Designing stable Metalloproteins using Deep Learning

Metal ions are essential cofactors for many proteins including enzymes. However it has been difficult to design metalloproteins due to the complicated electronic structure of metals using e.g Rosetta.  With the advent of deep learning based approaches we now have the tools at our disposal to adequately deal with metalloproteins. In this talk, we want to highlight our new Metal3D [1] model that together with a sequence design model [2] allows computational design of new zinc based metalloproteins without specifying the coordination motif and location of the binding site beforehand.  In addition, we will highlight recent advances in sharing deep learning models for proteins on the web.[1] Dürr, S.L., Levy, A. & Rothlisberger, U. Metal3D: a general deep learning framework for accurate metal ion location prediction in proteins. Nat Commun 14, 2713 (2023). https://doi.org/10.1038/s41467-023-37870-6[2] Anand, N., Eguchi, R., Mathews, I.I. et al. Protein sequence design with a learned potential. Nat Commun 13, 746 (2022). https://doi.org/10.1038/s41467-022-28313-9 Better culture when it comes to sharing models and weights