A multi-dimensional integrative scoring framework for predicting functional variants in the human genome

Abstract

Attempts to identify and prioritize functional DNA elements in coding and noncoding regions, particularly through use of in silico functional annotation data, continue to increase in popularity. However, specific functional roles may vary widely from one variant to another, making it challenging to summarize different aspects of variant function. Here we propose Multi-dimensional Annotation Class Integrative Estimation (MACIE), an unsupervised multivariate mixed model framework capable of integrating annotations of diverse origin to assess multi-dimensional functional roles for both coding and noncoding variants. Unlike existing one-dimensional scoring methods, MACIE views variant functionality as a composite attribute encompassing multiple different characteristics, and estimates the joint posterior functional probability vector of each genomic position, a quantity that offers richer and more interpretable information in the presence of multiple aspects of functionality. Applied to a variety of independent coding and non-coding datasets, MACIE demonstrates powerful and robust performance in discriminating between functional and non-functional variants. We also show an application of MACIE to fine-mapping using lipids GWAS summary statistics data from the European Network for Genetic and Genomic Epidemiology Consortium.

FAVOR: Functional Annotation of Variants Online Resource and Annotator for Variation across the Human Genome

“…A variety of functional annotations have been developed to measure multiple aspects of biological functionality of variants, including protein function (15)(16)(17), conservation (18,19), epigenetics (20,21), spatial genomics (22,23), network biology (24), mappability (25), local nucleotide diversity (26) and integrative composite annotations (4,(27)(28)(29). These annotations have successfully prioritized plausible causal variants of underlying GWAS signals according to their functional impact in experimental studies following GWAS findings (5,30), localizing causal variants in fine-mapping studies (4,8), partitioning heritability in GWAS (6), predicting genetic risk (6,7,9), and improving rare variant (RV) analysis of WGS association studies (12)(13)(14)31). For example, large-scale WGS/WES studies (1,3,32,33) assess the associations between complex diseases/traits and coding and non-coding rare variants across the genome.…”

“…For example, large-scale WGS/WES studies (1,3,32,33) assess the associations between complex diseases/traits and coding and non-coding rare variants across the genome. The recently developed STAAR method incorporates multi-faceted variant functional annotations to boost the power of rare variant association tests in WGS/WES studies (12)(13)(14).…”

“…Existing functional annotation databases and tools are not scalable for functionally annotating a massive number of variants in large-scale WGS/WES studies. Moreover, few of the currently available functional annotation tools can provide organized output in a format that is both storage efficient and ready to be used in downstream statistical genetic analyses, such as fine-mapping (4,11), heritability (6), rare-variants association tests (13,14). There is a pressing community need to develop a convenient and comprehensive functional annotation tool that annotates any WGS study dataset at scale and generates a functionally annotated genotype file in an organized and compressed format, that can be readily integrated into the downstream analysis.…”

“…Variant functional annotation represents a powerful and effective approach to leveraging functional information from many different bioinformatics sources to elucidate the multi-faceted functions of genetic variants for a wide range of analyses of WGS/WES studies and Genome-Wide Association Studies (GWAS) (4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). A variety of functional annotations have been developed to measure multiple aspects of biological functionality of variants, including protein function (15)(16)(17), conservation (18,19), epigenetics (20,21), spatial genomics (22,23), network biology (24), mappability (25), local nucleotide diversity (26) and integrative composite annotations (4,(27)(28)(29).…”

S-BEAM: A Semi-Supervised Ensemble Approach to Rank Potential Causal Variants and Their Target Genes in Microglia for Alzheimer’s Disease

“…Therefore, models that employ PN approaches can suffer from high false-negative prediction rates (Kolosov et al, 2021). Recently, unsupervised and semi-supervised techniques have demonstrated an advantage in the analysis of non-coding regions because they are not biased by the lack of available labeled training examples (Li et al, 2022). Particularly, positive-unlabeled (PU) and negative-unlabeled (NU) learning are semi-supervised methods suited to genomic classification problems, as they can train a model with limited high-quality labeled data.…”