TAGOOS: genome-wide supervised learning of non-coding loci associated to complex phenotypes

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Genome-wide association studies (GWAS) identify single nucleotide polymorphisms (SNPs) linked to complex traits.
• Most SNPs are in non-coding regions, likely regulatory, but linkage disequilibrium (LD) hinders functional SNP identification.
• Prioritizing functional SNPs requires annotation with molecular markers and prediction tools.

Purpose of the Study:

• To develop a supervised model, TAGOOS (TAG SNP bOOSting), for prioritizing putative functional SNPs.
• To integrate associated SNPs, LD blocks, and regulatory features into a predictive model.
• To assess the performance of TAGOOS in identifying functional regulatory SNPs.

Main Methods:

• Integrated associated SNPs, LD blocks, and regulatory features into a supervised model (TAGOOS).
• Computed genome-wide TAGOOS scores to prioritize SNPs.
• Evaluated TAGOOS performance using enrichment and prioritization metrics (odds ratio, AUC).

Main Results:

• TAGOOS scores significantly enriched and prioritized unseen associated SNPs (OR 4.3/3.5, AUC 0.65/0.6 for intronic/intergenic regions).
• TAGOOS scores correlated with SNP significance, expression quantitative trait loci (eQTLs), and regulatory feature annotations.
• Analysis of cleft lip and adult height loci identified known and predicted novel functional regions enriched in relevant transcription factors.

Conclusions:

• A supervised model (TAGOOS) was trained to prioritize putative functional regions based on associated SNPs.
• TAGOOS provides valuable scores and annotations for identifying functional SNPs in complex trait genetics.
• The TAGOOS scores, annotations, and UCSC genome tracks are publicly available for research.