Benchmark Study Finds LLMs Fall Short of Traditional Tools for Rare Disease Diagnosis

Summary

A comprehensive benchmarking study published in the European Journal of Human Genetics found that current large language models significantly underperform traditional bioinformatics decision support tools — specifically Exomiser — in differential diagnosis of rare genetic diseases. The research, led by Justin T. Reese, Peter N. Robinson, and collaborators from multiple institutions, evaluated a range of best-in-class LLMs used in medical applications against established computational genomics to