A Bayesian decision procedure for testing multiple hypotheses in DNA microarray experiments.

Area of Science:

• Genomics and Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• DNA microarray experiments involve simultaneous testing of thousands of hypotheses, necessitating robust multiple hypothesis testing procedures.
• Traditional methods can struggle with high-dimensional data, leading to increased error rates such as false positives and false negatives.

Purpose of the Study:

• To develop a Bayesian decision theory framework for multiple hypothesis testing in DNA microarrays.
• To propose a novel decision criterion that estimates the number of false null hypotheses (FNH) and minimizes a specific error measure.

Main Methods:

• A Bayesian decision theory perspective was adopted to address multiple hypothesis testing.
• A new decision criterion was proposed, estimating FNH and using the proportion of posterior expected false positives relative to true null hypotheses as an error measure.
• The methodology was applied to a Gaussian model for bilateral hypothesis testing, validated with simulated and real data.

Main Results:

• The proposed procedure significantly reduced the percentage of false negatives compared to existing Bayes rules.
• The percentage of false positives was maintained at an acceptable level.
• The method demonstrated effectiveness in both simulated and real-world DNA microarray data analysis.

Conclusions:

• The novel Bayesian decision criterion offers an effective approach for multiple hypothesis testing in DNA microarrays.
• This method enhances statistical power by reducing false negatives without compromising the control of false positives.
• The findings suggest improved accuracy and reliability in interpreting large-scale genomic data from microarrays.