Abstract:Experimental determination of membrane protein (MP) structures is challenging as they are often too large for nuclear magnetic resonance (NMR) experiments and difficult to crystallize. Currently there are only about 510 non-redundant MPs with solved structures in Protein Data Bank (PDB). To elucidate the MP structures computationally, we developed a novel web resource, denoted as PredMP (http://52.87.130.56:3001/#/proteinindex), that delivers one-dimensional (1D) annotation of the membrane topology and secondary structure, two-dimensional (2D) prediction of the contact/distance map, together with three-dimensional (3D) modeling of the MP structure in the lipid bilayer, for each MP target from a given model organism. The precision of the computationally constructed MP structures is leveraged by state-of-the-art deep learning methods as well as cutting-edge modeling strategies. In particular, (i) we annotate 1D property via DeepCNF (Deep Convolutional Neural Fields) that not only models complex sequence-structure relationship but also interdependency between adjacent property labels; (ii) we predict 2D contact/distance map through Deep Transfer Learning which learns the patterns as well as the complex relationship between contacts/distances and protein features from non-membrane proteins; and (iii) we model 3D structure by feeding its predicted contacts and secondary structure to the Crystallography & NMR System (CNS) suite combined with a membrane burial potential that is residue-specific and depth-dependent. PredMP currently contains more than 2,200 multi-pass transmembrane proteins (length<700 residues) from Human. These transmembrane proteins are classified according to IUPHAR/BPS Guide, which provides a hierarchical organization of receptors, channels, transporters, enzymes and other drug targets according to their molecular relationships and physiological functions. Among these MPs, we estimated that our approach could predict correct folds for 1,345-1,871 targets including a few hundred new folds, which shall facilitate the discovery of drugs targeting at MPs.

Deep Conditional Random Field Approach to Transmembrane Topology Prediction and Application to GPCR Three-Dimensional Structure Modeling.

CRF-TM: A Conditional Random Field Method for Predicting Transmembrane Topology.

Structure Prediction of Membrane Proteins

Topology Prediction Improvement of Α-Helical Transmembrane Proteins Through Helix-tail Modeling and Multiscale Deep Learning Fusion.

High-accuracy prediction of transmembrane inter-helix contacts and application to GPCR 3D structure modeling.

An Improved Topology Prediction of Alpha-Helical Transmembrane Protein Based on Deep Multi-Scale Convolutional Neural Network

Predicting Helix Boundaries of Α-Helix Transmembrane Protein with Feedback Conditional Random Fields.

DMCTOP: Topology Prediction of Alpha-Helical Transmembrane Protein Based on Deep Multi-Scale Convolutional Neural Network

Prediction of A. thaliana’s MCTP4 Structure using Deep Learning-Based tools and Exploration of Transmembrane domain Dynamics using Coarse-Grained Molecular Dynamics Simulations

Efficient and accurate prediction of transmembrane topology from amino acid sequence only

Machine learning in computational modelling of membrane protein sequences and structures: from methodologies to applications

Predmp: A Web Resource For Computationally Predicted Membrane Proteins Via Deep Learning

TMPSS: A Deep Learning-Based Predictor for Secondary Structure and Topology Structure Prediction of Alpha-Helical Transmembrane Proteins

Prediction of membrane protein types in a hybrid space.

Prediction of Membrane Protein Interaction Based on Deep Residual Learning.

Transmembrane Region Prediction by Using Sequence-Derived Features and Machine Learning Methods

Folding membrane proteins by deep transfer learning

Transmembrane Topology Identification by Fusing Evolutionary and Co-evolutionary Information with Cascaded Bidirectional Transformers

Predicting Residue-Residue Contacts And Helix-Helix Interactions In Transmembrane Proteins Using An Integrative Feature-Based Random Forest Approach

Deep Transfer Learning for Inter-Chain Contact Predictions of Transmembrane Protein Complexes

Tmp-Ssurface: A Deep Learning-Based Predictor For Surface Accessibility Of Transmembrane Protein Residues