Bias in microRNA functional enrichment analysis

Area of Science:

• Bioinformatics
• Molecular Biology
• Genomics

Background:

• MicroRNAs (miRNAs) are crucial regulators of gene expression, with differential expression studied across various biological contexts.
• Predicting shared pathways targeted by miRNAs is a common bioinformatics task.
• Standard functional enrichment tests often yield numerous, sometimes spurious, enriched categories for miRNA targets.

Purpose of the Study:

• To evaluate the appropriateness of standard functional enrichment tests for microRNA target analysis.
• To identify and address biases in current methods for analyzing miRNA target pathways.
• To develop and apply a more robust method for assessing functional enrichment of miRNA targets.

Main Methods:

• Critically assessed the hypergeometric distribution commonly used in functional enrichment analysis for miRNA targets.
• Developed a novel algorithm employing an empirical sampling approach to measure enrichment.
• Reanalyzed gene ontology classes for miRNA targets from 44 published studies using the new algorithm.

Main Results:

• The standard hypergeometric test is inappropriate for miRNA target analysis due to inherent bias, leading to inflated significance.
• Randomly selected miRNA targets frequently show significant enrichment, indicating a systematic issue with the standard method.
• After correcting for bias using the empirical approach, the vast majority of miRNA target sets showed no significant functional enrichment.

Conclusions:

• The standard functional enrichment method is unsuitable for analyzing sets of genes targeted by microRNAs.
• The identified bias in predicted miRNA targets necessitates a reevaluation of current bioinformatics practices.
• The developed empirical sampling approach offers a more reliable method for assessing functional enrichment in miRNA target analysis.