Enhancing gene set overrepresentation analysis with large language models.

This observation was highly significant for the GPT-4o model using shared bigrams (Wilcoxon rank sum, P < 10 −13 for KEGG, P < 10 −5 for Reactome, and P < 10 −5 for WikiPathways) and cosine similarity (Wilcoxon rank sum, P < 10 −11 for KEGG, P < 10 −17 for Reactome, and P < 10 −9 for WikiPathways) as evaluation metrics.

also in 132,142 other papers