Natural language processing of gene descriptions for overrepresentation analysis with GeneTEA

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Overrepresentation analysis is a common method for identifying biological enrichment in gene lists.
• Existing tools often struggle with controlling false discovery rates and can produce redundant results.

Purpose of the Study:

• To introduce GeneTEA, a novel model for gene set enrichment analysis.
• To develop a de novo gene set database using natural language processing on gene descriptions.
• To improve the accuracy and reduce redundancy in identifying biological enrichment.

Main Methods:

• GeneTEA ingests free-text gene descriptions.
• It employs natural language processing methods to learn a sparse gene-by-term embedding.
• The model's performance is benchmarked against existing overrepresentation analysis tools.

Main Results:

• GeneTEA effectively controls false discovery rates.
• The model consistently identifies the most relevant biological insights.
• GeneTEA achieves this with reduced redundancy compared to traditional methods.
• The approach is adaptable to other organisms and compounds.

Conclusions:

• GeneTEA offers a powerful and accurate method for overrepresentation analysis.
• The model provides a de novo gene set database derived from gene descriptions.
• GeneTEA improves upon existing tools by controlling false discoveries and minimizing redundancy.
• An interactive app and API are available for the GeneTEA model.