Fedor Nikolaev

This paper systematically evaluates metrics for protein generative models, identifying challenges and offering recommendations for robust evaluation.

PROSTATA predicts protein stability changes from point mutations using a transformer-based model, outperforming existing neural network approaches.