Data for DeepTMInter: Improved sequence-based prediction of interaction sites in α-helical transmembrane proteins by deep learning

Published: 24-12-2020| Version 2 | DOI: 10.17632/2t8kgwzp35.2
Contributors:
,
Dmitrij Frishman

Description

It contains five folders: 1). PDB. Crystallographic structures of TM proteins in the TrainData, IndepData, TestData, and CompData datasets. 2). Fasta. Sequences of TM proteins in the TrainData, IndepData, TestData, and CompData datasets. 3). Topology. Structure-derived and Phobius-predicted topologies of TM proteins in the TrainData, IndepData, TestData, and CompData datasets. 4). Cross Validation. 5-fold cross validations of the TrainData dataset by the stratified-shuffle method. 5). Prediction. Predictions of the 7 tools (DeepTMInter, DELPHI, MBPredAll, MBPredCombined, MBPredCyto, MBPredExtra, and MBPredTM) on the IndepData, TestData, and CompData datasets. Predictions of DeepTMInter on the human transmembrane proteome (5051 proteins). In addition, ten supplementary tables (19-28) in Excel format for the paper titled "Improved sequence-based prediction of interaction sites in α-helical transmembrane proteins by deep learning" are also available here.

Files