Recognition of protein allosteric states and residues: Machine learning approaches

Abstract

Allostery is a process by which proteins transmit the effect of perturbation at one site to a distal functional site upon certain perturbation. As an intrinsically global effect of protein dynamics, it is difficult to associate protein allostery with individual residues, hindering effective selection of key residues for mutagenesis studies. The machine learning models including decision tree (DT) and artificial neural network (ANN) models were applied to develop classification model for a cell signaling allosteric protein with two states showing extremely similar tertiary structures in both crystallographic structures and molecular dynamics simulations. Both DT and ANN models were developed with 75% and 80% of predicting accuracy, respectively. Good agreement between machine learning models and previous experimental as well as computational studies of the same protein validates this approach as an alternative way to analyze protein dynamics simulations and allostery. In addition, the difference of distributions of key features in two allosteric states also underlies the population shift hypothesis of dynamics-driven allostery model. © 2018 Wiley Periodicals, Inc.

Deciphering the Allosteric Process of the Phaeodactylum tricornutum Aureochrome 1a LOV Domain

“…However, in multi-task classification jobs, one-vs-one random forest model is more common and superior than random forest model by constructing one classifier for each pair of classes. 46 The overall output is the weighted sum of all base classifiers. In the current study, 10 macrostates were trained with 45 random forest models.…”

Predicted Hotspot Residues Involved in Allosteric Signal Transmission in Pro-Apoptotic Peptide—Mcl1 Complexes

“…Despite, our MD simulations exhibited the favorable energy contributing residues at the CBG of mMcl1—PAP complexes, the mechanism of the collective internal motion of these residues involved AST remain unclear. This investigation might provide the molecular origin of the regulation, network of interaction across dimer interface, and putative allosteric path from one site to distal functional site [ 77 ]. For this, the advanced post-processing method was applied to MD trajectory ( Figure 10 ).…”

Allosteric Regulation at the Crossroads of New Technologies: Multiscale Modeling, Networks, and Machine Learning

“…Dimension reduction is often performed using time-lagged independent component analysis (TICA) (Schwantes and Pande, 2014 ; Perez-Hernandez and Noe, 2016 ; Noe and Clementi, 2017 ; Olsson et al, 2017 ). In these approaches, the simulation samples can be divided into substates assuming that conformations within each substate share kinetic similarity and could interconvert rapidly (Bowman et al, 2009 ; Zhou and Tao, 2018 ; Zhou et al, 2018a , b ). t-SNE method was recently developed as a dimensionality reduction method with minimum structural information loss revealing that both one-dimensional (1D) and two-dimensional (2D) models of t-SNE method are superior to other tools in distinguishing functional states of allosteric proteins (Zhou et al, 2018a , b ).…”

“…In these approaches, the simulation samples can be divided into substates assuming that conformations within each substate share kinetic similarity and could interconvert rapidly (Bowman et al, 2009 ; Zhou and Tao, 2018 ; Zhou et al, 2018a , b ). t-SNE method was recently developed as a dimensionality reduction method with minimum structural information loss revealing that both one-dimensional (1D) and two-dimensional (2D) models of t-SNE method are superior to other tools in distinguishing functional states of allosteric proteins (Zhou et al, 2018a , b ). MSMs and transition network models are widely applied to extract kinetic descriptors from equilibrium simulations.…”

“…The integration of Markov modeling, simulations and ML approaches into robust and reproducible computational pipelines with the experimental feedback can be explored for atomistic modeling and classification of allosteric states ( Figure 3 ). Several ML algorithms including decision tree and artificial neural networks were employed in combination with MSM approaches to develop classification models of functionally relevant allosteric conformations that exhibit very similar tertiary structures (Zhou et al, 2018a , b ). Despite the lack of significant conformational change between allosteric states of the second PDZ domain (PDZ2) in the human PTP1E protein, which is a prototypical example of dynamics-driven allostery, it was demonstrated that both algorithms could build effective prediction models and provide reliable quantitative evaluation of the contributions from individual residues to the difference between the two allosteric states (Zhou et al, 2018a , b ).…”