seqLens: Optimizing Language Models for Genomic Predictions

A new study from the Computational Biology Institute has been published in Molecular Biology and Evolution. The research introduces seqLens, a framework for systematically evaluating genomic language models and identifying the design choices that most strongly influence performance. The study provides practical guidance for developing more accurate and efficient AI methods for genomics and advances research at the intersection of artificial intelligence, evolutionary biology, and computational biology.

The work was sponsored by the National Science Foundation, led by graduate student Mahdi Baghbanzadeh and researcher Brendan Mann, and designed and supervised by Professors Ali Rahnavard and Keith A. Crandall.

Paper: https://academic.oup.com/mbe/article/43/7/msag139/8715817