An empirical Bayesian ranking method, with applications to high throughput biology

Area of Science:

• Bioinformatics
• Genomics
• Statistical Genetics

Background:

• Genome-wide experiments identify biological differences between groups across many genomic locations.
• Traditional P-value based ranking may prioritize small effect sizes over larger, more significant biological findings.

Purpose of the Study:

• To develop a robust ranking algorithm for genomic locations in bioinformatics.
• To address the limitations of P-value based ranking in identifying biologically important effects.

Main Methods:

• Developed an Empirical Bayes ranking algorithm using the marginal distribution of data to estimate prior distributions.
• Algorithm utilizes the collective distribution of effect sizes for improved prior estimation.

Main Results:

• Demonstrated superior performance compared to P-value ranking and other methods via simulations and real data analysis.
• The algorithm is computationally efficient and adaptable for full or subset genomic location ranking.

Conclusions:

• Empirical Bayes ranking offers a more effective approach for prioritizing biologically significant genomic regions.
• The EBrank R-package provides an accessible implementation of this advanced ranking method.