Accurate deleteriousness prediction for nonsynonymous variants is crucial for distinguishing pathogenic mutations from background polymorphisms in whole exome sequencing (WES) studies. Although many deleteriousness prediction methods have been developed, their prediction results are sometimes inconsistent with each other and their relative merits are still unclear in practical applications. To address these issues, we comprehensively evaluated the predictive performance of 18 current deleteriousness-scoring methods, including 11 function prediction scores (PolyPhen-2, SIFT, MutationTaster, Mutation Assessor, FATHMM, LRT, PANTHER, PhD-SNP, SNAP, SNPs&GO and MutPred), 3 conservation scores (GERP++, SiPhy and PhyloP) and 4 ensemble scores (CADD, PON-P, KGGSeq and CONDEL). We found that FATHMM and KGGSeq had the highest discriminative power among independent scores and ensemble scores, respectively. Moreover, to ensure unbiased performance evaluation of these prediction scores, we manually collected three distinct testing datasets, on which no current prediction scores were tuned. In addition, we developed two new ensemble scores that integrate nine independent scores and allele frequency. Our scores achieved the highest discriminative power compared with all the deleteriousness prediction scores tested and showed low false-positive prediction rate for benign yet rare nonsynonymous variants, which demonstrated the value of combining information from multiple orthologous approaches. Finally, to facilitate variant prioritization in WES studies, we have pre-computed our ensemble scores for 87 347 044 possible variants in the whole-exome and made them publicly available through the ANNOVAR software and the dbNSFP database.

The purpose of the dbNSFP is to provide a one‐stop resource for functional predictions and annotations for human nonsynonymous single‐nucleotide variants (nsSNVs) and splice‐site variants (ssSNVs), and to facilitate the steps of filtering and prioritizing SNVs from a large list of SNVs discovered in an exome‐sequencing study. A list of all potential nsSNVs and ssSNVs based on the human reference sequence were created and functional predictions and annotations were curated and compiled for each SNV. Here, we report a recent major update of the database to version 3.0. The SNV list has been rebuilt based on GENCODE 22 and currently the database includes 82,832,027 nsSNVs and ssSNVs. An attached database dbscSNV, which compiled all potential human SNVs within splicing consensus regions and their deleteriousness predictions, add another 15,030,459 potentially functional SNVs. Eleven prediction scores (MetaSVM, MetaLR, CADD, VEST3, PROVEAN, 4× fitCons, fathmm‐MKL, and DANN) and allele frequencies from the UK10K cohorts and the Exome Aggregation Consortium (ExAC), among others, have been added. The original seven prediction scores in v2.0 (SIFT, 2× Polyphen2, LRT, MutationTaster, MutationAssessor, and FATHMM) as well as many SNV and gene functional annotations have been updated. dbNSFP v3.0 is freely available at http://sites.google.com/site/jpopgen/dbNSFP.

With the advance of sequencing technologies, whole exome sequencing has increasingly been used to identify mutations that cause human diseases, especially rare Mendelian diseases. Among the analysis steps, functional prediction (of being deleterious) plays an important role in filtering or prioritizing nonsynonymous SNP (NS) for further analysis. Unfortunately, different prediction algorithms use different information and each has its own strength and weakness. It has been suggested that investigators should use predictions from multiple algorithms instead of relying on a single one. However, querying predictions from different databases/Web‐servers for different algorithms is both tedious and time consuming, especially when dealing with a huge number of NSs identified by exome sequencing. To facilitate the process, we developed dbNSFP (database for nonsynonymous SNPs' functional predictions). It compiles prediction scores from four new and popular algorithms (SIFT, Polyphen2, LRT, and MutationTaster), along with a conservation score (PhyloP) and other related information, for every potential NS in the human genome (a total of 75,931,005). It is the first integrated database of functional predictions from multiple algorithms for the comprehensive collection of human NSs. dbNSFP is freely available for download at http://sites.google.com/site/jpopgen/dbNSFP. Hum Mutat 32:894–899, 2011.

Whole exome sequencing has been increasingly used in human disease studies. Prioritization based on appropriate functional annotations has been used as an indispensable step to select candidate variants. Here we present the latest updates to dbNSFP (version 4.1), a database designed to facilitate this step by providing deleteriousness prediction and functional annotation for all potential nonsynonymous and splice-site SNVs (a total of 84,013,093) in the human genome. The current version compiled 36 deleteriousness prediction scores, including 12 transcript-specific scores, and other variant and gene-level functional annotations. The database is available at http://database.liulab.science/dbNSFP with a downloadable version and a web-service.

DbNSFP is a database developed for functional prediction and annotation of all potential non‐synonymous single‐nucleotide variants (nsSNVs) in the human genome. This database significantly facilitates the process of querying predictions and annotations from different databases/web‐servers for large amounts of nsSNVs discovered in exome‐sequencing studies. Here we report a recent major update of the database to version 2.0. We have rebuilt the SNV collection based on GENCODE 9 and currently the database includes 87,347,043 nsSNVs and 2,270,742 essential splice site SNVs (an 18% increase compared to dbNSFP v1.0). For each nsSNV dbNSFP v2.0 has added two prediction scores (MutationAssessor and FATHMM) and two conservation scores (GERP++ and SiPhy). The original five prediction and conservation scores in v1.0 (SIFT, Polyphen2, LRT, MutationTaster and PhyloP) have been updated. Rich functional annotations for SNVs and genes have also been added into the new version, including allele frequencies observed in the 1000 Genomes Project phase 1 data and the NHLBI Exome Sequencing Project, various gene IDs from different databases, functional descriptions of genes, gene expression and gene interaction information, among others. dbNSFP v2.0 is freely available for download at http://sites.google.com/site/jpopgen/dbNSFP. ©2013 Wiley‐Liss, Inc.

In silico tools have been developed to predict variants that may have an impact on pre-mRNA splicing. The major limitation of the application of these tools to basic research and clinical practice is the difficulty in interpreting the output. Most tools only predict potential splice sites given a DNA sequence without measuring splicing signal changes caused by a variant. Another limitation is the lack of large-scale evaluation studies of these tools. We compared eight in silico tools on 2959 single nucleotide variants within splicing consensus regions (scSNVs) using receiver operating characteristic analysis. The Position Weight Matrix model and MaxEntScan outperformed other methods. Two ensemble learning methods, adaptive boosting and random forests, were used to construct models that take advantage of individual methods. Both models further improved prediction, with outputs of directly interpretable prediction scores. We applied our ensemble scores to scSNVs from the Catalogue of Somatic Mutations in Cancer database. Analysis showed that predicted splice-altering scSNVs are enriched in recurrent scSNVs and known cancer genes. We pre-computed our ensemble scores for all potential scSNVs across the human genome, providing a whole genome level resource for identifying splice-altering scSNVs discovered from large-scale sequencing studies.