TY - JOUR
AU - Kumar, Manoj
AB - Abstract Allele-specific siRNAs (ASP-siRNAs) have emerged as promising therapeutic molecules owing to their selectivity to inhibit the mutant allele or associated single-nucleotide polymorphisms (SNPs) sparing the expression of the wild-type counterpart. Thus, a dedicated bioinformatics platform encompassing updated ASP-siRNAs and an algorithm for the prediction of their inhibitory efficacy will be helpful in tackling currently intractable genetic disorders. In the present study, we have developed the ASPsiRNA resource (http://crdd.osdd.net/servers/aspsirna/) covering three components viz (i) ASPsiDb, (ii) ASPsiPred, and (iii) analysis tools like ASP-siOffTar. ASPsiDb is a manually curated database harboring 4543 (including 422 chemically modified) ASP-siRNAs targeting 78 unique genes involved in 51 different diseases. It furnishes comprehensive information from experimental studies on ASP-siRNAs along with multidimensional genetic and clinical information for numerous mutations. ASPsiPred is a two-layered algorithm to predict efficacy of ASP-siRNAs for fully complementary mutant (Effmut) and wild-type allele (Effwild) with one mismatch by ASPsiPredSVM and ASPsiPredmatrix, respectively. In ASPsiPredSVM, 922 unique ASP-siRNAs with experimentally validated quantitative Effmut were used. During 10-fold cross-validation (10nCV) employing various sequence features on the training/testing dataset (T737), the best predictive model achieved a maximum Pearson’s correlation coefficient (PCC) of 0.71. Further, the accuracy of the classifier to predict Effmut against novel genes was assessed by leave one target out cross-validation approach (LOTOCV). ASPsiPredmatrix was constructed from rule-based studies describing the effect of single siRNA:mRNA mismatches on the efficacy at 19 different locations of siRNA. Thus, ASPsiRNA encompasses the first database, prediction algorithm, and off-target analysis tool that is expected to accelerate research in the field of RNAi-based therapeutics for human genetic diseases. allele-specific siRNA, ASPsiDb, ASPsiPred, genetic disease database, prediction algorithm RNA interference (RNAi) is an evolutionarily conserved phenomenon to inhibit gene expression in eukaryotes including mammals (Fire et al. 1998; Paulson and Gonzalez-Alegre 2006). One of the most important implications of RNAi technology is the development of potent and highly effective siRNAs imparting exquisite specificity (Keiser et al. 2015). They have already been utilized as a vital research tool for loss-of-function studies and the suppression of phenotypes generated by dominantly acting mutant genes (Rodriguez-Lebron and Paulson 2006). Thus, siRNA-mediated selective suppression of dominantly inherited mRNA transcripts holds curative potential for gain-of-function human genetic diseases (Lopes et al. 2016; Loy et al. 2012). In this context, allele-specific RNAi (ASP-RNAi) is an innovative category of RNAi with the objective of suppressing the dominant mutant allele while sparing expression of the corresponding normal allele with the specificity of single-nucleotide differences between the two (Gonzalez-Alegre 2007). Therefore, allele-specific siRNAs (ASP-siRNAs) are potentially a novel and better remedial alternative for the treatment of autosomal dominant genetic disorders especially in cases where wild-type allele expression is crucial for organism survival (Miller et al. 2003). The mechanism of ASP-RNAi gene silencing is illustrated in Figure 1. Figure 1 Open in new tabDownload slide Mechanistic representation of ASP-RNAi. Numerous studies have been conducted to assess the potency and specificity of ASP-siRNAs for various neurodegenerative disorders like Huntington disease (HD) (Drouet et al. 2014; Miniarikova et al. 2016), DYT1 dystonia (Gonzalez-Alegre et al. 2003, 2005), Alzheimer’s disease (Sierant et al. 2011), Parkinson’s disease (PD) (Takahashi et al. 2015), amyloid lateral sclerosis (ALS) (Schwarz et al. 2006), and Machado–Joseph disease (Alves et al. 2008). Their therapeutic potential has also been assessed for various skin disorders like epidermolysis bullosa simplex (Atkinson et al. 2011), epidermolytic palmoplantar keratoderma (EPPK) (Lyu et al. 2016), and lattice corneal dystrophy type I (LCDI) (Courtney et al. 2014). They have also been utilized to suppress the mutations associated with other diseases like cancer (Iyer et al. 2016), viral diseases (Teng et al. 2011), and sex-linked disorders (Caplen et al. 2002). Various in-vivo studies have been reported in different animal models, for e.g., HD (Drouet et al. 2014), EPPK (Miniarikova et al. 2016), and hyper-trophic cardiomyopathy (Miniarikova et al. 2016). The potential of this therapeutic modality has been studied in human embryonic stem cells (Miniarikova et al. 2016), and allele-specific gene silencing (ASGS) approaches have started to move from the laboratory to the clinic (Liu et al. 2016). The first ASP-siRNA TD101 for the human skin disorder pachyonychia congenita (PC) has entered into phase1b clinical trials (Leachman et al. 2008). Currently there is no cure available for dominant negative genetic maladies (Squitieri and de Yebenes 2015). Although, a few symptomatic pharmacological and nonpharmacological drugs have been used in clinical practice (Marelli and Maschat 2016), they were aimed at temporary relief and delay of disease progression (Jamwal and Kumar 2015; Kulshreshtha and Piplani 2016; LeWitt et al. 2016). Similarly peptide-based drugs have been used to suppress the aggregate formation of toxic mutant protein (Aharony et al. 2015; Arribat et al. 2013). However, it is reported that indiscriminate sustained suppression at the protein level may have harmful effects on the cell (Rodriguez-Lebron and Paulson 2006), and they are not aimed at disease reversal. Likewise, traditional antisense molecules are also candidates for mutant-specific suppression (Pandey et al. 2015). However, the one-to-one ratio of binding to target requires high concentrations of these molecules in the cell, which may result in toxic situations (Allen et al. 2013). On the other hand, ASP-siRNAs exhibit multiplicity i.e., a single siRNA can cause cleavage of multiple copies of the target mRNA (Allen et al. 2013). Moreover, antisense molecules exhibit irreversible binding to their target making them poor candidates for ASP-RNAi, especially when the system demands one nucleotide discrimination (Allen et al. 2013). Antisense Oligonucleotide (ASO), being single stranded, is unstable and less potent, thus requiring high concentrations and, consequently, leading to off-target effects more severe than dsRNA (Watts and Corey 2012). Despite unprecedented specificity and immense therapeutic utility of ASP-siRNAs, bioinformatics repositories in the field are lacking. Although there are several resources available for siRNAs like siRECORDS (Ren et al. 2006), HusiDa (Truss et al. 2005), HIVsirDB (Tyagi et al. 2011), VIRsiRNAdb (Thakur et al. 2012b), siRNAmod (Dar et al. 2016b), and RNAiAtlas (Mazur et al. 2012), they lack information related to ASP-siRNAs (Supplemental Material, Table S1 in File S1). Likewise, there are numerous algorithms (Ahmed and Raghava 2011; Dar et al. 2016a; Filhol et al. 2012; Huesken et al. 2005; Kaur et al. 2016; McQuisten and Peek 2009; Mysara et al. 2011; Pan et al. 2011; Peek 2007; Qureshi et al. 2013; Saetrom 2004; Shabalina et al. 2006; Vert et al. 2006) and design rules (Amarzguioui and Prydz 2004; Elbashir et al. 2001a; Reynolds et al. 2004; Ui-Tei et al. 2004) for siRNA efficacy prediction. But, none of the available web servers was dedicated to predicting two efficacies associated with a single siRNA. This prompted us to develop ASPsiRNA, a web resource offering multiple modules. The first module, ASPsiDb, delivers updated and manually curated ASP-siRNA sequences targeted against human genetic diseases available in the literature, coupled with clinicopathogenic information about various mutations and the annotation of genes. In the second module ASPsiPred, using data from the database, we have developed a two-layered algorithm for prediction of inhibitory efficacy of ASP-siRNA for mutant and wild-type alleles. We have provided Support Vector Machine (SVM) and matrix-based algorithms for the prediction of the efficacy of ASP-siRNA for both diseased (Effmut) and wild-type alleles (Effwild). This algorithm is aimed to help experimental biologists in designing optimum allele discriminatory siRNAs along with minimum off-targets. In the third module, we have integrated useful analysis tools like ASP-siOffTar (seed and full sequence based), BLAST, and ASP-siMAP. Materials and Methods ASPsiDb database development Data collection: Information extraction was primarily divided into four parallel data systems (Supplemental Methods Section I and II in File S1): (a) ASP-siRNA data extraction: An extensive literature search was executed to obtain articles indexed in PubMed using the following combination of keywords (((Allele)) AND (((((((sirna) OR shrna) OR small interfering RNA) OR short interfering RNA) OR RNA interference) OR RNAi) OR silencing)) AND (((specific) OR mismatch*) OR discrimination). Patents pertaining to ASP-siRNAs were extracted from “The Lens” (www.lens.org). (b) Clinical information regarding various mutations: Clinical data associated with different mutations were mined from ClinVar (Landrum et al. 2014), dbVar (Lappalainen et al. 2013), dbSNP (Sherry et al. 1999), and OMIM (Hamosh et al. 2000). (c) Annotation of genes targeted by ASP-siRNAs: It involves standard nomenclature of every gene from HGNC (HUGO Gene Nomenclature Committee), cytogenic/chromosomal coordinates of a gene from UniProt, UCSC genome browser. (d) Molecular/biological/genetic information regarding diverse human genes and corresponding diseases: Information about the genetic basis of disorders was compiled from various resources; e.g., OMIM, ClinVar, and KEGG disease modules. Database schema: Database content is systematically organized to provide easy access of ASP-siRNAs data coupled with comprehensive information of clinical and genetic data. It is maintained using MySQL and launched on Apache HTTP Server installed on an IBM machine under Red Hat Enterprise Linux5 background. The responsive front end was implemented with CSS, PHP, HTML5, and JavaScript as employed in our previous resources (Qureshi et al. 2014). Detailed architecture of the resource is depicted in Figure 2. Figure 2 Open in new tabDownload slide ASPsiRNA architecture. ASPsiDb web interface: searching and browsing: Proficient searching and browsing is provided in the resource “Search” section that provides three suboptions for convenient data mining in the database, i.e., (i) keyword search, (ii) literature search, and (iii) sequence mapping based search (Figure S1 in File S1). Additionally, we have also offered database browsing in six categories: disease, gene, mutation, cell line, mismatch, and Pubmed ID (Supplemental Methods Section III in File S1). The output of the searching and browsing page provides a list of ASP-siRNAs matching the input query. By clicking on the individual ASP-siRNA ID, the user can get complete details of the respective entry structured in nine modules (Supplemental Methods Section IV and Figures S2–S5 in File S1). ASPsiPred: prediction algorithm development Dataset preparation: Since designing effective and discriminatory ASP-siRNAs is associated with two efficacy values, i.e., one for a fully complementary target allele and a second for the nontarget allele, we have integrated a two-tiered algorithm in ASPsiPred (ASPsiPredSVM and ASPsiPredmatrix) to predict Effmut and Effwild, respectively (Figure 3). Figure 3 Open in new tabDownload slide Computational workflow employed to extract ASP-siRNAs and developing the algorithm for the prediction of inhibitory efficacy: left arm describes the development of the SVM-based algorithm (ASPsiPredSVM) for prediction of efficacy for fully complementary mutant allele (Effmut), while the right arm depicts the process of making ASPsiPredmatrix for the prediction of the efficacy for wild-type allele having one mismatch (Effwild). In the first layer, i.e., ASPsiPredSVM, we have screened ASPsiDb with 4543 ASP-siRNAs to get a unique and representative working dataset. After removing the 422 chemically modified (cm) ASP-siRNAs, we have processed the remaining 4121 sequences to extract 922 nonredundant 19mer siRNA sequences with quantitative efficacies (D922) (Supplemental Methods Section V and Table S2 in File S1). From D922, we have randomly extracted 185 sequences as independent/validation datasets (V185), while the remaining 737 sequences were used for the 10-fold cross-validation (10nCV) training/testing datasets (T737) (Tables S3 and S4 in File S1). This process was repeated five times to generate five training/testing and external validation sets. Features used for model development: Nucleotide composition and position-related features, thermodynamic stability and secondary structure based features were used in this study (see Supplemental Methods Section VII in File S1). We have selected these models/features and applied 10nCV on these sets. Once we obtained optimal results on selected hyper-parameters, we applied 10nCV on the full T922 dataset as a final classifier (Table S4 in File S1). Algorithm development and validation: The SVMlight (http://svmlight.joachims.org) software package was used to train the different siRNA features and develop predictive models using 10nCV. In this study, we have used the radial basis function kernel for development of ASPsiPredSVM. We have evaluated the performance of our models using the Pearson correlation coefficient (PCC) (Supplemental Methods Section VIII and IX in File S1). For the prediction of Effwild, i.e., the efficacy to inhibit target sequences with one mismatch, we have developed ASPsiPredmatrix (Tables S5–S8 in File S1) utilizing data from the following articles (Birmingham et al. 2006; Huang et al. 2009; Ohnishi et al. 2008; Schwarz et al. 2006) (Supplemental Methods Section X in File S1). Implementation of ASPsiPred webserver: ASPsiPred was developed on a SUN server using PERL, HTML, and CGI-PERL (Qureshi et al. 2013; Thakur et al. 2012a). Upon clicking ASPsiPred, a user is asked to enter the target and wild-type allele in FASTA format with the nucleotide mutation in lower case. For user convenience, we have provided a clickable example sequence. Our tool will generate ASP-siRNAs against mutation at all possible 19 locations followed by the prediction of Effmut and Effwild using ASPsiPredSVM and ASPsiPredmatrix. We have integrated the ASP-siOffTar tool on the output page to provide seed-based off-targets for all predicted 19 ASP-siRNAs against user-provided mutation. This will give an idea about the potency as well as specificity of ASP-siRNA (Figure 4A). Thus, a user can select optimal allele-differentiating siRNAs with minimum off-target effects. The result is also displayed in a graphical format to analyze at which position ASP-siRNA displays relatively high discrimination for both alleles (Figure 4B). Figure 4 Open in new tabDownload slide Description of ASPsiPred web server with result output. (A) Screenshot demonstrating ASP-siRNAs generated against a T > G mutation at all possible 19 positions along with Effmut and Effwild predicted from ASPsiPredSVM and ASPsiPredmatrix, respectively. Their relative difference between the two efficacies is also displayed along with the prediction of seed-based off-targets for all 19 ASP-siRNAs. (B) The output of the Effmut and Effwild of 19 ASP-siRNAs in graphical form. Analysis tools ASP-siOffTar (seed based): This provides a list of off-targets based on the alignment of hexamer (2–7) or heptamer (2–8) seed regions of ASP-siRNA or any siRNA on the human genome (build GRCh37). Since off targeting is majorly associated with the presence of perfectly complementary 3′-UTR matches with the seed region of the antisense strand of the siRNA (Birmingham et al. 2006), we have not allowed any mismatch in the alignment of seed regions on the human genome (Figure S6 in File S1). ASP-siOffTar (full sequence based): Full sequence based off-targets are also integrated as a separate tool on the web interface with a maximum of three allowed mismatches (Figure S7 in File S1). ASP-siRNA-BLAST: This matches a user-provided siRNA sequence against the ASPsiRNA database to find out whether similar siRNA/s are already reported. ASP-siMAP: Experimental biologists who seek to design an ASP-siRNA on their target gene can take advantage of the ASP-siMAP tool. It simply maps ASP-siRNAs reported in our archive to a user-specified target gene along with its start position. Data availability All the data necessary for the results and conclusions in this paper are provided in the article or ASPsiRNA repository (http://crdd.osdd.net/servers/aspsirna/). Results ASPsiDb Database statistics: ASPsiDb is a manually curated and highly annotated depository of 4543 experimentally validated ASP-siRNA entries including 422 chemically modified (cm) ASP-siRNAs affecting 78 unique genes causing 51 various diseases out of which hemolytic uremic syndrome, HD, ALS, cancer, and PD were the top five diseases targeted (Figure S8a in File S1). Likewise, the CD46 gene followed by HTT, SOD1, DBI, and PPIB were the top five genes (Figure S8b in File S1). ASP-siRNAs were transfected using diverse transfection reagents; out of these lipofectamine 2000 was the most commonly used. Among the various methods reported to deliver ASP-siRNAs to the target locus, transfection (87.80%) was the major delivery method followed by shRNA expression vector (19.85%), lentiviral vector (1.66%), electroporation (1.38%), stereotaxic injection (0.76%), atelocollagen (0.57%) mediated delivery, and other methods (0.42%) (Figure S9 in File S1). The efficacy of various ASP-siRNAs was determined using 45 different cell-lines, among them HEK followed by HeLa, fibroblast, AD293, DU145, and HaCaT were most frequently used (Figure 5A). Animal models were also employed for in vivo studies including the transgenic mouse model, male Wistar rat, and Caenorhabditis elegans, out of which the mouse model was most common. In a particular study, human plantar calluses were also used to assess the potency of ASP-siRNA TD-101 targeting PC in a phase1b clinical trial (Leachman et al. 2010). Both RNA and protein level experimental methods were used for evaluating the efficacy; however DLRA (dual luciferase reporter assay) was reported in the majority of studies followed by western blot, RT-PCR, fluorescence microscopy, and microarray (Figure 5B). Figure 5 Open in new tabDownload slide ASPsiRNA database statistics. (A and B) Pie charts exemplifying the distribution of cell lines and experimental methods used for validation of ASP-siRNAs. (C and D) Bar graphs describing the percentage coverage of different categories of genetic diseases and the statistical distribution of the clinical significance of diverse types of gene variants reported in the archive, respectively; described in ASPsiRNA. Dominant genetic disorders are ideal candidates of ASGS due to its capability to target mutant alleles selectively. Our resource covers these disorders from seven different categories namely neurological disorders (ND) (51%), followed by skin (16%), skeletal (10%), cancer (5%), muscular disorders (4%), autoimmune diseases (3%), and others (11%) as depicted in Figure 5C. For the design of effective and specific ASP-siRNAs, we have to select such an siRNA that causes least harm to the wild-type allele while keeping the mutant allele inhibition at the maximum level and displaying optimum allele discrimination (Davidson and Paulson 2004). Therefore, to analyze and find the discriminatory siRNAs, we have plotted the Effmutvs. the Effwild efficacies in the form of a scatter plot (Figure S10 in File S1). Statistical inspection reveals that the lower right section of the plot is quite dense as compared to the other quartiles. This section represents a high Effmut but low Effwild. Thus, these sequences exhibit experimentally validated allelic discrimination most helpful for experimental biologists to target specific mutant alleles. Statistical analysis of gene variants/mutations: We have analyzed the pathogenic status of various gene variants/mutations and found that ∼64% of ASP-siRNAs target pathogenic mutations (Figure 5D). We have also sketched all mutations and their associated molecular changes collected from ClinVar in the form of 3D-line graphs represented in Figure 6. It shows the statistical distribution of different sequence variations such as single-nucleotide variation (snv), microsatellite (expansion mutations), deletion (del), copy number gain (CNG), and insertion-deletion (InDel), which are associated with molecular consequences like missense mutation, frame shift variation (fsv), synonymous mutation, and 3′-UTR variant (variation in 3′ UTR region). Investigation of the graph indicates that: (i) in siRNAs targeting snvs, the molecular consequence is missense mutation in ∼98% of the cases; (ii) similarly, siRNAs targeting deletion variants cause fsv in ∼98% of cases; and (iii) siRNAs targeting microsatellite mutations mostly have a tendency to show fsv and missense mutations. Figure 6 Open in new tabDownload slide Different mutations and molecular consequences represented by 3D line-graph. fsv, frame shift variation; del, deletion; CNG, copy number gain; Indel, insertion-deletion. A mutational landscape was summarized to investigate all gene variants/mutations examined by ASP-siRNAs with the help of circos plot (Krzywinski et al. 2009). It shows that ASP-siRNAs mostly target genes that had single-nucleotide substitutions (SNPs) and missense mutations (Figure 7). This observation is in accordance with the Human Genome Database (HGDB), which states that out of 73,411 reported mutations responsible for causing genetic diseases, >60% are caused by SNPs (Seyhan 2011). Figure 7 Open in new tabDownload slide Mutational landscape of different genes described in ASPsiRNA epitomized by a circos plot: left and right hemi circle represents the mutation categories and gene names, respectively. The length of the main circular segments is proportional to the total number of ASP-siRNAs belonging to that segment, while the width of the ribbon connecting the gene with the mutation represents the proportion of ASP-siRNA sequences belonging to the particular mutation type. The two outer rings are contribution tracks, i.e., stacked bar plots with a gradient of color signifying the proportion of entries from different genes. ASPsiPred performance evaluation ASPsiPredSVM: performance during 10nCV: Selected sequence features (mdtt+binary) (see Supplemental Methods Section VII in File S1) were used to perform 10nCV on five random training/testing sets (T737). Their performance was measured on an independent validation dataset (V185) (Table S3 in File S1). After confirming that all five sets performed approximately similarly, we have selected Random Set-2 to build final classifier without any bias (random set-2). During 10nCV on the selected set, predictive models based upon sequence composition based features like mono-, di-, tri-, tetra-, and penta-nucleotide compositions achieved a maximum correlation of 0.53, 0.68, 0.70, 0.69, and 0.68, respectively. Position-based features like the binary pattern of nucleotides attained a PCC of 0.55. We have also developed hybrid models using >1 nucleotide features as input, e.g., hybrid of mono- (m) and dinucleotide (d) composition (md). We achieved correlations of 0.67, 0.70, 0.71, 0.71, 0.71, and 0.71 in the md, mdt, mdtt, mdttp, mdtt + binary, and mdttp + binary hybrid models, respectively (see Table 1). Accordingly, performance of thermodynamic and secondary structure based features achieved a PCC of 0.41 and 0.24, respectively; however, their hybrid with our best model did not lead to an improvement in correlations (Table 1, model 12+13, 12+14, and 12+13+14). The sequence features, which performed best on set-2, i.e., ASPsiPredSVM (mdtt + binary), were applied to the total dataset (D922) as a final classifier on the webserver termed as ASPsiPredSVM# (Table S4 in File S1). Performance of different predictive models on the training/testing dataset of 737 sequences (T737) during 10-fold cross-validation. Evaluation of the models on an independent validation dataset (V185) Table 1 Performance of different predictive models on the training/testing dataset of 737 sequences (T737) during 10-fold cross-validation. Evaluation of the models on an independent validation dataset (V185) . . . PCC on Training/Testing Sets (T737) and Independent Validation Sets (V185) Using 10nCV . Predictive Model No. . siRNA Feature Name . No. of Features . T737 . V185 . 1 Mononucleotide composition 4 0.53 0.54 2 Dinucleotide composition 16 0.68 0.64 3 Trinucleotide composition 64 0.70 0.66 4 Tetranucleotide composition 256 0.69 0.65 5 Pentanucleotide composition 1024 0.68 0.63 6 Binary 76 0.55 0.56 7 1+2 20 0.67 0.63 8 1+2+3 84 0.70 0.63 9 1+2+3+4 340 0.71 0.65 10 1+2+3+4+5 1364 0.71 0.65 11 1+2+3+4+6 (ASPsiPredSVM) 416 0.71 0.65 12 1+2+3+4+5+6 1440 0.71 0.65 13 Thermodynamic feature 21 0.41 0.30 14 Secondary structure 19 0.24 0.07 15 13+14 40 0.35 0.23 16 12+13 437 0.71 0.65 17 12+14 435 0.71 0.65 18 12+13+14 456 0.71 0.65 19 ASPsiPredmatrix Matrix based Developed on rules-based studies 0.63 . . . PCC on Training/Testing Sets (T737) and Independent Validation Sets (V185) Using 10nCV . Predictive Model No. . siRNA Feature Name . No. of Features . T737 . V185 . 1 Mononucleotide composition 4 0.53 0.54 2 Dinucleotide composition 16 0.68 0.64 3 Trinucleotide composition 64 0.70 0.66 4 Tetranucleotide composition 256 0.69 0.65 5 Pentanucleotide composition 1024 0.68 0.63 6 Binary 76 0.55 0.56 7 1+2 20 0.67 0.63 8 1+2+3 84 0.70 0.63 9 1+2+3+4 340 0.71 0.65 10 1+2+3+4+5 1364 0.71 0.65 11 1+2+3+4+6 (ASPsiPredSVM) 416 0.71 0.65 12 1+2+3+4+5+6 1440 0.71 0.65 13 Thermodynamic feature 21 0.41 0.30 14 Secondary structure 19 0.24 0.07 15 13+14 40 0.35 0.23 16 12+13 437 0.71 0.65 17 12+14 435 0.71 0.65 18 12+13+14 456 0.71 0.65 19 ASPsiPredmatrix Matrix based Developed on rules-based studies 0.63 PCC, Pearson correlation coefficient; 10nCV, 10-fold cross-validation; T737, training/testing dataset for 10-fold cross-validation; V185, independent validation dataset. PCC is between actual and observed Effmut. Training/testing dataset is used to train different predictive models, while independent validation dataset was not used anywhere during training/testing of algorithm. Open in new tab Table 1 Performance of different predictive models on the training/testing dataset of 737 sequences (T737) during 10-fold cross-validation. Evaluation of the models on an independent validation dataset (V185) . . . PCC on Training/Testing Sets (T737) and Independent Validation Sets (V185) Using 10nCV . Predictive Model No. . siRNA Feature Name . No. of Features . T737 . V185 . 1 Mononucleotide composition 4 0.53 0.54 2 Dinucleotide composition 16 0.68 0.64 3 Trinucleotide composition 64 0.70 0.66 4 Tetranucleotide composition 256 0.69 0.65 5 Pentanucleotide composition 1024 0.68 0.63 6 Binary 76 0.55 0.56 7 1+2 20 0.67 0.63 8 1+2+3 84 0.70 0.63 9 1+2+3+4 340 0.71 0.65 10 1+2+3+4+5 1364 0.71 0.65 11 1+2+3+4+6 (ASPsiPredSVM) 416 0.71 0.65 12 1+2+3+4+5+6 1440 0.71 0.65 13 Thermodynamic feature 21 0.41 0.30 14 Secondary structure 19 0.24 0.07 15 13+14 40 0.35 0.23 16 12+13 437 0.71 0.65 17 12+14 435 0.71 0.65 18 12+13+14 456 0.71 0.65 19 ASPsiPredmatrix Matrix based Developed on rules-based studies 0.63 . . . PCC on Training/Testing Sets (T737) and Independent Validation Sets (V185) Using 10nCV . Predictive Model No. . siRNA Feature Name . No. of Features . T737 . V185 . 1 Mononucleotide composition 4 0.53 0.54 2 Dinucleotide composition 16 0.68 0.64 3 Trinucleotide composition 64 0.70 0.66 4 Tetranucleotide composition 256 0.69 0.65 5 Pentanucleotide composition 1024 0.68 0.63 6 Binary 76 0.55 0.56 7 1+2 20 0.67 0.63 8 1+2+3 84 0.70 0.63 9 1+2+3+4 340 0.71 0.65 10 1+2+3+4+5 1364 0.71 0.65 11 1+2+3+4+6 (ASPsiPredSVM) 416 0.71 0.65 12 1+2+3+4+5+6 1440 0.71 0.65 13 Thermodynamic feature 21 0.41 0.30 14 Secondary structure 19 0.24 0.07 15 13+14 40 0.35 0.23 16 12+13 437 0.71 0.65 17 12+14 435 0.71 0.65 18 12+13+14 456 0.71 0.65 19 ASPsiPredmatrix Matrix based Developed on rules-based studies 0.63 PCC, Pearson correlation coefficient; 10nCV, 10-fold cross-validation; T737, training/testing dataset for 10-fold cross-validation; V185, independent validation dataset. PCC is between actual and observed Effmut. Training/testing dataset is used to train different predictive models, while independent validation dataset was not used anywhere during training/testing of algorithm. Open in new tab Performance on independent validation dataset (V185): The performance of the predictive models was assessed on V185. Our best model achieved a maximum (PCC) of 0.71 during 10nCV on the training dataset (T737) termed as ASPsiPredSVM. On V185, a comparable PCC of 0.65 was obtained (Table 1). Scatter plots depicting the correlation between the actual and predicted efficacy during 10nCV and independent validation are shown as Figures S11 and S12 in File S1. Performance during leave one target out cross-validation (LOTOCV): Since D922 contains sequences having single-nucleotide sliding difference (see more in Supplemental Methods Section VI in File S1), a simple 10nCV on random training/testing dataset in which some sequences are in the training dataset while others are in the test set can inflate the performance of classifier. Therefore, to deal with overlapping sequences and to check the predictive contribution of each target gene in the D922, we have used the LOTOCV method. In this method, we have assigned ASP-siRNAs targeting a particular gene in the validation dataset, while sequences from other genes were assigned to the training set. In total, 22 different sets have been made including one heterogeneous set titled “Others” which includes genes for which fewer ASP-siRNAs (<10) were reported (Table 4). Overall performance during 10nCV ranged from PCC values of a minimum of 0.53 to a maximum of 0.74 with an average PCC of 0.66. Performance on validation sets ranged from a PCC value of 0.20 to 0.88 with an average PCC of 0.40. Comparison of ASPsiPredSVM with other webservers: While comparing the performance of any two algorithms, one should use the same dataset for training and testing (Ahmed and Raghava 2011). In the literature, second-generation siRNA efficacy prediction tools were developed using the Huesken dataset and exhibit a very good PCC in the range of 0.56–0.85 (Train# column of Table 2). On the other hand, ASPsiPredSVM is developed on an updated ASP-siRNA dataset. Therefore, finding no similarity in the datasets employed to develop these tools, we have done comparative evaluation in three ways, i.e., by assessing the performance of (i) our algorithm with previously developed methods, (ii) cross-replacement of datasets, and (iii) our algorithm on an independent benchmarking dataset designated as “V419” (Ichihara et al. 2007). Performance of second-generation siRNA efficacy prediction algorithms on T737, V185, and V419 Table 2 Performance of second-generation siRNA efficacy prediction algorithms on T737, V185, and V419 . . . . . Pearson Correlation Coefficient (PCC) . S. No. . Reference . Technique . siRNA Dataset . ASP-siRNA Dataset . Train# . Val# . T737 . V185 . V419* . 1 Huesken et al. (2005) ANN Huesken2431 ✗ 0.67 0.66 Webserver not working 0.54 2 Vert et al. (2006) LR Huesken2431 ✗ 0.67 0.57 Webserver not working 0.55 3 Jiang et al. (2007) RFR 3589 ✗ 0.85 0.59 Webserver not working NA 4 Ichihara et al. (2007) LR Huesken2431 ✗ 0.72 NA 0.18 0.14 0.56 5 Ahmed and Raghava (2011) SVM Huesken2431 ✗ 0.65 0.65 0.27 0.25 0.55 6 siRNApred Kumar et al., (2009) SVM Huesken2431 ✗ 0.56 0.47 0.27 0.09 0.23 . . . . . Pearson Correlation Coefficient (PCC) . S. No. . Reference . Technique . siRNA Dataset . ASP-siRNA Dataset . Train# . Val# . T737 . V185 . V419* . 1 Huesken et al. (2005) ANN Huesken2431 ✗ 0.67 0.66 Webserver not working 0.54 2 Vert et al. (2006) LR Huesken2431 ✗ 0.67 0.57 Webserver not working 0.55 3 Jiang et al. (2007) RFR 3589 ✗ 0.85 0.59 Webserver not working NA 4 Ichihara et al. (2007) LR Huesken2431 ✗ 0.72 NA 0.18 0.14 0.56 5 Ahmed and Raghava (2011) SVM Huesken2431 ✗ 0.65 0.65 0.27 0.25 0.55 6 siRNApred Kumar et al., (2009) SVM Huesken2431 ✗ 0.56 0.47 0.27 0.09 0.23 Second-generation siRNA efficacy algorithms were developed on the Huesken dataset. S.No., Serial number; RFR, random forest regression; ANN, artificial neural network; LR, linear regression; Train# and Val# is the performance during n-fold cross-validation and independent validation of a particular algorithm. T737 and V185 column reflects the performance of algorithms on training/testing and independent validation sets of ASPsiPredSVM (in bold italics), while extreme right column indicates performance of algorithms on benchmarking dataset V419. Open in new tab Table 2 Performance of second-generation siRNA efficacy prediction algorithms on T737, V185, and V419 . . . . . Pearson Correlation Coefficient (PCC) . S. No. . Reference . Technique . siRNA Dataset . ASP-siRNA Dataset . Train# . Val# . T737 . V185 . V419* . 1 Huesken et al. (2005) ANN Huesken2431 ✗ 0.67 0.66 Webserver not working 0.54 2 Vert et al. (2006) LR Huesken2431 ✗ 0.67 0.57 Webserver not working 0.55 3 Jiang et al. (2007) RFR 3589 ✗ 0.85 0.59 Webserver not working NA 4 Ichihara et al. (2007) LR Huesken2431 ✗ 0.72 NA 0.18 0.14 0.56 5 Ahmed and Raghava (2011) SVM Huesken2431 ✗ 0.65 0.65 0.27 0.25 0.55 6 siRNApred Kumar et al., (2009) SVM Huesken2431 ✗ 0.56 0.47 0.27 0.09 0.23 . . . . . Pearson Correlation Coefficient (PCC) . S. No. . Reference . Technique . siRNA Dataset . ASP-siRNA Dataset . Train# . Val# . T737 . V185 . V419* . 1 Huesken et al. (2005) ANN Huesken2431 ✗ 0.67 0.66 Webserver not working 0.54 2 Vert et al. (2006) LR Huesken2431 ✗ 0.67 0.57 Webserver not working 0.55 3 Jiang et al. (2007) RFR 3589 ✗ 0.85 0.59 Webserver not working NA 4 Ichihara et al. (2007) LR Huesken2431 ✗ 0.72 NA 0.18 0.14 0.56 5 Ahmed and Raghava (2011) SVM Huesken2431 ✗ 0.65 0.65 0.27 0.25 0.55 6 siRNApred Kumar et al., (2009) SVM Huesken2431 ✗ 0.56 0.47 0.27 0.09 0.23 Second-generation siRNA efficacy algorithms were developed on the Huesken dataset. S.No., Serial number; RFR, random forest regression; ANN, artificial neural network; LR, linear regression; Train# and Val# is the performance during n-fold cross-validation and independent validation of a particular algorithm. T737 and V185 column reflects the performance of algorithms on training/testing and independent validation sets of ASPsiPredSVM (in bold italics), while extreme right column indicates performance of algorithms on benchmarking dataset V419. Open in new tab Our best model has achieved a maximum PCC of 0.71 on 10nCV and 0.65 on independent validation; which is comparable to previously developed siRNA efficacy prediction methods (Table 1). In the cross-replacement strategy, we have assessed the performance of available algorithms on our dataset (Table 2) and ASPsiPredSVM on theirs (Table 3). We found algorithms developed on Huesken2431 achieved PCCs in the range of 0.18 to 0.27 and 0.09 to 0.25 on our T737 and V185 datasets, respectively (see Table 2). On the other hand, ASPsiPredSVM has achieved PCCs of 0.23 and 0.26 on Huesken2431 (T2182/V249) (Table 3). Performance of ASPsiPredSVM on Huesken2431 and V419 Table 3 Performance of ASPsiPredSVM on Huesken2431 and V419 S. No. . Reference . Technique . siRNA dataset . ASP-siRNA . T737 . V185 . T2182 . V249 . V419 . 1 ASPsiPredSVM SVM ASP-siRNA (D922) ✓ 0.71 0.65 0.23 0.26 0.22 S. No. . Reference . Technique . siRNA dataset . ASP-siRNA . T737 . V185 . T2182 . V249 . V419 . 1 ASPsiPredSVM SVM ASP-siRNA (D922) ✓ 0.71 0.65 0.23 0.26 0.22 S.No., Serial number. The Huesken2431 dataset is divided into T2182 and V249 as training/testing and independent validation set. T737 and V185 column reflects the performance of ASPsiPredSVM on training/testing and independent validation sets; while V419 indicates performance on benchmarking dataset. Open in new tab Table 3 Performance of ASPsiPredSVM on Huesken2431 and V419 S. No. . Reference . Technique . siRNA dataset . ASP-siRNA . T737 . V185 . T2182 . V249 . V419 . 1 ASPsiPredSVM SVM ASP-siRNA (D922) ✓ 0.71 0.65 0.23 0.26 0.22 S. No. . Reference . Technique . siRNA dataset . ASP-siRNA . T737 . V185 . T2182 . V249 . V419 . 1 ASPsiPredSVM SVM ASP-siRNA (D922) ✓ 0.71 0.65 0.23 0.26 0.22 S.No., Serial number. The Huesken2431 dataset is divided into T2182 and V249 as training/testing and independent validation set. T737 and V185 column reflects the performance of ASPsiPredSVM on training/testing and independent validation sets; while V419 indicates performance on benchmarking dataset. Open in new tab Further, we have checked the performance of our algorithm on an independent benchmarking dataset, V419 (Ichihara et al. 2007). This dataset has also been utilized in previous tools to assess their performance. While Huesken-based methods have achieved correlation of 0.23 to 0.56 on V419 (extreme right column in Table 2), we attained a PCC of 0.22 (Table 3). ASPsiPredmatrix: performance evaluation of ASPsiPredmatrix on validation datasets: The second tier of our algorithm is the mismatch information matrix generated from the rule-based studies. It had achieved a PCC of 0.63 on V185 (Table S8 in File S1). Comparison of ASPsiPredMatrix with other webservers: Currently, there is no webserver to predict Effwild, although one method desiRm exists that describes the improvement in the efficacy of an siRNA after introducing mismatches in it. On the other hand, our method has the same ASP-siRNA but assessed against mismatches with the wild-type allele. Therefore, we have compared the performance of both methods using four experimental studies in which 19mer ASP-siRNAs complementary to a sliding window across a mutation were assessed. Performance of desiRm was not satisfactory on single-nucleotide sliding trails, while the matrix-based method attained a collective PCC in the range of 0.35–0.52 (Table S8 in File S1). Discussion Post-ENCODE (Lussier et al. 2013; Venter et al. 2001), a plethora of information has been released about genome sequence, structure and multifaceted ways of its regulation. This information has provided new opportunities to understand complex genetic disorders at the molecular level. Thus, it will be useful for tailoring the conventional gene therapy into a custom-made one (Lander 2011). In this context, RNA targeting approaches up to the precision of single-nucleotide discrimination are emerging as a potential and therapeutic alternative to traditionally undruggable targets (Keiser et al. 2016). ASGS is a progressive technique for tailored treatment of dominantly inherited disorders. An ASP-siRNA is designed to target an allele of interest/mutant allele at any location where it differs from its wild-type counterpart (Lombardi et al. 2009). Despite its immense medical importance, a dedicated informatics resource in this field was lacking, which encouraged us to develop resources on ASP-siRNAs implicated in various genetic diseases. While existing archives hold information about siRNAs targeted against one gene with a single inhibitory efficacy (Table S1 in File S1), ASPsiDb harbors ASP-siRNAs targeted against the mutant and wild-type alleles of a gene and hence associated with two inhibitory efficacies (Effmut/Effwild). It was after the breakthrough discovery that RISC-mediated cleavage occurs at the phosphodiester bond of the 10th nucleotide position on the guide strand (Elbashir et al. 2001b; Haley and Zamore 2004) that researchers around the world started utilizing its role in achieving ASGS by placing the nucleotide complementary to the mutation at the 10th or central positions of siRNAs to make it less accessible to the normal allele. This scrutiny was employed in achieving ASGS by directly targeting disease-causing mutations (Jiang et al. 2013; Lyu et al. 2016) or indirectly targeting disease-associated SNPs in linkage disequilibrium (Drouet et al. 2014; Yu et al. 2012). Moreover, mutation-specific suppression has also been accomplished for mutant alleles exhibiting deletions by placing mutation-specific nucleotides at the central positions (Gonzalez-Alegre et al. 2003). Although there were several reports studying the effect of placing nucleotides complementary at the mutation on the efficacy of the mutant allele (Effmut), but an algorithm employing these studies was lacking. Correspondingly, there were some rule-based studies reporting the effect of siRNA: mRNA residue clash on efficacy at all 19 locations of the siRNA guide strand (Birmingham et al. 2006; Huang et al. 2009; Ohnishi et al. 2008; Schwarz et al. 2006). It is also testified that siRNA: mRNA residue clash of purine: purine (pur:pur) type is less tolerable than pyrimidine: pyrimidine (pyr:pyr) clash. For example, siRNA “siC7/8” having G: G clash with the wild-type allele suppresses the mutant allele three fold more than its counterpart (Miller et al. 2003). In some cases, when siRNA: mRNA have a pyr:pyr or pyr:pur clash, an additional mismatch is introduced in the siRNA to make it more discriminative (Miller et al. 2004). Despite these rule-based studies, there is no algorithm employing these findings for prediction of Effmut and Effwild. We have developed ASPsiPred, the first web server in this field incorporating a two-tiered algorithm (ASPsiPredSVM and ASPsiPredmatrix) for predicting efficacies Effmut and Effwild. In the literature, initially many mammalian siRNA efficacy prediction algorithms were developed using heterogeneous siRNA datasets and achieved a good PCC of 0.46–0.56 (Holen 2006; Saetrom 2004; Shabalina et al. 2006). Thereafter, algorithms to predict siRNA efficacies were reported using the Huesken dataset (Huesken et al. 2005) and exhibited very good PCC values in the range of 0.56–0.85. Likewise, ASPsiPredSVM has achieved a correlation of 0.71 on 10nCV and 0.65 on an independent validation set (Table 1). The ASP-siRNA dataset (D922) has not been employed anywhere in the present mammalian siRNA efficacy algorithms. Moreover, our algorithm has not utilized currently available siRNA datasets other than D922. Further, it has been reported that siRNA algorithms perform less well on datasets in which they have not been trained (Qureshi et al. 2013). Correspondingly, the performance of other available algorithms on our dataset (Table 2) and ASPsiPredSVM on their datasets was lower (Table 3). ASPsiPredSVM performed better on the ASP-siRNA datasets including T737 and V185 sets (Table 3). However, it achieved a PCC of 0.23 and 0.26 on the Huesken2431 dataset (T2182/V249). This may be because it has only been trained on an allele-specific dataset and suggests the need of an ASP-siRNA efficacy prediction algorithm. Thus, ASPsiPredSVM will be helpful for researchers in designing and predicting Effmut for consecutive single-nucleotide sliding siRNAs for a given gene that is not necessarily linked to disease. For this purpose, we have provided our best predictive model as a general siRNA efficacy predictor under the separate ASPsiPredSVM section on the web server. As the D922 dataset covers sequences with single-nucleotide sliding differences, there is overlap among them. Therefore, the simple 10nCV in which overlapping sequences are randomly assigned to training and test sets could inflate the performance of the algorithm. Thus, to further address this issue, we have used the LOTOCV method in which ASP-siRNAs from each target gene are iteratively excluded and the classifier is trained on sequences from the remaining genes followed by testing on the sequences from the excluded gene (Table 4). Out of the 21 genes, predictive performance of 14 genes was satisfactory despite the fact that data from that gene were not present in the training set. Therefore, results from the above strategy show that ASPsiPredSVM can act as a general ASP-siRNA efficacy prediction algorithm for other genes (Table 4). However, predictive performance of some of the genes was less than satisfactory. This may be due to the difference in the pattern of the target gene mutation, which might be improved in the future based on the availability of more data. Performances of the SVM models during 10-fold cross-validation using LOTOCV method Table 4 Performances of the SVM models during 10-fold cross-validation using LOTOCV method . . No. of ASP-siRNAs . Pearson Correlation Coefficient (PCC) During 10nCV and IV . S. No. . Gene Name . Training Dataset . Validation Dataset . 10nCV . IV . 1 APP 907 15 0.71 0.88 2 AR 912 10 0.71 0.19 3 COL1A1 912 10 0.71 0.49 4 COL3A1 903 19 0.71 0.34 5 COL6A3 911 11 0.70 0.24 6 COL7A1 903 19 0.71 0.55 7 HTT 883 39 0.56 0.28 8 KRAS 844 78 0.68 0.31 9 KRT12 884 38 0.71 0.48 10 KRT5 884 38 0.71 0.24 11 KRT6a 903 19 0.70 0.31 12 KRT9 830 92 0.63 0.26 13 LRRK2 901 21 0.71 0.26 14 Others 844 78 0.74 0.20 15 P. Luciferase 865 57 0.71 0.23 16 PPIB 695 227 0.53 0.61 17 PRNP 904 18 0.71 0.79 18 PSEN1 903 19 0.43 0.30 19 SNCA 906 16 0.71 0.50 20 SOD1 881 41 0.53 0.34 21 TGFBI 903 19 0.55 0.64 22 TP63 884 38 0.58 0.33 . . No. of ASP-siRNAs . Pearson Correlation Coefficient (PCC) During 10nCV and IV . S. No. . Gene Name . Training Dataset . Validation Dataset . 10nCV . IV . 1 APP 907 15 0.71 0.88 2 AR 912 10 0.71 0.19 3 COL1A1 912 10 0.71 0.49 4 COL3A1 903 19 0.71 0.34 5 COL6A3 911 11 0.70 0.24 6 COL7A1 903 19 0.71 0.55 7 HTT 883 39 0.56 0.28 8 KRAS 844 78 0.68 0.31 9 KRT12 884 38 0.71 0.48 10 KRT5 884 38 0.71 0.24 11 KRT6a 903 19 0.70 0.31 12 KRT9 830 92 0.63 0.26 13 LRRK2 901 21 0.71 0.26 14 Others 844 78 0.74 0.20 15 P. Luciferase 865 57 0.71 0.23 16 PPIB 695 227 0.53 0.61 17 PRNP 904 18 0.71 0.79 18 PSEN1 903 19 0.43 0.30 19 SNCA 906 16 0.71 0.50 20 SOD1 881 41 0.53 0.34 21 TGFBI 903 19 0.55 0.64 22 TP63 884 38 0.58 0.33 ASP-siRNAs targeting a particular gene are assigned to the validation dataset, while sequences from other genes were assigned to the training set. Validation of the models was done using respective gene in the independent validation set. Standard HGNC gene symbols have been used. PCC is between the actual and observed Effmut. The training dataset is used to train different predictive models, while independent validation datasets were not used in any training algorithms. S.No., Serial number; 10nCV, ten-fold cross-validation; IV, independent validation. Open in new tab Table 4 Performances of the SVM models during 10-fold cross-validation using LOTOCV method . . No. of ASP-siRNAs . Pearson Correlation Coefficient (PCC) During 10nCV and IV . S. No. . Gene Name . Training Dataset . Validation Dataset . 10nCV . IV . 1 APP 907 15 0.71 0.88 2 AR 912 10 0.71 0.19 3 COL1A1 912 10 0.71 0.49 4 COL3A1 903 19 0.71 0.34 5 COL6A3 911 11 0.70 0.24 6 COL7A1 903 19 0.71 0.55 7 HTT 883 39 0.56 0.28 8 KRAS 844 78 0.68 0.31 9 KRT12 884 38 0.71 0.48 10 KRT5 884 38 0.71 0.24 11 KRT6a 903 19 0.70 0.31 12 KRT9 830 92 0.63 0.26 13 LRRK2 901 21 0.71 0.26 14 Others 844 78 0.74 0.20 15 P. Luciferase 865 57 0.71 0.23 16 PPIB 695 227 0.53 0.61 17 PRNP 904 18 0.71 0.79 18 PSEN1 903 19 0.43 0.30 19 SNCA 906 16 0.71 0.50 20 SOD1 881 41 0.53 0.34 21 TGFBI 903 19 0.55 0.64 22 TP63 884 38 0.58 0.33 . . No. of ASP-siRNAs . Pearson Correlation Coefficient (PCC) During 10nCV and IV . S. No. . Gene Name . Training Dataset . Validation Dataset . 10nCV . IV . 1 APP 907 15 0.71 0.88 2 AR 912 10 0.71 0.19 3 COL1A1 912 10 0.71 0.49 4 COL3A1 903 19 0.71 0.34 5 COL6A3 911 11 0.70 0.24 6 COL7A1 903 19 0.71 0.55 7 HTT 883 39 0.56 0.28 8 KRAS 844 78 0.68 0.31 9 KRT12 884 38 0.71 0.48 10 KRT5 884 38 0.71 0.24 11 KRT6a 903 19 0.70 0.31 12 KRT9 830 92 0.63 0.26 13 LRRK2 901 21 0.71 0.26 14 Others 844 78 0.74 0.20 15 P. Luciferase 865 57 0.71 0.23 16 PPIB 695 227 0.53 0.61 17 PRNP 904 18 0.71 0.79 18 PSEN1 903 19 0.43 0.30 19 SNCA 906 16 0.71 0.50 20 SOD1 881 41 0.53 0.34 21 TGFBI 903 19 0.55 0.64 22 TP63 884 38 0.58 0.33 ASP-siRNAs targeting a particular gene are assigned to the validation dataset, while sequences from other genes were assigned to the training set. Validation of the models was done using respective gene in the independent validation set. Standard HGNC gene symbols have been used. PCC is between the actual and observed Effmut. The training dataset is used to train different predictive models, while independent validation datasets were not used in any training algorithms. S.No., Serial number; 10nCV, ten-fold cross-validation; IV, independent validation. Open in new tab Additionally, there is no web server to predict the efficacy of ASP-siRNAs with a wild-type allele having a single mismatch (Effwild). Though desiRm also deals with mismatches and efficacy, it aims to improve the efficacy of an siRNA by introducing mismatches in the same target sequence. On the other hand, ASPsiPredmatrix is intended to predict the efficacy of ASP-siRNA targeting a wild-type allele (Effwild) with one mismatch. desiRm is associated with one efficacy value at a time, while ASPsiPred predicts two efficacies (Effmut/Effwild) simultaneously from two methods. In the former, a mismatch is introduced in the siRNA for the same target sequence to improve efficacy, while in the latter case, a mismatch is present between wild-type allele and ASP-siRNA. desiRm was developed on the Huesken dataset and ASPsiPred is developed using ASPs-RNAs, which is a novel siRNA dataset in the literature. We have also compared the performance of both methods on four experimental studies of multiple 19mer siRNAs offset along a target and found that ASPsiPredmatrix performs better in predicting single-nucleotide sliding 19mer trails (Table S9 in File S1). It is well established that off-target effects are a major issue during siRNA-based gene silencing and seed regions are a key determinant for these effects (Birmingham et al. 2006; Jackson et al. 2003; Kamola et al. 2015). Therefore, to deal with off-targets, we have also integrated the ASP-siOffTar tool to deliver a list of off-target hits based on the alignment of the seed regions of ASP-siRNA or any siRNA to the human genome. To extend the off-targets repertoire of particular siRNAs, a full sequence based off-target tool is also integrated on the web interface with a maximum of three allowed mismatches. Furthermore, many chemical modifications (cm) on siRNAs have been used to reduce off-target effects and increase the half-life of siRNAs by making it nuclease resistant (Dar et al. 2016b). We have also compiled a list of 422 cm ASP-siRNAs and provided it on our web server. Although ASP-RNAi is a powerful tool, various factors must be taken into account before it enters clinic, such as binding of siRNAs to unintended off-targets via partial sequence complementarity (Kamola et al. 2015), stability, and half-life (Dar et al. 2016b). Successful siRNA delivery is also an important contributing factor, which depends upon choice of transfection reagent and the intrinsic susceptibility of the target cell type (Nabzdyk et al. 2011). Thus, the ASPsiRNA resource would be immensely helpful for in silico design and predicting efficacy of ASP-siRNAs for various maladies, e.g., in cancer-associated SNPs (Iyer et al. 2016; Mook et al. 2009), for treatment of genetic diseases, e.g., from currently incurable autosomal dominant (Miller et al. 2004) to severe sex-linked disorders (Caplen et al. 2002), in combating viral drug resistance (Teng et al. 2011), and many more. It will also be beneficial for researchers who wish to study the function of alleles. Currently, our method is limited to the prediction of Effwild with a single mismatch due to limited data on multiple mismatches. It also has limited performance on unseen or novel genes owing to a limited number of target genes in the dataset. In the future, there would be a need to develop an algorithm for >1 mismatch, which can improve allelic discrimination. Nevertheless, the upcoming use of ASP selectivity will not only be useful to suppress disease-associated SNPs, but can also be applied as a research tool where you can silence one splice variant from other (Trochet et al. 2015). Conclusion and future implications Understanding distinctive aspects of ASGS by ASP-siRNAs may be exploited in the treatment of currently incurable dominant genetic disorders. In this ASPsiRNA resource, ASPsiDb provides a highly annotated dataset of ASP-siRNAs and their associated targets. It also provides a two-layered algorithm to design effective and discriminatory siRNAs against heterozygous SNPs (ASPsiPredSVM) and wild-type alleles (ASPsiPredmatrix) coupled with useful tools like ASP-siOffTar for off-target analysis. We hope ASPsiPred will be immensely helpful to target not only disease-causing mutations, but also to study the biological function of alleles that are not necessarily linked to disease. Acknowledgments This work was partially supported by the Council of Scientific and Industrial Research (CSIR)-Institute of Microbial Technology (OLP 0083) and the Department of Biotechnology, Government of India (GAP0001) provided financial support for this work. Funding for open-access charge was provided by CSIR-Institute of Microbial Technology, Sector 39-A, Chandigarh, India. The authors declare that they have no competing interests. Author contributions: M.K. and N.T. conceived the idea and execution strategy. I.M. manually collected and curated the data. I.M., A.Q., and N.T. designed the web server part of the database. I.M., N.T., and A.K.G. performed the execution of prediction algorithm. I.M. and M.K. performed data analysis, interpreted results, and drafted the manuscript. M.K. coordinated the entire project. All authors read and approved the final manuscript. Footnotes Supplemental material is available online at www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.044024/-/DC1. Communicating editor: J. Prendergast Literature Cited Aharony I , Ehrnhoefer D E, Shruster A, Qiu X, Franciosi S et al. , 2015   A Huntingtin-based peptide inhibitor of caspase-6 provides protection from mutant Huntingtin-induced motor and behavioral deficits. Hum. Mol. Genet. 24 : 2604 – 2614 . Google Scholar Crossref Search ADS PubMed WorldCat Ahmed F , Raghava G P, 2011   Designing of highly effective complementary and mismatch siRNAs for silencing a gene. PLoS One 6 : e23443 . Google Scholar Crossref Search ADS PubMed WorldCat Allen E H , Atkinson S D, Liao H, Moore J E, Leslie Pedrioli D M et al. , 2013   Allele-specific siRNA silencing for the common keratin 12 founder mutation in Meesmann epithelial corneal dystrophy. Invest. Ophthalmol. Vis. Sci. 54 : 494 – 502 . Google Scholar Crossref Search ADS PubMed WorldCat Alves S , Nascimento-Ferreira I, Auregan G, Hassig R, Dufour N et al. , 2008   Allele-specific RNA silencing of mutant ataxin-3 mediates neuroprotection in a rat model of Machado-Joseph disease. PLoS One 3 : e3341 . Google Scholar Crossref Search ADS PubMed WorldCat Amarzguioui M , Prydz H, 2004   An algorithm for selection of functional siRNA sequences. Biochem. Biophys. Res. Commun. 316 : 1050 – 1058 . Google Scholar Crossref Search ADS PubMed WorldCat Arribat Y , Bonneaud N, Talmat-Amar Y, Layalle S, Parmentier M L et al. , 2013   A huntingtin peptide inhibits polyQ-huntingtin associated defects. PLoS One 8 : e68775 . Google Scholar Crossref Search ADS PubMed WorldCat Atkinson S D , McGilligan V E, Liao H, Szeverenyi I, Smith F J et al. , 2011   Development of allele-specific therapeutic siRNA for keratin 5 mutations in epidermolysis bullosa simplex. J. Invest. Dermatol. 131 : 2079 – 2086 . Google Scholar Crossref Search ADS PubMed WorldCat Birmingham A , Anderson E M, Reynolds A, Ilsley-Tyree D, Leake D et al. , 2006   3′ UTR seed matches, but not overall identity, are associated with RNAi off-targets. Nat. Methods 3 : 199 – 204 . Google Scholar Crossref Search ADS PubMed WorldCat Caplen N J , Taylor J P, Statham V S, Tanaka F, Fire A et al. , 2002   Rescue of polyglutamine-mediated cytotoxicity by double-stranded RNA-mediated RNA interference. Hum. Mol. Genet. 11 : 175 – 184 . Google Scholar Crossref Search ADS PubMed WorldCat Courtney D G , Atkinson S D, Moore J E, Maurizi E, Serafini C et al. , 2014   Development of allele-specific gene-silencing siRNAs for TGFBI Arg124Cys in lattice corneal dystrophy type I. Invest. Ophthalmol. Vis. Sci. 55 : 977 – 985 . Google Scholar Crossref Search ADS PubMed WorldCat Dar S A , Gupta A K, Thakur A, Kumar M, 2016 a   SMEpred workbench: a web server for predicting efficacy of chemically modified siRNAs. RNA Biol. 13 : 1144 – 1151 . Google Scholar Crossref Search ADS PubMed WorldCat Dar S A , Thakur A, Qureshi A, Kumar M, 2016 b   siRNAmod: a database of experimentally validated chemically modified siRNAs. Sci. Rep. 6 : 20031 . Google Scholar Crossref Search ADS PubMed WorldCat Davidson B L , Paulson H L, 2004   Molecular medicine for the brain: silencing of disease genes with RNA interference. Lancet Neurol. 3 : 145 – 149 . Google Scholar Crossref Search ADS PubMed WorldCat Drouet V , Ruiz M, Zala D, Feyeux M, Auregan G et al. , 2014   Allele-specific silencing of mutant huntingtin in rodent brain and human stem cells. PLoS One 9 : e99341 . Google Scholar Crossref Search ADS PubMed WorldCat Elbashir S M , Harborth J, Lendeckel W, Yalcin A, Weber K et al. , 2001 a   Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 411 : 494 – 498 . Google Scholar Crossref Search ADS PubMed WorldCat Elbashir S M , Martinez J, Patkaniowska A, Lendeckel W, Tuschl T, 2001 b   Functional anatomy of siRNAs for mediating efficient RNAi in Drosophila melanogaster embryo lysate. EMBO J. 20 : 6877 – 6888 . Google Scholar Crossref Search ADS PubMed WorldCat Filhol O , Ciais D, Lajaunie C, Charbonnier P, Foveau N et al. , 2012   DSIR: assessing the design of highly potent siRNA by testing a set of cancer-relevant target genes. PLoS One 7 : e48057 . Google Scholar Crossref Search ADS PubMed WorldCat Fire A , Xu S, Montgomery M K, Kostas S A, Driver S E et al. , 1998   Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391 : 806 – 811 . Google Scholar Crossref Search ADS PubMed WorldCat Gonzalez-Alegre P , 2007   Therapeutic RNA interference for neurodegenerative diseases: from promise to progress. Pharmacol. Ther. 114 : 34 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat Gonzalez-Alegre P , Miller V M, Davidson B L, Paulson H L, 2003   Toward therapy for DYT1 dystonia: allele-specific silencing of mutant TorsinA. Ann. Neurol. 53 : 781 – 787 . Google Scholar Crossref Search ADS PubMed WorldCat Gonzalez-Alegre P , Bode N, Davidson B L, Paulson H L, 2005   Silencing primary dystonia: lentiviral-mediated RNA interference therapy for DYT1 dystonia. J. Neurosci. 25 : 10502 – 10509 . Google Scholar Crossref Search ADS PubMed WorldCat Haley B , Zamore P D, 2004   Kinetic analysis of the RNAi enzyme complex. Nat. Struct. Mol. Biol. 11 : 599 – 606 . Google Scholar Crossref Search ADS PubMed WorldCat Hamosh A , Scott A F, Amberger J, Valle D, McKusick V A, 2000   Online mendelian inheritance in man (OMIM). Hum. Mutat. 15 : 57 – 61 . Google Scholar Crossref Search ADS PubMed WorldCat Holen T , 2006   Efficient prediction of siRNAs with siRNArules 1.0: an open-source JAVA approach to siRNA algorithms. RNA 12 : 1620 – 1625 . Google Scholar Crossref Search ADS PubMed WorldCat Huang H , Qiao R, Zhao D, Zhang T, Li Y et al. , 2009   Profiling of mismatch discrimination in RNAi enabled rational design of allele-specific siRNAs. Nucleic Acids Res. 37 : 7560 – 7569 . Google Scholar Crossref Search ADS PubMed WorldCat Huesken D , Lange J, Mickanin C, Weiler J, Asselbergs F et al. , 2005   Design of a genome-wide siRNA library using an artificial neural network. Nat. Biotechnol. 23 : 995 – 1001 . Google Scholar Crossref Search ADS PubMed WorldCat Ichihara M , Murakumo Y, Masuda A, Matsuura T, Asai N et al. , 2007   Thermodynamic instability of siRNA duplex is a prerequisite for dependable prediction of siRNA activities. Nucleic Acids Res. 35 : e123 . Google Scholar Crossref Search ADS PubMed WorldCat Iyer S V , Parrales A, Begani P, Narkar A, Adhikari A S et al. , 2016   Allele-specific silencing of mutant p53 attenuates dominant-negative and gain-of-function activities. Oncotarget 7 : 5401 – 5415 . Google Scholar Crossref Search ADS PubMed WorldCat Jackson A L , Bartz S R, Schelter J, Kobayashi S V, Burchard J et al. , 2003   Expression profiling reveals off-target gene regulation by RNAi. Nat. Biotechnol. 21 : 635 – 637 . Google Scholar Crossref Search ADS PubMed WorldCat Jamwal S , Kumar P, 2015   Antidepressants for neuroprotection in Huntington’s disease: a review. Eur. J. Pharmacol. 769 : 33 – 42 . Google Scholar Crossref Search ADS PubMed WorldCat Jiang, P., H. Wu, Y. Da, F. Sang, J. Wei, et al. 2007 RFRCDB-siRNA: improved design of siRNAs by random forest regression model coupled with database searching. Comput Methods Programs Biomed 87: 230–238. Jiang J , Wakimoto H, Seidman J G, Seidman C E, 2013   Allele-specific silencing of mutant Myh6 transcripts in mice suppresses hypertrophic cardiomyopathy. Science 342 : 111 – 114 . Google Scholar Crossref Search ADS PubMed WorldCat Kamola P J , Nakano Y, Takahashi T, Wilson P A, Ui-Tei K, 2015   The siRNA non-seed region and its target sequences are auxiliary determinants of off-target effects. PLoS Comput. Biol. 11 : e1004656 . Google Scholar Crossref Search ADS PubMed WorldCat Kaur K , Gupta A K, Rajput A, Kumar M, 2016   ge-CRISPR – an integrated pipeline for the prediction and analysis of sgRNAs genome editing efficiency for CRISPR/Cas system. Sci. Rep. 6 : 30870 . Google Scholar Crossref Search ADS PubMed WorldCat Keiser M S , Kordower J H, Gonzalez-Alegre P, Davidson B L, 2015   Broad distribution of ataxin 1 silencing in rhesus cerebella for spinocerebellar ataxia type 1 therapy. Brain 138 : 3555 – 3566 . Google Scholar Crossref Search ADS PubMed WorldCat Keiser M S , Kordasiewicz H B, McBride J L, 2016   Gene suppression strategies for dominantly inherited neurodegenerative diseases: lessons from Huntington’s disease and spinocerebellar ataxia. Hum. Mol. Genet. 25 : R53 – R64 . Google Scholar Crossref Search ADS PubMed WorldCat Krzywinski M , Schein J, Birol I, Connors J, Gascoyne R et al. , 2009   Circos: an information aesthetic for comparative genomics. Genome Res. 19 : 1639 – 1645 . Google Scholar Crossref Search ADS PubMed WorldCat Kulshreshtha A , Piplani P, 2016   Current pharmacotherapy and putative disease-modifying therapy for Alzheimer’s disease. Neurol. Sci. 37 : 1403 – 1435 . Google Scholar Crossref Search ADS PubMed WorldCat Kumar, M., S. Lata, and G. Raghava, 2009 siRNApred: SVM based method for predicting efficacy value of siRNA. Paper presented at: Proceedings of the first international conference on Open Source for Computer Aided Drug Discovery (OSCADD) (CSIR-IMTECH). Lander E S , 2011   Initial impact of the sequencing of the human genome. Nature 470 : 187 – 197 . Google Scholar Crossref Search ADS PubMed WorldCat Landrum M J , Lee J M, Riley G R, Jang W, Rubinstein W S et al. , 2014   ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 42 : D980 – D985 . Google Scholar Crossref Search ADS PubMed WorldCat Lappalainen I , Lopez J, Skipper L, Hefferon T, Spalding J D et al. , 2013   DbVar and DGVa: public archives for genomic structural variation. Nucleic Acids Res. 41 : D936 – D941 . Google Scholar Crossref Search ADS PubMed WorldCat Leachman S A , Hickerson R P, Hull P R, Smith F J, Milstone L M et al. , 2008   Therapeutic siRNAs for dominant genetic skin disorders including pachyonychia congenita. J. Dermatol. Sci. 51 : 151 – 157 . Google Scholar Crossref Search ADS PubMed WorldCat Leachman S A , Hickerson R P, Schwartz M E, Bullough E E, Hutcherson S L et al. , 2010   First-in-human mutation-targeted siRNA phase Ib trial of an inherited skin disorder. Mol. Ther. 18 : 442 – 446 . Google Scholar Crossref Search ADS PubMed WorldCat LeWitt P A , Hauser R A, Grosset D G, Stocchi F, Saint-Hilaire M H et al. , 2016   A randomized trial of inhaled levodopa (CVT-301) for motor fluctuations in Parkinson’s disease. Mov. Disord. 31 : 1356 – 1365 . Google Scholar Crossref Search ADS PubMed WorldCat Liu Y , Snedecor E R, Zhang X, Xu Y, Huang L et al. , 2016   Correction of hair shaft defects through allele-specific silencing of mutant Krt75. J. Invest. Dermatol. 136 : 45 – 51 . Google Scholar Crossref Search ADS PubMed WorldCat Lombardi M S , Jaspers L, Spronkmans C, Gellera C, Taroni F et al. , 2009   A majority of Huntington’s disease patients may be treatable by individualized allele-specific RNA interference. Exp. Neurol. 217 : 312 – 319 . Google Scholar Crossref Search ADS PubMed WorldCat Lopes C , Aubert S, Bourgois-Rocha F, Barnat M, Rego A C et al. , 2016   Dominant-negative effects of adult-onset huntingtin mutations alter the division of human embryonic stem cells-derived neural cells. PLoS One 11 : e0148680 . Google Scholar Crossref Search ADS PubMed WorldCat Loy R E , Lueck J D, Mostajo-Radji M A, Carrell E M, Dirksen R T, 2012   Allele-specific gene silencing in two mouse models of autosomal dominant skeletal myopathy. PLoS One 7 : e49757 . Google Scholar Crossref Search ADS PubMed WorldCat Lussier Y A , Li H, Maienschein-Cline M, 2013   Conquering computational challenges of omics data and post-ENCODE paradigms. Genome Biol. 14 : 310 . Google Scholar Crossref Search ADS PubMed WorldCat Lyu Y S , Shi P L, Chen X L, Tang Y X, Wang Y F et al. , 2016   A small indel mutant mouse model of epidermolytic palmoplantar keratoderma and its application to mutant-specific shRNA therapy. Mol. Ther. Nucleic Acids 5 : e299 . Google Scholar Crossref Search ADS PubMed WorldCat Marelli C , Maschat F, 2016   The P42 peptide and peptide-based therapies for Huntington’s disease. Orphanet J. Rare Dis. 11 : 24 . Google Scholar Crossref Search ADS PubMed WorldCat Mazur S , Csucs G, Kozak K, 2012   RNAiAtlas: a database for RNAi (siRNA) libraries and their specificity. Database (Oxford) 2012 : bas027 . Google Scholar Crossref Search ADS PubMed WorldCat McQuisten K A , Peek A S, 2009   Comparing artificial neural networks, general linear models and support vector machines in building predictive models for small interfering RNAs. PLoS One 4 : e7522 . Google Scholar Crossref Search ADS PubMed WorldCat Miller V M , Xia H, Marrs G L, Gouvion C M, Lee G et al. , 2003   Allele-specific silencing of dominant disease genes. Proc. Natl. Acad. Sci. USA 100 : 7195 – 7200 . Google Scholar Crossref Search ADS WorldCat Miller V M , Gouvion C M, Davidson B L, Paulson H L, 2004   Targeting Alzheimer’s disease genes with RNA interference: an efficient strategy for silencing mutant alleles. Nucleic Acids Res. 32 : 661 – 668 . Google Scholar Crossref Search ADS PubMed WorldCat Miniarikova J , Zanella I, Huseinovic A, van der Zon T, Hanemaaijer E et al. , 2016   Design, characterization, and lead selection of therapeutic miRNAs targeting huntingtin for development of gene therapy for Huntington’s disease. Mol. Ther. Nucleic Acids 5 : e297 . Google Scholar Crossref Search ADS PubMed WorldCat Mook O R , Baas F, de Wissel M B, Fluiter K, 2009   Allele-specific cancer cell killing in vitro and in vivo targeting a single-nucleotide polymorphism in POLR2A. Cancer Gene Ther. 16 : 532 – 538 . Google Scholar Crossref Search ADS PubMed WorldCat Mysara M , Garibaldi J M, Elhefnawi M, 2011   MysiRNA-designer: a workflow for efficient siRNA design. PLoS One 6 : e25642 . Google Scholar Crossref Search ADS PubMed WorldCat Nabzdyk C S , Chun M, Pradhan L, Logerfo F W, 2011   High throughput RNAi assay optimization using adherent cell cytometry. J. Transl. Med. 9 : 48 . Google Scholar Crossref Search ADS PubMed WorldCat Ohnishi Y , Tamura Y, Yoshida M, Tokunaga K, Hohjoh H, 2008   Enhancement of allele discrimination by introduction of nucleotide mismatches into siRNA in allele-specific gene silencing by RNAi. PLoS One 3 : e2248 . Google Scholar Crossref Search ADS PubMed WorldCat Pan W J , Chen C W, Chu Y W, 2011   siPRED: predicting siRNA efficacy using various characteristic methods. PLoS One 6 : e27602 . Google Scholar Crossref Search ADS PubMed WorldCat Pandey S K , Wheeler T M, Justice S L, Kim A, Younis H S et al. , 2015   Identification and characterization of modified antisense oligonucleotides targeting DMPK in mice and nonhuman primates for the treatment of myotonic dystrophy type 1. J. Pharmacol. Exp. Ther. 355 : 329 – 340 . Google Scholar Crossref Search ADS PubMed WorldCat Paulson H , Gonzalez-Alegre P, 2006   RNAi gets its prize. Lancet Neurol. 5 : 997 – 999 . Google Scholar Crossref Search ADS PubMed WorldCat Peek A S , 2007   Improving model predictions for RNA interference activities that use support vector machine regression by combining and filtering features. BMC Bioinformatics 8 : 182 . Google Scholar Crossref Search ADS PubMed WorldCat Qureshi A , Thakur N, Kumar M, 2013   VIRsiRNApred: a web server for predicting inhibition efficacy of siRNAs targeting human viruses. J. Transl. Med. 11 : 305 . Google Scholar Crossref Search ADS PubMed WorldCat Qureshi A , Thakur N, Monga I, Thakur A, Kumar M, 2014   VIRmiRNA: a comprehensive resource for experimentally validated viral miRNAs and their targets. Database (Oxford) 2014 : bau103 . Google Scholar Crossref Search ADS PubMed WorldCat Ren Y , Gong W, Xu Q, Zheng X, Lin D et al. , 2006   siRecords: an extensive database of mammalian siRNAs with efficacy ratings. Bioinformatics 22 : 1027 – 1028 . Google Scholar Crossref Search ADS PubMed WorldCat Reynolds A , Leake D, Boese Q, Scaringe S, Marshall W S et al. , 2004   Rational siRNA design for RNA interference. Nat. Biotechnol. 22 : 326 – 330 . Google Scholar Crossref Search ADS PubMed WorldCat Rodriguez-Lebron E , Paulson H L, 2006   Allele-specific RNA interference for neurological disease. Gene Ther. 13 : 576 – 581 . Google Scholar Crossref Search ADS PubMed WorldCat Saetrom P , 2004   Predicting the efficacy of short oligonucleotides in antisense and RNAi experiments with boosted genetic programming. Bioinformatics 20 : 3055 – 3063 . Google Scholar Crossref Search ADS PubMed WorldCat Schwarz D S , Ding H, Kennington L, Moore J T, Schelter J et al. , 2006   Designing siRNA that distinguish between genes that differ by a single nucleotide. PLoS Genet. 2 : e140 . Google Scholar Crossref Search ADS PubMed WorldCat Seyhan A A , 2011   RNAi: a potential new class of therapeutic for human genetic disease. Hum. Genet. 130 : 583 – 605 . Google Scholar Crossref Search ADS PubMed WorldCat Shabalina S A , Spiridonov A N, Ogurtsov A Y, 2006   Computational models with thermodynamic and composition features improve siRNA design. BMC Bioinformatics 7 : 65 . Google Scholar Crossref Search ADS PubMed WorldCat Sherry S T , Ward M, Sirotkin K, 1999   dbSNP-database for single nucleotide polymorphisms and other classes of minor genetic variation. Genome Res. 9 : 677 – 679 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Sierant M , Paduszynska A, Kazmierczak-Baranska J, Nacmias B, Sorbi S et al. , 2011   Specific silencing of L392V PSEN1 mutant allele by RNA interference. Int. J. Alzheimers Dis. 2011 : 809218 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Squitieri F , de Yebenes J G, 2015   Profile of pridopidine and its potential in the treatment of Huntington disease: the evidence to date. Drug Des. Devel. Ther. 9 : 5827 – 5833 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Takahashi M , Suzuki M, Fukuoka M, Fujikake N, Watanabe S et al. , 2015   Normalization of overexpressed alpha-synuclein causing Parkinson’s disease by a moderate gene silencing with RNA interference. Mol. Ther. Nucleic Acids 4 : e241 . Google Scholar Crossref Search ADS PubMed WorldCat Teng X , Liu J Y, Li D, Fang Y, Wang X Y et al. , 2011   Application of allele-specific RNAi in hepatitis B virus lamivudine resistance. J. Viral Hepat. 18 : e491 – e498 . Google Scholar Crossref Search ADS PubMed WorldCat Thakur N , Qureshi A, Kumar M, 2012 a   AVPpred: collection and prediction of highly effective antiviral peptides. Nucleic Acids Res. 40 : W199 – W204 . Google Scholar Crossref Search ADS PubMed WorldCat Thakur N , Qureshi A, Kumar M, 2012 b   VIRsiRNAdb: a curated database of experimentally validated viral siRNA/shRNA. Nucleic Acids Res. 40 : D230 – D236 . Google Scholar Crossref Search ADS PubMed WorldCat Trochet D , Prudhon B, Vassilopoulos S, Bitoun M, 2015   Therapy for dominant inherited diseases by allele-specific RNA interference: successes and pitfalls. Curr. Gene Ther. 15 : 503 – 510 . Google Scholar Crossref Search ADS PubMed WorldCat Truss M , Swat M, Kielbasa S M, Schafer R, Herzel H et al. , 2005   HuSiDa—the human siRNA database: an open-access database for published functional siRNA sequences and technical details of efficient transfer into recipient cells. Nucleic Acids Res. 33 : D108 – D111 . Google Scholar Crossref Search ADS PubMed WorldCat Tyagi A , Ahmed F, Thakur N, Sharma A, Raghava G P et al. , 2011   HIVsirDB: a database of HIV inhibiting siRNAs. PLoS One 6 : e25917 . Google Scholar Crossref Search ADS PubMed WorldCat Ui-Tei K , Naito Y, Takahashi F, Haraguchi T, Ohki-Hamazaki H et al. , 2004   Guidelines for the selection of highly effective siRNA sequences for mammalian and chick RNA interference. Nucleic Acids Res. 32 : 936 – 948 . Google Scholar Crossref Search ADS PubMed WorldCat Venter J C , Adams M D, Myers E W, Li P W, Mural R J et al. , 2001   The sequence of the human genome. Science 291 : 1304 – 1351 . Google Scholar Crossref Search ADS PubMed WorldCat Vert J P , Foveau N, Lajaunie C, Vandenbrouck Y, 2006   An accurate and interpretable model for siRNA efficacy prediction. BMC Bioinformatics 7 : 520 . Google Scholar Crossref Search ADS PubMed WorldCat Watts J K , Corey D R, 2012   Silencing disease genes in the laboratory and the clinic. J. Pathol. 226 : 365 – 379 . Google Scholar Crossref Search ADS PubMed WorldCat Yu D , Pendergraff H, Liu J, Kordasiewicz H B, Cleveland D W et al. , 2012   Single-stranded RNAs use RNAi to potently and allele-selectively inhibit mutant huntingtin expression. Cell 150 : 895 – 908 . Google Scholar Crossref Search ADS PubMed WorldCat © 2017 Monga et al. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2017 Monga et al.
TI - ASPsiRNA: A Resource of ASP-siRNAs Having Therapeutic Potential for Human Genetic Disorders and Algorithm for Prediction of Their Inhibitory Efficacy
JF - "G3: Genes, Genomes, Genetics"
DO - 10.1534/g3.117.044024
DA - 2017-09-01
UR - https://www.deepdyve.com/lp/oxford-university-press/aspsirna-a-resource-of-asp-sirnas-having-therapeutic-potential-for-lNPc9tXye2
SP - 2931
EP - 2943
VL - 7
IS - 9
DP - DeepDyve
ER -