BiGER: Bayesian rank aggregation in genomics with extended ranking schemes

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Large-scale genomic studies generate extensive gene lists relevant to complex diseases.
• Individual study evaluation is limited; robust meta-analytic methods are needed for aggregating evidence.
• Existing methods struggle with the diverse assumptions and protocols inherent in different genomic studies.

Purpose of the Study:

• To introduce BiGER, a novel Bayesian rank aggregation method for inferring latent global gene rankings.
• To address limitations of existing methods by accommodating mixed gene lists, including top-ranked, top-unranked, bottom-tied, and missing genes.
• To provide a flexible yet rigorous approach for aggregating large-scale gene lists in genomic meta-analysis.

Main Methods:

• BiGER employs a Bayesian hierarchical framework for robust statistical inference.
• Variational inference is utilized for efficient aggregation of large gene lists.
• The method is designed to handle complex gene list structures, including missing and tied ranks.

Main Results:

• BiGER demonstrates state-of-the-art accuracy in gene list meta-analysis.
• The method successfully integrates diverse genomic datasets, yielding reliable global rankings.
• BiGER provides valuable insights into the reliability of individual data sources.

Conclusions:

• BiGER is an effective tool for reliable meta-analysis in large-scale genomic studies.
• The method enhances the discovery of disease-relevant genes by synthesizing evidence from multiple sources.
• BiGER offers a flexible and accurate approach to handle complex gene list data in bioinformatics.