Bayesian decision theoretic multiple comparison procedures: an application to phage display data.

Area of Science:

• Bioinformatics
• Statistical Inference
• Computational Biology

Background:

• Massive multiple comparisons present challenges in analyzing large biological datasets.
• Phage display experiments generate complex data requiring robust statistical methods.
• Identifying high-affinity ligands is crucial for biological and medical research.

Purpose of the Study:

• To evaluate a Bayesian decision theoretic approach for massive multiple comparisons.
• To address limitations in identifying significant biological signals in multi-stage experiments.
• To develop improved inference methods for high-throughput biological data.

Main Methods:

• Applied a principled Bayesian decision theoretic framework to analyze tripeptide count data from a mouse phage display experiment.
• Investigated an approach controlling the posterior expected false discovery rate.
• Developed a second approach incorporating a utility function with weights for the magnitude of biological signal increase.

Main Results:

• The initial Bayesian approach, while principled, overlooked the magnitude of observed increases.
• The refined utility-function-based approach offers a more nuanced selection of significant peptide-tissue pairs.
• Demonstrated the practical application and limitations of these Bayesian methods in a real-world case study.

Conclusions:

• Bayesian decision theory provides a flexible framework for massive multiple comparisons in biological research.
• Accounting for the magnitude of biological effects is essential for effective ligand identification.
• The proposed utility-based method enhances the identification of biologically relevant findings from complex datasets.