IW-Scoring: an Integrative Weighted Scoring framework for annotating and prioritizing genetic variations in the noncoding genome

Abstract

IW-Scoring represents a new Integrative Weighted Scoring model to annotate and prioritise functionally relevant noncoding variations. The pipeline integrates 11 popular algorithms and outperforms individual methods in three independent data sets, including variants in ClinVar database and GWAS studies, and cancer mutations.Using IW-Scoring, we located 11 recurrently mutated noncoding regions enriched for at least three functional mutations in 14 follicular lymphoma genomes, and validated 9 clusters (82%) in the International Cancer Genome Consortium cohort (n=36), including promoter and enhancer regions of PAX5. IW-Scoring offers greater versatility to identify trait and disease associated noncoding variants.

Dissecting molecular regulatory mechanisms underlying noncoding susceptibility SNPs associated with 19 autoimmune diseases using multi-omics integrative analysis

“…To further assess the performance of our epigenetic functional scoring, we compared the functional support on multiple immune-cell associated regulatory evidence between SNPs prioritized by our method and other five functional scoring methods [11-15]. Table S15 summarized the main characteristics between our method and other scoring methods (see discussion for comparison in detail).…”

“…In this study, we developed a new improved epigenetic functional scoring method to prioritize functional autoimmune SNPs through incorporating hundreds of immune cell-specific active epigenetic information. Some other comparable scoring methods are also developed, such as 3DSNP [13], FIRE [11], GWAS4D [14], IW-Scoring [15] or RegulomeDB [12]. Compared with these approaches, one distinct feature of our method was the integrating of immune cell-specific epigenetic information (Table S15), which might provide better evaluation for disease-specific functional autoimmune SNPs.…”

“…However, the prior labelled training data may be inaccurate and impractical due to the current knowledge limitation in functional roles underlying noncoding SNPs. Some other methods like RegulomeDB [12], 3DSNP [13], GWAS4D [14], IW-Scoring [15], Eigen [16], and FunSeq2 [17] either directly combined various epigenetic/regulatory features to rank SNP functionality or adopted a weighted scoring scheme by considering the relative importance of each feature to assign SNP functionality scores. However, these approaches principally incorporated epigenetic or transcriptional annotation across all cells or tissues, while omitting the cell-specific or context-specific regulation.…”

Whole genome base-wise aggregation and functional prediction for human non-coding regulatory variants

“…Moreover, when predicting effective MPRA alleles, tools learned by deep learning or unsupervised model, such as DeepSEA, GenoCanyon, Eigen_PC and Basset, obtained a higher AUC than our regBase_REG model ( Figure 4F and Supplementary Table S13), probably due to the fact that deep learning and unsupervised methods could capture unknown features that explain the in-vitro activity of regulatory allele. We also evaluated the performance of our newly trained models with existing ensemble methods including IW-Scoring (12) and our previous PRVCS (11). We found that regBase_REG_Common model obtained Supplementary Table S14).…”

“…Building on different predictive assumptions, abundant annotation datasets as well as complementary statistical models, these algorithms have achieved great successes to prioritize functional, pathogenic and cancer-relevant non-coding regulatory variants (7)(8)(9)(10). However, the state-of-the-art benchmarks showed poor concordance among the prediction scores of several existing methods (11)(12)(13). To comprehensively evaluate the regulatory potential or pathogenesis of certain SNV outside the protein-coding region, researchers now have to collect and compare scores from different resources, even need to download huge pre-computed files or manually calculate prediction scores.…”