Predictions - Domain-Peptide and Protein-Protein Interactions - HSM
datasetposted on 06.01.2020, 14:21 by Joseph Cunningham
Protein interaction predictions derived from hierarchical statistical mechanical modelling, a machine learning framework used to approximate biophysical models.
Data are broken into two parts - protein level predictions (hsm_p.predictions.csv) and domain level predictions (hsm_d.predictions.tar.gz). Protein-protein predictions are triplets of two UniProt ids and the likelihood of binding. Domain-peptide interactions are chunked (arbitrarily) over JSON files. The JSON files are lists with each item containing the domain-peptide interactions of a single protein-protein interaction in the form:
UID1,UID2 ,[[DTYPE, DUID1 , DSEQ , PSEQ, VAL ], ...]
Cunningham JM, Koytiger G, Sorger PK, AlQuraishi M. "Biophysical prediction of protein-peptide interactions and signaling networks using machine learning." Nature Methods (2020).