Using machine learning to detect the differential usage of novel gene isoforms.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Differential gene isoform usage contributes to phenotypic diversity and disease.
• Accurate detection is challenging, especially in under-annotated genomes.

Purpose of the Study:

• To evaluate machine learning for detecting differential isoform usage using read distribution data.
• To identify genes with distinct transcript usage between European and African populations.

Main Methods:

• Applied gradient boosting and elastic net machine learning models.
• Analyzed read distribution across gene regions to infer isoform usage.
• Compared findings with existing differential expression analysis methods.

Main Results:

• Machine learning successfully identified numerous genes with differential isoform usage between populations.
• Detected differences were enriched in relevant biological pathways.
• Transcript diversity at the 3' and 5' gene ends significantly drove population differences.

Conclusions:

• Machine learning provides an effective method for detecting differential isoform usage from read fraction data.
• This approach offers novel insights into population-specific biological variations.