Efficient protein structure prediction fromcompact computers to datacenters withOpenFold-TRT

Kieran Didi; Prashant Sohani; Fabian Berressem; Alexander Nesterovskiy; Boris Fomitchev; Robert Ohannessian; Mohamed Elbalkini; Jonathan Cogan; Anthony B. Costa; Arash Vahdat; Felix Kallenborn; Bertil Schmidt; Milot Mirdita; Martin Steinegger; Christian Dallago; Alejandro Chacón

doi:10.64898/2026.03.11.711233

BioinformaticsbioRxivMar 15, 2026

Preprint

Efficient protein structure prediction fromcompact computers to datacenters withOpenFold-TRT

Kieran Didi iD, Prashant Sohani iD, Fabian Berressem iD, Alexander Nesterovskiy, Boris Fomitchev, Robert Ohannessian, Mohamed Elbalkini, Jonathan Cogan, Anthony B. Costa iD, Arash Vahdat iD, Felix Kallenborn iD, Bertil Schmidt iD, Milot Mirdita iD, Martin Steinegger iD, Christian Dallago iD, Alejandro Chacón iD

Abstract

We introduce accelerations for deep learning inference with OpenFold and TensorRT that, combined with MMseqs2-GPU on an x86 system with one NVIDIA RTX PRO 6000 Blackwell Server Edition GPU, reach up to 131x faster inference compared to AlphaFold2. ARM-optimizations enable homology search on power efficient DGX Spark, and allow search beyond available GPU RAM on NVIDIA Grace Hopper Superchip at comparable overall folding speed. These accelerations enable high-throughput protein structure inference at no accuracy cost using familiar tools.

PDF DOI