A latent model for prioritization of SNPs for functional studies

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Prioritizing single nucleotide polymorphisms (SNPs) from genetic association studies for functional analysis is challenging.
• Current methods often rely solely on p-values, limiting the integration of diverse data types.
• Handling results from multiple complex disease studies and incorporating biological knowledge remains difficult.

Purpose of the Study:

• To introduce a Bayesian latent variable model (BLVM) for prioritizing SNPs.
• To systematically integrate SNP features and biological knowledge for improved marker selection.
• To estimate a latent quality score for each SNP and rank them based on posterior probabilities.

Main Methods:

• Development and application of a Bayesian latent variable model (BLVM).
• Incorporation of SNP features (e.g., location in gene exons or regulatory regions).
• Estimation of SNP quality scores and their posterior probability distributions for ranking.

Main Results:

• The BLVM successfully prioritized SNPs in an ovarian cancer genome-wide association study (GWAS).
• The top-ranked SNP by BLVM showed improved ranking compared to p-value-based methods across different analyses.
• Application to simulated data demonstrated the model's utility in multi-GWAS marker prioritization.

Conclusions:

• The BLVM offers a systematic approach to integrate multiple SNP features for prioritization.
• This method accounts for ranking uncertainty, enhancing the selection of loci for fine-mapping and functional studies.
• The BLVM provides a robust framework for leveraging diverse data in genetic research.