Work

BeoFold: Simulating How Proteins Fold Using Distributed Computing and State Networks

Year: 2026
Type: preprint
Abstract: BeoFold (v1.4.0) is an open-source molecular simulation and cluster-networking program built to investigate how short amino-acid chains fold into stable, functional three-dimensional protein structures. Simulating protein folding inside a single conventional computer at normal biological temperatures is remarkably slow because flexible molecules spend almost all of their time bouncing inside shallow, partially folded pockets separated by tall energy walls. BeoFold solves that sampling trap by orchestrating parallel replicas across multiple temperatures—allowing warm copies to cross energy barriers and trade shapes cleanly with cooler copies—while optionally linking ordinary lab desktops over encrypted QUIC and Raft consensus networking. Once snapshots accumulate, BeoFold converts thousands of noisy atomic coordinates into a clear network of stable states and folding speeds. On the classic 20-amino-acid Trp-cage model protein, BAOAB thermal integration held kinetic motion steady near 282.3828 K, four temperature tiers from 285 K to 405 K achieved a 47.89% exchange rate down to 156.6821 kcal/mol, and spectral state clustering mapped three distinct structural basins with near-exact time-reversibility. CRV Science Apps Hub: https://crvscience.com/appsCanonical Preprint URL: https://crvscience.com/apps/preprints/beofold-distributed-protein-msmMaster Concept DOI: 10.5281/zenodo.24222370Repository: https://github.com/bpwhite/crvscience-beofold
Institution: CRV (Netherlands)
Language: English
Open Access status: green
OpenAlex ID: W7233597410
Cites:
Cited by:
Related to: