A Multitask Deep-Learning Method for Predicting Membrane Associations and Secondary Structures of Proteins

Abstract

Prediction of residue-level structural attributes and protein-level structural classes helps model protein tertiary structures and understand protein functions. Existing methods are either specialized on only one class of proteins or developed to predict only a specific type of residue-level attribute. In this work, we develop a new deep-learning method, named Membrane Association and Secondary Structure Predictor (MASSP), for accurately predicting both residue-level structural attributes (secondary structure, location, orientation, and topology) and protein-level structural classes (bitopic, α-helical, β-barrel, and soluble). MASSP integrates a multilayer two-dimensional convolutional neural network (2D-CNN) with a long short-term memory (LSTM) neural network into a multitasking framework. Our comparison shows that MASSP performs equally well or better than the state-of-the-art methods in predicting residue-level secondary structures, boundaries of transmembrane segments, and topology. Furthermore, it achieves outstanding accuracy in predicting protein-level structural classes. MASSP automatically distinguishes the structural classes of input sequences and identifies transmembrane segments and topologies if present, making it broadly applicable to different classes of proteins. In summary, MASSP's good performance and broad applicability make it well suited for annotating residue-level attributes and protein-level structural classes at the proteome scale.

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

“…Rather than focusing on single property prediction, a few studies have sought to predict a number of properties in combination, such as solvent accessibility, secondary structures, and torsion angles. These methods include AllesTM [242] , MASSP [243] , and TopProperty [244] , which all use deep learning methods to keep abreast of any possible advances in prediction performance ( Table 6 ). For example, in the AllesTM work, the ensemble of conventional machine learning methods (random forest) and deep learning methods (CNNs and bidirectional LSTM NNs) leads to superior performance in predicting Z-coordinates, flexibility, and topology, and its performance in predicting torsion angles, secondary structures, and monomer relative solvent accessibility is roughly similar to that of SPOT-1D.…”

Deep learning for protein secondary structure prediction: Pre and post-AlphaFold

“…Proteins from the Genbank database [80] were used as training data to build the PSSP model developed by Xavier and Thirunavukarasu [81] . Li et al [82] predicted the secondary structure of transmembrane proteins and took the protein sequences from the OPM database [83] .…”

“…CNN-2D is used in PSSP model architecture to extract temporal and spatial features of the input sequences better. Feature vectors (PSSM and one-hot encoding) of a fixed length residue window were employed in [102] , [103] , [104] , [82] as input to the two-dimensional CNN.…”

“…NetSurfP-2.0 [37] was developed using a single model to predict three-state PSSP, eight-state PSSP, relative solvent accessibility, dihedral angles, and protein disorder. MASSP [82] was designed to be able to predict both residue-level structural attributes (secondary structure, location, orientation, and topology) and protein-level structural classes (bitopic, -helical, -barrel, and soluble).…”

Secondary and Topological Structural Merge Prediction of Alpha-Helical Transmembrane Proteins Using a Hybrid Model Based on Hidden Markov and Long Short-Term Memory Neural Networks

“…However, there are only a few methods for the merge prediction of the topological and secondary structures of TMPs. TMPSS [ 23 ] and MASSP [ 24 ] achieved simultaneous prediction by using the DL of Convolutional Neural Networks (CNNs) or Long Short-Term Memory (LSTM) layers, and such studies usually employed multi-task learning to predict structures separately, creating conflicting prediction results. Furthermore, DL is not well suited for modeling temporal phenomena.…”