute to disease phenotype and can be directly targeted present some limitations. They only consider known by a drug (i.e. small molecule or monoclonal antibody). cancer driver genes for drug prescription, they are based For instance, BRAF is a direct target for vemurafenib [21]. exclusively on somatic DNA alterations, the therapeutic op- “Biomarkers” refers to genes that have a genetic status as- tions are restricted to “one target - one drug” ignoring mul- sociated with drug response (according to clinical or tiple targetable mutations and the protein pathway-specific pre-clinical evidence) but the protein product is not the activity, and they do not provide a prioritized list of treat- drug target itself. For example, BRCA-mutated cancers ments based on clinical, biological, and pharmacological that respond to poly-ADP-ribose polymerase (PARP)in- evidence. hibitors [22], PTEN loss that is associated with decreased Here we introduce PanDrugs, a new computational sensitivity of colorectal cancer tumors to anti-EGFR resource to propose drug therapies from genome-wide antibodies [23], or mutations in TSC1/2 as clinically ap- experimental results, including variant and gene lists. Pan- proved biomarkers of PI3K/Akt/mTOR inhibitor response Drugs expands cancer therapeutic options by taking into [24, 25]. PanDrugs “biomarkers” information was obtained account multiple genomic events potentially responsive to from manually curated databases (see Additional file 1: Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 3 of 11 Fig. 1 PanDrugs score calculation. a Overview of the DScore and GScore calculation and their corresponding annotation sources. PanDrugs considers drug indication and status, gene–drug associations and number of hits to calculate the DScore. GScore is estimated according to gene essentiality and tumoral vulnerability, gene relevance in cancer, the biological impact of mutations, the frequency of gene alterations, and their clinical implications. b PanDrugs considers the “Best therapeutic candidates” based on the accumulated and weighted scoring of GScore and DScore Materials and Methods for details) and from experimental Finally, PanDrugs analyzes the “collective gene impact” assays in cancer cell lines (GDSC and CTRP). defined as the number of druggable genes (direct targets, Targeted therapies may target cell signals that are biomarkers, and pathway members) in the input list that needed for cancer cells to develop, proliferate, and in- points to a particular drug. Those drugs capable of target- vade. Drugs targeting the activity of the surrounding ing the largest number of druggable genes are prioritized. interactors in the biological pathway of a mutated gene could: (1) produce the same downstream effect as target- PanDrugs uses two scores to prioritize cancer treatments ing the mutated gene itself; (2) enhance response by syn- PanDrugs calculates two scores integrating a variety of ergistic effects; and (3) be used in combination to avoid clinical, biological, and pharmacological sources and possible compensatory drug resistance mechanisms databases to suggest tailored anticancer therapies based [26–29]. Following this paradigm, PanDrugs includes on user supplied variant lists and PanDrugsdb (Fig. 1a). “pathway member” referring to all those downstream Gene Score (GScore) is in the range of 0–1 based on the druggable targets taking advantage of the pathway back- level of evidence supporting gene clinical implication ground underlying the user’s gene list. Interestingly, this and its biological relevance in cancer (Additional file 1: paradigm unlocks alternative therapeutic ways for untar- Figure S3A). Drug Score (DScore) estimates drug re- getable genes. sponse and treatment suitability (Additional file 1: Figure Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 4 of 11 S3B). A larger number of supporting databases, curated Although it is known that only few genomic alterations annotation, and clinical impact enhance the weight in will follow these stringent criteria, the emerging “drug- both GScore and DScore calculation. Full descriptions gable genome” concept opens the whole genome to of GScore and DScore calculations are available in therapeutic intervention. In other words, both a given Additional file 1: Materials and Methods. mutated gene and its interactions are putative drug tar- GScore has been implemented to consider: (1) gen- gets [37, 38]. Following this paradigm, PanDrugs offers a omic feature evidence by mutation consequence, func- systems biology framework to propose drugs that arise tional impact, and population allele frequency; (2) as rational therapeutic candidates. For example, MET relevance in cancer estimated by Cancer Gene Census amplification plays a role in acquired resistance to EGFR (CGC) of COSMIC v84 [30], TumorPortal resource [31], inhibitors of patients with EGFR-mutated tumors by ac- Tamborero et al. [32], and OncoScape [33]; (3) essential- tivating MAPK and PI3K/AKT signaling pathways [39, ity from RNA interference (RNAi) experiments in cancer 40]. Combination therapy of EGFR and MET inhibitors cell lines from Achilles project [34, 35] and; (4) clinical is used to block both MET and EGFR signaling pathways implications based on its pathogenicity supporting evi- [41]. In this scenario, PanDrugs proposes the following dence (taken from COSMIC and ClinVar). GScore as therapeutic options: (1) avoiding EGFR inhibitors weight assignation for non-ranked gene lists and for alone due to the known lack of sensitivity; (2) the usage VCF files are described in Additional file 1: Tables S2 of MET inhibitors that can overcome resistance of and S3, respectively. EGFR-TKIs; and (3) targeting downstream druggable DScore is calculated using PanDrugsdb to evaluate the genes (i.e. RAF, MEK) with available drugs (i.e. Sorafenib, therapeutic implications of those altered genes previ- Trametinib) blocking the oncogenic consequences of the ously employed for GScore calculation. DScore takes pathway overstimulation [42, 43] (Fig. 2). into account: (1) drug-cancer type indication (from the FDA and clinicaltrials.gov); (2) drug clinical status (ap- PanDrugs web tool and application programming interface proved by the FDA, clinical trials, or preclinical); (3) PanDrugs is available as a user-friendly web tool with pre- gene–drug relationship (i.e. direct target, biomarker, or loaded queries and demo examples at http://www.pandrug- pathway member); (4) number of curated databases sup- s.org. The detailed user manual is accessible online at porting that relationship (i.e. database factor); and (5) https://www.pandrugs.org/#!/help. The server supports three collective gene impact (Additional file 1: Figure S3C). alternative types of input: (1) single and multiple queries DScore has values from − 1 to 1 where negative values (gene lists); (2) standard VCF files; and (3) a ranked list of correspond to drug unresponsiveness and positive values genes, where ranking is obtained from experimental obser- to drug sensitivity (Additional file 1: Figure S3B). vations (i.e. RNA-sequencing experiments). Regular analysis PanDrugs provides a prioritized list of candidate drugs of a 500-line VCF file takes an average time of ~ 6 min in considering GScore and DScore values. Those drug ther- thePanDrugsserver. Theresults page provides a panel with apies supported by scores nearest to 1 in both GScore basic statistics of the query, pie-charts depicting clinical and DScore will have more evidence for their effective- status distribution and families for the drugs proposed, and ness in cancer treatment and will be considered “Best abubbleplotrepresentingGScoreand DScore together therapeutic candidates” by PanDrugs (Fig. 1b). with the best candidate therapies suggested by PanDrugs. Moreover, the tool generates a ranked summary table of Exploiting pathway information to increase therapeutic the treatments with raw scores and links to external data- options bases (Additional file 1:FigureS4).InadditiontoGScore PanDrugs expands the anticancer therapeutic arsenal and DScore, the summary table displays comprehensive suggesting drugs to target genes located downstream to and sortable information about genes, drugs, type of gene– the altered gene(s). To this end, PanDrugs integrates drug interaction, drug family, drug clinical status, type of previously modelled biological circuit information (e.g. therapy available (only for the approved drugs), and sources signaling pathways) [36], the interaction types between of annotation employed. PanDrugs uses KEGG pathways to nodes (activation or inhibition), and the gene functional map the relationship between input genes and pathway role (oncogene or tumor suppressor). Ideally, a perfect members suggested as candidate drug targets. Our tool also gene–drug(s) solution will meet the following criteria: supports drug queries to explore the gene–drug interac- (1) the gene is affected by activating point mutations tions available in PanDrugsdb providing the subsequent (predicted by functional impact algorithms or confirmed summary table. All results generated by PanDrugs are easily by databases/literature); (2) the gene is essential in syn- downloadable in standard formats (i.e. CSV, PNG, PDF, thetic lethal screenings; (3) the gene is sensitive to spe- etc.). PanDrugs provides a REST-based Application Pro- cific drugs included in PanDrugs; and (4) an FDA drug gramming Interface (API) allowing developers to make is approved or under clinical trial that targets the gene. queries directly to PanDrugsdb, to incorporate output from Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 5 of 11 Fig. 2 Possible scenarios for PanDrugs therapeutic candidates. PanDrugs proposes three potential types of druggable candidates. This includes: (1) direct targets, a gene that contributes to a disease phenotype and can be directly targeted by a drug; (2) drug-resistance biomarkers, a gene which genetic status is associated with a drug response from clinical or pre-clinical evidence but its protein product is not the direct target of the drug; and (3) pathway members, a targetable gene located downstream to the altered one. To illustrate this, tumors mutated in EGFR carrying MET amplifications will not respond to EGFR inhibitors (red). PanDrugs proposes as therapeutic strategy MET inhibitors and targeting MET downstream proteins (green) to drive tumor cell death PanDrugs within their own algorithms, and to combine the drops drastically at GScore = 0.4. In agreement with previ- tool as a novel module in NGS analysis pipelines integrat- ous studies, this observation suggests that most genomic al- ing genetic data and therapeutic alternatives. PanDrugs is terationsinTCGApatientshavelittleevidenceof being also available as a docker image at https://github.com/ associated with cancer and, in consequence, are poorly an- sing-group/pandrugs-docker. notated in public databases [31, 44](Additional file 1:Figure S6A). Genes with a GScore > 0.4 and carrying at least Results one mutational and/or CNV event were used to identify PanDrugs in The Cancer Genome Atlas (TCGA) data potential therapies and were found present in > 6000 of PanDrugs has been systematically applied to a cohort of the 7069 TCGA samples (Additional file 1: Figure S6B, 7069 samples from the TCGA project that correspond to S6C; Table S6). We decided to use this threshold for 20 different tumor types (Additional file 1: Figure S5). TCGA analysis to establish a compromise solution be- File sources employed for the TCGA analysis are listed tween gene annotation quality and retaining the largest in Additional file 1: Table S4. Databases used in the number of patients. For instance, EGFR mutations across study and their corresponding versions are detailed in different TCGA tumor types exhibit a 0.23 < GScore Additional file 1: Table S5. Full results may be inter- < 0.82 while mutations in KRAS have GScore values in actively accessed at the PanDrugs website (http:// the range of 0.36–0.97. These GScores underline the bio- www.pandrugs.org/). logical relevance and clinical utility of both KRAS and ThePanDrugsanalysisofTCGAtumorsshowedthatthe EGFR genomic alterations in cancer. Differences in GScore distribution of genes affected by genomic alteration GScore values reflect the higher frequency of KRAS hot- events (SNVs + indels + Copy Number Variations (CNVs)) spot mutations and their well-known pathogenicity in Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 6 of 11 contrast to the EGFR mutational heterogeneity and its Gene–drug associations per database are compared in broader functional impact and pathological diversity Fig. 3b. PanDrugs supports single and multiple queries (Additional file 1: Figure S7). In particular, EGFR (i.e. gene lists) as well as standard VCF files, while GScore ~ 0.8 reveals those well-known mutations with clinical DGIdb, CR, and PCT do not support variants and significance related to drug response (L858R) while OncoKB, PCT, and CR do not accept multiple gene EGFR ~ 0.4 corresponds to mutations residues that queries. To make tool comparisons viable, we selected as GScore are not located in the protein kinase domain and that input only those altered genes annotated in cBioPortal. are less frequently found in cancer patients but that may The comparison was carried out by selecting an have a deleterious functional effect (i.e. EGFR p.L62R). EGFR-mutant lung adenocarcinoma patient with known KRAS ~ 0.9 represents well-described and very fre- drug-resistance mechanisms to EGFR inhibitors via GScore quent driver mutations located in exon 2 codons 12 and MET amplifications from the TCGA cohort (NSCLC, 13 with clinical significance as diagnostic, prognostic, TCGA-38-4629). Our results show that only PanDrugs and predictive biomarkers. and CGI alert of the risk of a possible resistance mech- Using a GScore threshold our observations showed that anism to EGFR inhibitors. By contrast, the other tools an in silico prescription of approved drugs for direct tar- offer EGFR inhibitors as main therapeutic options since gets plus biomarkers offered treatments for 65% of pa- they do not support co-occurring alterations among tients when point mutations, indels, and CNVs are their functionalities. considered simultaneously. (Additional file 1:Figure S8A). Remarkably, only PanDrugs suggests clinically ap- Notably, PanDrugs is able to extend drug prescription for proved treatments and drugs in clinical trials for genes 86% of TCGA patients by exploiting pathway member– not considered by the other tools using biomarkers and drug associations (Additional file 1:Figure S8B). pathway members. For instance, PanDrugs prescribes The PanDrugs TCGA analysis was also used to evaluate Palbociclib (Additional file 1: Figure S10A), a selective the potential of non-driver genes as effective targets for inhibitor of the cyclin-dependent kinases CDK4 and cancer treatment by selecting the most frequently altered CDK6, to treat this NSCLC patient as a result of the fol- genes in TCGA patients annotated in PanDrugsdb. This lowing evidence: (1) CDK6 is a direct target; (2) CCND1, selection was carried out for every TCGA tumor type by CDKN2A, and CDKN2B are biomarkers; and (iii) CDK4 considering the top five genes according to frequency in is a downstream pathway member gene. Another clear small variants (point mutations and indels) and CNVs sep- example is Navitoclax, a BCL2 family inhibitor currently arately. Following these criteria, we obtained 200 alter- tested in clinical trials for NSCLC (NCT02520778). Pan- ations located in 100 genes (Additional file 1: Table S7). Drugs suggests Navitoclax since CDK6 is a biomarker of Interestingly, 54% of these most frequently altered genes Navitoclax response and BCL2 as pathway member be- were found druggable but are currently labelled as cause it is downstream to the TP53 and ERBB2 genes non-driver genes [45, 32]. Our results strongly suggest which are altered in this particular NSCLC patient. that the extension of genomic events detection beyond Interestingly, PanDrugs is also capable of expanding the known cancer driver genes can help in finding add- drug prescription beyond known cancer genes. To illus- itional effective therapies for cancer treatment. trate this, we use the list of novel candidate cancer genes We compared PanDrugs’ performance by applying our provided by Martincorena et al. PanDrugs analysis re- methodology to the TCGA cohort previously used in vealed 436 gene–drug associations not reported by the other studies [46, 47] This comparative TCGA analysis other tools (i.e. MAP2K7-Lenalidomide, BMPR2-Serdeme- showed that the PanDrugs pathway member approach tan, or ZFP36L2-Embelin) (Additional file 1: Figure S10B). expands therapeutic options for FDA-approved drugs to As expected, all these associations have low GScores due an average percentage of 93.41% in TCGA patients to the limited clinical and biological gene annotations; (Additional file 1: Figure S9). This result shows that the however, 32 associations corresponding to 13 non-driver PanDrugs pathway member strategy might be useful to genes showed DScore > 0.7, suggesting their viability as complement current in silico prescription tools. potential targets for cancer treatment. PanDrugs has been integrated within an online resource Comparison with other tools for PanCancer Analysis of Whole Genomes (PCAWG) We compared PanDrugs’ performance to DGIdb, covering 2658 donors from 48 cancer types [50]. Among OncoKB, CGI, CancerResource (CR) [48], Personalized these donors, we chose three patients without druggable Cancer Therapy (PCT, https://pct.mdanderson.org/), cancer driver-altered genes from colorectal, breast, and JAX-Clinical Knowledgebase [49], and Precision Medi- prostate cancer to evaluate PanDrugs therapeutic proposals. cine Knowledgebase (https://pmkb.weill.cornell.edu/). The colorectal cancer patient (DO10486) showed 32 pre- The tools included in the comparison and the descrip- dicted damaged genes. None of the six altered drivers tion of their main functionalities are shown in Fig. 3a. (STAT3, SOX9, ARID1A, TGFBR2, RTN4, PPP2R1A)are Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 7 of 11 Fig. 3 a Comparison of current in silico drug prescription tools based on genomic data. b Venn diagram for drug–gene associations available in DGIdb v3.0.2, Cancer Genome Interpreter, OncoKB, and PanDrugs. Global data for associations from CancerResource and Personalized Cancer Therapy is not accessible. Total numbers for non-redundant drug–gene interactions after drug standardization using PubChem to compare the resources are 29,197 (DGidb), 349 (CGI), 129 (OncoKB), and 43,909 (PanDrugs) currently targeted with approved drugs although STAT3 in- PanDrugs application in a cancer case study hibitors are under clinical trial. PanDrugs was queried with Unfortunately, detailed clinical annotations for patients the 32 damaged genes and proposed, among others, Dabra- in cancer genomics international consortiums are not fenib (LIMK1 as the direct target) and Paclitaxel (CDK5R1 publicly available to validate PanDrugs results. To over- as the pathway member). come this limitation, we have experimentally validated In the TP53-deficient breast cancer patient (DO5375) PanDrugs results using a patient-derived cancer mouse with 15 damaged genes detected (none of them known xenograft (PDX) model on which several therapeutic drivers), Vismodegib, an inhibitor of the Hedgehog sig- strategies have been tested as part of a personalized naling pathway, is proposed by PanDrugs as best thera- medicine protocol. The protocol includes whole exome peutic candidate driven by LRP2, a damaged gene that sequencing analysis and the development of PDX models belongs to this pathway. Paclitaxel and Doxorubicin are as described elsewhere [51]. also proposed by PanDrugs as the best therapeutic can- For this validation, our case study was a 58-year-old didates. Interestingly, the combination of Vismodegib man diagnosed with advanced squamous cell lung carcin- plus Paclitaxel and Epirubicin (an analog of Doxorubi- oma (SCLC; stage IV with brain metastasis). After surgery cin) is currently under clinical trial as neoadjuvant (R0), he received a first-line chemotherapy with carbopla- chemotherapy in triple negative breast cancer patients tin/Paclitaxel. Pemetrexed/Erlotinib was administered as a [https://clinicaltrials.gov/ct2/show/NCT02694224]. Eight second-line therapy to treat the progression of the disease. non-driver damaged genes were found in the prostate Tumor and normal samples of this patient were se- cancer case (DO46813). Currently these genes have no quenced to identify tumor-specific sequence alterations. drugs available to directly target them. Here, PanDrugs as- We found 965 somatic mutations and 501 somatic copy signs the best DScores to approved MEK inhibitors and number alterations (389 gain regions and 112 loss regions) Vinblastine, an antitumoral alkaloid, to target pathway (Additional file 1: Figure S11A). We detected 318 genes members downstream to the damaged TRAF2 gene. that would have proteins classified as damaged by mutation These examples highlight PanDrugs’ capability for propos- consequences such as stop gains, frameshifts, and deleteri- ing drugs used in clinical practice in those cases with no ous missense variants. Additional file 1: Table S8 summa- known driver mutations and limited molecular evidence. rizes the 46 gene mutations predicted as deleterious. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 8 of 11 The patient’s variant list (e.g. the complete VCF file) pathway member) for the genes under assessment. In was evaluated by PanDrugs to identify druggable gen- addition, the DScore calculation integrates in vitro omic alterations. MAPK pathway inhibitors were sug- drug-screening data from GDSC and CTRP. GScore gested as the best candidates. (Additional file 1: Figure and DScore are finally evaluated together to create a S11B). Indeed, likely underlying this suggestion, examin- gene–drug ranking offering personalized candidate ation of the genomic events in this patient revealed dele- treatments for the variant input list. This approach esti- terious somatic mutations in HRAS (G13 V), NF1 mates the treatment adequacy based on the gene–drug (K297*), and RAF1 (M562I) proteins. These mutations associations covering more biomedical sources than are predicted as damaging and may produce an activa- any other current in silico prescription tool. tion of MAPK/ERK pathway. Constitutive activation of Unfortunately, current anticancer therapies are based this pathway has been associated with cancers of the on single biomarkers that do not consider the mutational lung, colon, melanoma, lung, thyroid, leukemia, and landscape of the tumor and intratumoral clonal hetero- pancreas [52] what makes it a suitable target to treat geneity [55]. Additionally, cancer genomics studies have these tumours. MAPK inhibitors include compounds clearly revealed that tumoral survival and progression are targeting MAP2K1 (MEK). Also, MAPK activity can still mainly activated by accumulation of genetic alterations in occur as a result of PI3K activation through RAS. Dual crucial molecular pathways rather than driven by single activation of these two pathways is observed in a num- gene alterations [56, 57]. Thus, an accurate assessment of ber of cancer types including melanoma, prostate, and tumor fingerprints is essential for the development of ef- colorectal cancer, and provides the rationale for combin- fective therapies taking into account the collective gene ing therapeutic agents [53]. impact and pathway context of genomic alterations from We then performed an in vivo evaluation of the effi- cancer patients [58, 59]. Our methodology evaluates the cacy of several targeted antitumor agents—PI3K inhibi- collective gene impact by assigning a higher DScore to tor (PI3Ki), MEK inhibitor (MEKi), rapamycin, dasatinib, that drug capable to target the highest number of genes and lapatinib—in a low passage PDX mouse model for found in the input list. PanDrugs also provides prioritized this SCLC patient. Statistically significant (p < 0.05) treatments beyond single direct targets and biomarkers tumor growth inhibition was reported for MEKi and found in variant lists by exploiting the context of pathway PI3Ki treatments compared with the control group at members. Following the druggable genome paradigm Pan- the time point considered. Overall, benefit was reported Drugs offers a systems biology knowledge-based layer that with the combination of MEKi and PI3Ki towards the automatically inspects biological circuits. Interestingly, avatar model tested (Additional file 1: Figure S11C). this expands cancer candidate therapies from beyond lim- ited cancer-related gene lists to the whole druggable path- Discussion way. To our knowledge, there is no other current tool Precision oncology requires novel resources and tools to with similar characteristics. translate cancer genomic landscapes to clinical utility in PanDrugs-assisted therapeutic strategies have been order to prescribe rational, efficient, and tailored treatments systematically applied to large patient cohorts using to individual cancer patients [54]. The PanDrugs method TCGA patients. The feasibility of our candidate treat- has been implemented to address the interpretation gap be- ment proposals has been also tested in PDX experimen- tween raw genomic data and clinical usefulness. To this tal models. In these analyses, we found that the pathway end, our methodology relies on PanDrugsdb, the largest member paradigm is able to expand in silico drug pre- catalogue of drug-target associations currently available. scriptions for already approved drugs. This might have a This database is publicly accessible and relates druggable direct impact on improving clinical decision-making by genes to already approved treatments, well-known targeted extending treatment opportunities to those patients therapies, and preclinical drugs. without a clear approved pharmacogenomics biomarker. Starting from user-supplied gene or variant lists, Pan- The TCGA analysis was carried out establishing a Drugs identifies and prioritizes both direct or indirect compromise GScore threshold to retrieve highly reliable targetable genomic alterations in tumors using a novel candidate treatments for well-known genes. This strat- approach based on two scores: GScore and DScore. The egy avoids handling huge lists of results by preserving GScore calculates target suitability for each variant (or best candidates, but also discards drug–gene associations gene) by considering its essentiality using RNAi experimen- found in poorly annotated genes. However, Martincor- tal data from the Achilles project, gene relevance in cancer, ena et al. have recently reported that half of the coding tumor frequency, and the biological and clinical impacts. driver mutations occur outside of known cancer driver The DScore evaluates drug applicability by considering genes [7]. If this is true, precision oncology will demand its clinical indications, drug status, collective gene im- the implementation of novel methodologies capable of pact, and druggability (e.g. direct target, biomarker, or prescribing therapies beyond known cancer genes. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 9 of 11 PanDrugs offers novel therapeutic strategies for such tumors. Indeed, PanDrugs represents the first drug pre- genes; lowering the GScore threshold while keeping the scription tool that proposes cancer therapies with a ra- default DScore cut-off is enough to uncover reliable tionale based on pathway context, collective gene therapeutic options that can target poorly annotated impact, and information provided by functional experi- genes. This would allow the discovery of novel clinically ments. PanDrugs has demonstrated its adaptability by significant and actionable mutations that could become being systematically applied to large cohorts of patients new genetic predictive and prognostic biomarkers. and by providing candidate treatments directed to drug- It is important to remark that PanDrugs is more than gable genes beyond cancer driver genes. Overall, our an organized catalogue of known gene–drug relation- method highlights new areas of opportunity for advan- ships. PanDrugs is the first method to systematically cing precision cancer medicine, providing a novel and infer novel targeted treatments following a rational fully accessible method that could be useful in decreas- framework supported by multi-gene markers, molecular ing the complexity of the interpretation of genomic data pathway context, and pharmacological evidence. Our re- and clinical decision-making. PanDrugs is freely available sults show that in silico prescription approaches focused at http://www.pandrugs.org. uniquely on known cancer genes should be complemen- ted by incorporating drug information associated to gen- Availability and requirements omic alterations located in non-cancer genes. Our Project name: PanDrugs. approach extends the treatment opportunities of cancer Project home page: http://www.pandrugs.org patients by enriching the therapeutic arsenal against tu- Operating system(s): Platform independent. mors and opens new avenues for personalized medicine. Programming language: Java, MySQL RDBMS, Perl. Although PanDrugs offers a valuable methodology for Other requirements: Chrome, Firefox, Safari. in silico prescription, further efforts are required to im- License: GPLv3. prove cancer treatment by proposing more effective drugs and anticipating drug resistance. Since drug effi- Additional file cacy substantially depends on tissue-, cell-, and molecular-specific context [60], precision oncology tools Additional file 1: Supplementary materials and methods, Figures S1–S11 and Tables S1–S8. (PDF 14845 kb) should integrate data beyond pure genomics [61] includ- ing the combination of high-throughput drug screenings Abbreviations and functional experiments to unravel heterogeneous CGI: Cancer Genome Interpreter; CR: CancerResource; NSCLC: Non-small cell multi-omic dependencies influencing response to ther- lung cancer; PCT: Personalized cancer therapy; PDX: Patient-derived apy [62, 63]. In addition, the integration of additional xenograft; PMKB: Precision Medicine Knowledgebase; RNAi: RNA interference; SCLC: Small-cell lung cancer; TCGA: The Cancer Genome Atlas; TKI: Tyrosine biological relationship layers such as protein interaction kinase inhibitor; VCF: Variant call format networks [64], transcriptional regulatory modules [65], or pathway activity footprints [66] should improve drug Acknowledgements The authors thank Joaquín Dopazo, Patricia León, and José Carbonell for prioritization and will help to propose alternative thera- kindly providing the modelled pathways employed in PanDrugs peutic strategies. It is also crucial to have a comprehen- implementation; and Michael Tress for his helpful comments and sive and well-structured drug ontology available that suggestions in the earlier version of the manuscript. provides drug annotations (i.e. drug indication, mecha- Funding nisms of action, chemical structure, side effects, Part of our own work has been supported by a Marie-Curie Career drug-target associations, and drug families) for a more Integration Grant (CIG) CIG334361. KT and JP-P are supported by a Severo accurate drug prescription [67]. Ochoa FPI grant doctoral fellowship by the Spanish Ministry of Economy and Competitiveness. MR-J and DG-P are supported by the Biomedical Research Finally, it should be emphasized that current in silico Centre (“Centro Singular de Galicia,” www.cinbio.es) funded by “Consellería drug prescription tools are limited by the lack of large de Cultura, Educación e Ordenación Universitaria,”“Xunta de Galicia,” and longitudinal precision medicine studies with accessible FEDER (European Union) and by the CITI (Centro de Investigación Transferencia e Innovación) from the University of Vigo. JMR was supported by the Spanish clinical records. Such studies are crucial to assess and National Institute of Bioinformatics (www.inab.org), a platform of the “Instituto validate drug proposals and refine in silico prescription de Salud Carlos III” [INB-ISCIII, PRB2]. algorithms to consider additional factors such as mode Availability of data and materials of drug administration, combinatorial therapies, drug re- The webtool is freely accessible at http://www.pandrugs.org and through its positioning, and side effects. programmatic API or docker image. Conclusions Authors’ contributions EP-Y, MR-J, DG-P, KT, and JMR designed PanDrugsdb, developed the PanDrugs provides a feasible method to guide computational framework, and continue to maintain PanDrugs; EP-Y and MR-J genomic-hypothesis therapies as well as to prioritize led and executed the benchmarking analyses with input from JP-P, HT, JC, GG-L, multiple druggable alterations in genomically complex and MH and supervision by DG-P and FA; PPL-C and MH carried out the Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 10 of 11 experimental validation in PDX models; TS and JC provide fruitful discussion 13. Eilbeck K, Quinlan A, Yandell M. Settling the score: variant prioritization and and valuable suggestions to improve the methodology; GG-L and FA designed Mendelian disease. Nat Rev Genet. 2017;18(10):599–612. https://doi.org/10. the study and wrote the manuscript. FA initiated and led the study. All authors 1038/nrg.2017.52 read and approved the final manuscript. 14. Griffith M, Griffith OL, Coffman AC, Weible JV, McMichael JF, Spies NC, et al. DGIdb 2.0: mining clinically relevant drug–gene interactions. Nat Methods. Ethics approval and consent to participate 2013;10:1209–10. Animals were maintained at the Spanish National Cancer Research Centre 15. Chakravarty D, Gao J, Phillips SM, Kundra R, Zhang H, Wang J, et al. OncoKB: (CNIO) Animal Facility (awarded with the AAALAC accreditation) in A Precision Oncology Knowledge Base. JCO Precis Oncol. 2017; https://doi. accordance with the guidelines stated in the International Guiding Principles org/10.1200/PO.17.00011 for Biomedical Research Involving Animals, developed by the Council for 16. Tamborero D, Rubio-Perez C, Deu-Pons J, Schroeder MP, Vivancos A, Rovira International Organizations of Medical Sciences (CIOMS). Four- to six-week- A, et al. Cancer Genome Interpreter annotates the biological and clinical old female athymic nude-Foxn1nu (nu/nu) mice were purchased from Envigo relevance of tumor alterations. Genome Med. 2018;10:25. https://doi.org/10. (Barcelona, Spain). All animal experiments were approved by the Competent 1101/140475 Authority of Comunidad de Madrid (project PROEX 104/16). Patients pro- 17. Amadoz A, Sebastian-Leon P, Vidal E, Salavert F, Dopazo J. Using activation vided written informed consent before study enrollment. The inform consent status of signaling pathways as mechanism-based biomarkers to predict form was reviewed and approved by the IRB of Hospital de Madrid in ac- drug sensitivity. Sci Rep. 2015;5:18494. https://doi.org/10.1038/srep18494 cordance with the guiding principles of the Declaration of Helsinki. 18. Garnett MJ, Edelman EJ, Heidorn SJ, Greenman CD, Dastur A, Lau KW, et al. Systematic identification of genomic markers of drug sensitivity in cancer Competing interests cells. Nature. 2012;483(7391):570–5. The authors declare that they have no competing interests. 19. Basu A, Bodycombe NE, Cheah JH, Price EV, Liu K, Schaefer GI, et al. An interactive resource to identify cancer genetic and lineage dependencies targeted by small molecules. Cell. 2013;154:1151–61. Publisher’sNote 20. Iorio F,Knijnenburg TA, Vis DJ,Bignell GR,Menden MP,Schubert M, et al.A Springer Nature remains neutral with regard to jurisdictional claims in published landscape of pharmacogenomic interactions in cancer. Cell. 2016;166(3):740–54. maps and institutional affiliations. 21. Bollag G, Tsai J, Zhang J, Zhang C, Ibrahim P, Nolop K, et al. Vemurafenib: the first drug approved for BRAF-mutant cancer. Nat Rev Drug Discov. 2012; Author details 11(11):873–86. Spanish National Cancer Research Centre (CNIO), 3rd Melchor Fernandez 22. Bryant HE, Schultz N, Thomas HD, Parker KM, Flower D, Lopez E, et al. Almagro st., E-28029 Madrid, Spain. Computer Science Department - Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP- University of Vigo, Vigo, Spain. Biomedical Research Centre (CINBIO), Vigo, ribose) polymerase. Nature. 2005;434(7035):913–7. Spain. Spanish National Bioinformatics Institute (INB), Madrid, Spain. 5 6 23. De Roock W, De Vriendt V, Normanno N, Ciardiello F, Tejpar S. KRAS, BRAF, Department of Physiology - University of Valencia, Valencia, Spain. Loyola 7 PIK3CA, and PTEN mutations: implications for targeted therapies in University Chicago Stritch School of Medicine, Maywood, IL, USA. Beth Israel metastatic colorectal cancer. Lancet Oncol. 2011;12(6):594–603. Deaconess Medical Center, Boston, USA. 24. Iyer G, Hanrahan AJ, Milowsky MI, Al-Ahmadie H, Scott SN, Janakiraman M. Genome sequencing identifies a basis for everolimus sensitivity. Science. Received: 28 December 2017 Accepted: 4 May 2018 2012;338(6104):221. 25. Perini GF, Campregher PV, Ross JS, Ali S, Hamerschlak N, Santos FPS. Clinical response to everolimus in a patient with Hodgkin's lymphoma harboring a TSC2 References mutation. Blood Cancer J. 2016;6(e420) https://doi.org/10.1038/bcj.2016.25. 1. Kamalakaran S, Varadan V, Janevski A, Banerjee N, Tuck D, McCombie WR, 26. Faber AC, Li D, Song Y, Liang MC, Yeap BY, Bronson RT, et al. Differential et al. Translating next generation sequencing to practice: opportunities and induction of apoptosis in HER2 and EGFR addicted cancers following PI3K necessary steps. Mol Oncol. 2013;7(4):743–55. inhibition. Proc Natl Acad Sci U S A. 2009;106(46):19503–8. 2. Lyon GJ, Wang K. Identifying disease mutations in genomic medicine 27. Zhu Z, Aref AR, Cohoon TJ, Barbie TU, Imamura Y, Yang S, et al. Inhibition of settings: current challenges and how to accelerate progress. Genome Med. KRAS-driven tumorigenicity by interruption of an autocrine cytokine circuit. 2012;4:7–58. Cancer Discov. 2014;4(4):452–65. 3. Andre F, Mardis E, Salm M, Soria JC, Siu LL, Swanton C. Prioritizing targets 28. Butler DE, Marlein C, Walker HF, Frame FM, Mann VM, Simms MS, et al. for precision cancer medicine. Ann Oncol. 2014;25:2295–303. Inhibition of the PI3K/AKT/mTOR pathway activates autophagy and 4. Hyman DM, Taylor BS, Baselga J. Implementing genome-driven oncology. compensatory Ras/Raf/MEK/ERK signalling in prostate cancer. Oncotarget. Cell. 2017;168(4):584–99. 2017;8(34):56698–713. 5. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA Jr, Kinzler KW. 29. Lee HJ, Zhuang G, Cao Y, Du P, Kim HJ, Settleman J. Drug resistance via Cancer genome landscapes. Science. 2013;339(6127):1546–58. feedback activation of Stat3 in oncogene-addicted cancer cells. Cancer Cell. 6. Van Allen EM, Wagle N, Stojanov P, Perrin DL, Cibulskis K, Marlow S, et al. 2014;26(2):207–21. Whole-exome sequencing and clinical interpretation of formalin-fixed, 30. Forbes SA, Beare D, Boutselakis H, Bamford S, Bindal N, Tate J, et al. COSMIC: paraffin-embedded tumor samples to guide precision cancer medicine. Nat somatic cancer genetics at high-resolution. Nucleic Acids Res. 2017;45(D1): Med. 2014;20(6):682–8. D777–83. 7. Martincorena I, Raine KM, Gerstung M, Dawson KJ, Haase K, Van Loo P, et al. 31. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, Golub TR, Universal patterns of selection in cancer and somatic tissues. Cell. 2017; et al. Discovery and saturation analysis of cancer genes across 21 tumour 171(5):1029–41. types. Nature. 2014;505(7484):495–501. https://doi.org/10.1038/nature12912 8. Cheng J, Demeulemeester J, Wedge DC, Vollan HKM, Pitt JJ, Russnes HG, 32. Tamborero D, Gonzalez-Perez A, Perez-Llamas C, Deu-Pons J, Kandoth C, et al. Pan-cancer analysis of homozygous deletions in primary tumours Reimand J, et al. Comprehensive identification of mutational cancer driver uncovers rare tumour suppressors. Nat Commun. 2017;1:1221. genes across 12 tumor types. Sci Rep. 2013;3:2650. 9. Peterson TA, Gauran IIM, Park J, Park D, Kann MG. Oncodomains: a protein 33. Schlicker A, Michaut M, Rahman R, Wessels LF. OncoScape: Exploring domain-centric framework for analyzing rare variants in tumor samples. PLoS the cancer aberration landscape by genomic data fusion. Sci Rep. Comput Biol. 2017;4:e1005428. https://doi.org/10.1371/journal.pcbi.1005428 2016;6:28103. 10. Wei CH, Phan L, Feltz J, Maiti R, Hefferon T, Lu Z. tmVar 2.0: integrating genomic variant information from literature with dbSNP and ClinVar for 34. Cowley GS, Weir BA, Vazquez F, Tamayo P, Scott JA, Rusin S, et al. Parallel precision medicine. Bioinformatics. 2018;34:80–7. https://doi.org/10.1093/ genome-scale loss of function screens in 216 cancer cell lines for the bioinformatics/btx541 identification of context-specific genetic dependencies. Sci Data. 2014;1: 11. Yang H, Robinson PN, Wang K. Phenolyzer: phenotype-based prioritization 140035. of candidate genes for human diseases. Nat Methods. 2015;12(9):841–3. 35. Shao DD, Tsherniak A, Gopal S, Weir BA, Tamayo P, Stransky N, et al. ATARiS: 12. Ghosh R, Oak N, Plon SE. Evaluation of in silico algorithms for use with ACMG/ computational quantification of gene suppression phenotypes from AMP clinical variant interpretation guidelines. Genome Biol. 2017;18:225. multisample RNAi screens. Genome Res. 2013;23(4):665–78. Piñeiro-Yáñez et al. Genome Medicine (2018) 10:41 Page 11 of 11 36. Salavert F, Hidalgo MR, Amadoz A, Çubuk C, Medina I, Crespo D, et al. 62. Tsherniak A, Vazquez F, Montgomery PG, Weir BA, Kryukov G, Cowley GS, Actionable pathways: interactive discovery of therapeutic targets using et al. Defining a cancer dependency map. Cell. 2017;170(3):564–76. signaling pathway models. Nucleic Acids Res. 2016;44(W1):W212–6. 63. McDonald ER, de Weck A, Schlabach MR, Billy E, Mavrakis KJ, Hoffman GR, 37. Workman P, Al-Lazikani B. Drugging cancer genomes. Nat Rev Drug Discov. et al. Project DRIVE: a compendium of cancer dependencies and synthetic 2013;12(12):889–90. lethal relationships uncovered by large-scale, deep RNAi screening. Cell. 38. Editorial. It's all druggable. Nat Genet. 2017;49(2):169. https://doi.org/10. 2017;170(3):577–92. 1038/ng.3788. 64. Ivanov AA, Revennaugh B, Rusnak L, Gonzalez-Pecchi V, Mo X, Johns MA, et al. The OncoPPi Portal: an integrative resource to explore and prioritize 39. Engelman JA, Zejnullahu K, Mitsudomi T, Song Y, Hyland C, Park J, et al. protein-protein interactions for cancer target discovery. Bioinformatics. 2018; MET amplification leads to gefitinib resistance in lung cancer by activating 34:1183–91. https://doi.org/10.1093/bioinformatics/btx743 ERBB3 signaling. Science. 2007;316(5827):1039–43. 65. Garcia-Alonso L, Iorio F, Matchan A, Fonseca N, Jaaks P, Peat G, et al. 40. Turke AB, Zejnullahu K, Wu YL, Song Y, Dias-Santagata D, Lifshits E, et al. Transcription factor activities enhance markers of drug sensitivity in cancer. Preexistence and clonal selection of MET amplification in EGFR mutant Cancer Res. 2018;78:3,769–80. NSCLC. Cancer Cell. 2010;17(1):77–88. 66. Schubert M, Klinger B, Klünemann M, Sieber A, Uhlitz F, Sauer S, et al. 41. Gainor JF, Niederst MJ, Lennerz JK, Dagogo-Jack I, Stevens S, Shaw AT, et al. Perturbation-response genes reveal signaling footprints in cancer gene Dramatic response to combination erlotinib and crizotinib in a patient with expression. Nat Commun. 2018;9(1):20. advanced, EGFR-mutant lung cancer harboring De Novo MET amplification. 67. Sharp ME. Toward a comprehensive drug ontology: extraction of drug- J Thorac Oncol. 2016;11(7):83–5. indication relations from diverse information sources. J Biomed Semantics. 42. Tricker EM, Xu C, Uddin S, Capelletti M, Ercan D, Ogino A, et al. Combined 2017;8(1):2. EGFR/MEK inhibition prevents the emergence of resistance in EGFR-Mutant lung cancer. Cancer Discov. 2015;5(9):960–71. 43. Ortiz-Cuaran S, Scheffler M, Plenker D, Dahmen L, Scheel AH, Fernandez-Cuesta L, et al. Heterogeneous mechanisms of primary and acquired resistance to third-generation EGFR inhibitors. Clin Cancer Res. 2016;22(19):4837–47. 44. Kandoth C, McLellan MD, Vandin F, Ye K, Niu B, Lu C, et al. Mutational landscape and significance across 12 major cancer types. Nature. 2013; 502(7471):333–9. https://doi.org/10.1038/nature12634 45. Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R, et al. A census of human cancer genes. Nat Rev Cancer. 2004;4(3):177–83. 46. Dienstmann R, Jang IS, Bot B, Friend S, Guinney J. Database of genomic biomarkers for cancer drugs and clinical targetability in solid tumors. Cancer Discov. 2015;5(2):118–23. 47. Rubio-Perez C, Tamborero D, Schroeder MP, Antolín AA, Deu-Pons J, Perez- Llamas C, et al. In silico prescription of anticancer drugs to cohorts of 28 tumor types reveals targeting opportunities. Cancer Cell. 2015;27(3):382–96. 48. Gohlke BO, Nickel J, Otto R, Dunkel M, Preissner R. CancerResource–updated database of cancer-relevant proteins, mutations and interacting drugs. Nucleic Acids Res. 2016;44(D1):D932–7. 49. Patterson SE, Liu R, Statz CM, Durkin D, Lakshminarayana A, Mockus SM. The clinical trial landscape in oncology and connectivity of somatic mutational profiles to targeted therapies. Hum Genomics. 2016;10:4. 50. Goldman M, Zhang J, Fonseca NA, Xiang Q, Craft B, Piñeiro-Yáñez E, et al. Online resources for PCAWG data exploration, visualization, and discovery. https://www.biorxiv.org/content/early/2017/10/20/163907 51. Garralda E, Paz K, López-Casas PP, Jones S, Katz A, Kann LM, et al. Integrated next-generation sequencing and avatar mouse models for personalized cancer treatment. Clin Cancer Res. 2014;20(9):2476–84. 52. Dhillon AS, Hagan S, Rath O, Kolch W. MAP kinase signalling pathways in cancer. Oncogene. 2007;26(22):3279–90. 53. Mendoza MC, Er EE, Blenis J. The Ras-ERK and PI3K-mTOR pathways: cross- talk and compensation. Trends Biochem Sci. 2011;36(6):320–8. 54. Gómez-López G, Dopazo J, Cigudosa JC, Valencia A, Al-Shahrour F. Precision medicine needs pioneering clinical bioinformaticians. Brief Bioinform. 2017; https://doi.org/10.1093/bib/bbx144. 55. Dagogo-Jack I, Shaw AT. Tumour heterogeneity and resistance to cancer therapies. Nat Rev Clin Oncol. 2018:15, 81–94. https://doi.org/10.1038/ nrclinonc.2017.166 56. Garraway LA, Verweij J, Ballman KV. Precision oncology: an overview. J Clin Oncol. 2013;31(15):1803–5. 57. Califano A, Alvarez MJ. The recurrent architecture of tumour initiation, progression and drug sensitivity. Nat Rev Cancer. 2017;17(2):116–30. 58. Naulaerts S, Dang CC, Ballester PJ. Precision and recall oncology: combining multiple gene mutations for improved identification of drug-sensitive tumours. Oncotarget. 2017;8:97025–40. 59. Schneider G, Schmidt-Supprian M, Rad R, Saur D. Tissue-specific tumorigenesis: context matters. Nat Rev Cancer. 2017;17(4):239–53. https://doi.org/10.1038/nrc.2017.5. 60. Mina M, Raynaud F, Tavernari D, Battistello E, Sungalee S, Saghafinia S, et al. Conditional selection of genomic alterations dictates cancer evolution and oncogenic dependencies. Cancer Cell. 2017;32(2):155–68. 61. Letai A. Functional precision cancer medicine-moving beyond pure genomics. Nat Med. 2017;23(9):1028–35.

Journal

Genome MedicineSpringer Journals

Published: May 31, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off