Protein conformational flexibility prediction using machine learning

Prediction of order parameters based on protein NMR structure ensemble and machine learning

The fast motions of proteins at the picosecond to nanosecond timescale, known as fast dynamics, are closely related to protein conformational entropy and rearrangement, which in turn affect catalysis, ligand binding and protein allosteric effects. The most used NMR approach to study fast protein dynamics is the model free method, which uses order parameter S2 to describe the amplitude of the internal motion of local group. However, to obtain order parameter through NMR experiments is quite complex and lengthy. In this paper, we present a machine learning approach for predicting backbone 1H-15N order parameters based on protein NMR structure ensemble. A random forest model is used to learn the relationship between order parameters and structural features. Our method achieves high accuracy in predicting backbone 1H-15N order parameters for a test dataset of 10 proteins, with a Pearson correlation coefficient of 0.817 and a root-mean-square error of 0.131.

Protein Function Analysis through Machine Learning

Machine learning (ML) has been an important arsenal in computational biology used to elucidate protein function for decades. With the recent burgeoning of novel ML methods and applications, new ML approaches have been incorporated into many areas of computational biology dealing with protein function. We examine how ML has been integrated into a wide range of computational models to improve prediction accuracy and gain a better understanding of protein function. The applications discussed are protein structure prediction, protein engineering using sequence modifications to achieve stability and druggability characteristics, molecular docking in terms of protein-ligand binding, including allosteric effects, protein-protein interactions and protein-centric drug discovery. To quantify the mechanisms underlying protein function, a holistic approach that takes structure, flexibility, stability, and dynamics into account is required, as these aspects become inseparable through their interdependence. Another key component of protein function is conformational dynamics, which often manifest as protein kinetics. Computational methods that use ML to generate representative conformational ensembles and quantify differences in conformational ensembles important for function are included in this review. Future opportunities are highlighted for each of these topics.

Detection of chemical exchange in methyl groups of macromolecules

The zero- and double-quantum methyl TROSY Hahn-echo and the methyl ¹H-¹H dipole-dipole cross-correlation nuclear magnetic resonance experiments enable estimation of multiple quantum chemical exchange broadening in methyl groups in proteins. The two relaxation rate constants are established to be linearly dependent using molecular dynamics simulations and empirical analysis of experimental data. This relationship allows chemical exchange broadening to be recognized as an increase in the Hahn-echo relaxation rate constant. The approach is illustrated by analyzing relaxation data collected at three temperatures for E. coli ribonuclease HI and by analyzing relaxation data collected for different cofactor and substrate complexes of E. coli AlkB.

In silico prediction of protein flexibility with local structure approach

Flexibility is an intrinsic essential feature of protein structures, directly linked to their functions. To this day, most of the prediction methods use the crystallographic data (namely B-factors) as the only indicator of protein's inner flexibility and predicts them as rigid or flexible. PredyFlexy stands differently from other approaches as it relies on the definition of protein flexibility (i) not only taken from crystallographic data, but also (ii) from Root Mean Square Fluctuation (RMSFs) observed in Molecular Dynamics simulations. It also uses a specific representation of protein structures, named Long Structural Prototypes (LSPs). From Position-Specific Scoring Matrix, the 120 LSPs are predicted with a good accuracy and directly used to predict (i) the protein flexibility in three categories (flexible, intermediate and rigid), (ii) the normalized B-factors, (iii) the normalized RMSFs, and (iv) a confidence index. Prediction accuracy among these three classes is equivalent to the best two class prediction methods, while the normalized B-factors and normalized RMSFs have a good correlation with experimental and in silico values. Thus, PredyFlexy is a unique approach, which is of major utility for the scientific community. It support parallelization features and can be run on a local cluster using multiple cores.

Intrinsically disordered proteins in various hypotheses on the pathogenesis of Alzheimer's and Parkinson's diseases

Amyloid-β (Aβ) and α-synuclein (αS) are two intrinsically disordered proteins (IDPs) at the centers of the pathogenesis of Alzheimer's and Parkinson's diseases, respectively. Different hypotheses have been proposed for explanation of the molecular mechanisms of the pathogenesis of these two diseases, with these two IDPs being involved in many of these hypotheses. Currently, we do not know, which of these hypothesis is more accurate. Experiments face challenges due to the rapid conformational changes, fast aggregation processes, solvent and paramagnetic effects in studying these two IDPs in detail. Furthermore, pathological modifications impact their structures and energetics. Theoretical studies using computational chemistry and computational biology have been utilized to understand the structures and energetics of Aβ and αS. In this chapter, we introduce Aβ and αS in light of various hypotheses, and discuss different experimental and theoretical techniques that are used to study these two proteins along with their weaknesses and strengths. We suggest that a promising solution for studying Aβ and αS at the center of varying hypotheses could be provided by developing new techniques that link quantum mechanics, statistical mechanics, thermodynamics, bioinformatics to machine learning. Such new developments could also lead to development in experimental techniques.

Protein flexibility in the light of structural alphabets

Protein structures are valuable tools to understand protein function. Nonetheless, proteins are often considered as rigid macromolecules while their structures exhibit specific flexibility, which is essential to complete their functions. Analyses of protein structures and dynamics are often performed with a simplified three-state description, i.e., the classical secondary structures. More precise and complete description of protein backbone conformation can be obtained using libraries of small protein fragments that are able to approximate every part of protein structures. These libraries, called structural alphabets (SAs), have been widely used in structure analysis field, from definition of ligand binding sites to superimposition of protein structures. SAs are also well suited to analyze the dynamics of protein structures. Here, we review innovative approaches that investigate protein flexibility based on SAs description. Coupled to various sources of experimental data (e.g., B-factor) and computational methodology (e.g., Molecular Dynamic simulation), SAs turn out to be powerful tools to analyze protein dynamics, e.g., to examine allosteric mechanisms in large set of structures in complexes, to identify order/disorder transition. SAs were also shown to be quite efficient to predict protein flexibility from amino-acid sequence. Finally, in this review, we exemplify the interest of SAs for studying flexibility with different cases of proteins implicated in pathologies and diseases.

The DynaMine webserver: predicting protein dynamics from sequence

Protein dynamics are important for understanding protein function. Unfortunately, accurate protein dynamics information is difficult to obtain: here we present the DynaMine webserver, which provides predictions for the fast backbone movements of proteins directly from their amino-acid sequence. DynaMine rapidly produces a profile describing the statistical potential for such movements at residue-level resolution. The predicted values have meaning on an absolute scale and go beyond the traditional binary classification of residues as ordered or disordered, thus allowing for direct dynamics comparisons between protein regions. Through this webserver, we provide molecular biologists with an efficient and easy to use tool for predicting the dynamical characteristics of any protein of interest, even in the absence of experimental observations. The prediction results are visualized and can be directly downloaded. The DynaMine webserver, including instructive examples describing the meaning of the profiles, is available at http://dynamine.ibsquare.be.

PredyFlexy: flexibility and local structure prediction from sequence

Protein structures are necessary for understanding protein function at a molecular level. Dynamics and flexibility of protein structures are also key elements of protein function. So, we have proposed to look at protein flexibility using novel methods: (i) using a structural alphabet and (ii) combining classical X-ray B-factor data and molecular dynamics simulations. First, we established a library composed of structural prototypes (LSPs) to describe protein structure by a limited set of recurring local structures. We developed a prediction method that proposes structural candidates in terms of LSPs and predict protein flexibility along a given sequence. Second, we examine flexibility according to two different descriptors: X-ray B-factors considered as good indicators of flexibility and the root mean square fluctuations, based on molecular dynamics simulations. We then define three flexibility classes and propose a method based on the LSP prediction method for predicting flexibility along the sequence. This method does not resort to sophisticate learning of flexibility but predicts flexibility from average flexibility of predicted local structures. The method is implemented in PredyFlexy web server. Results are similar to those obtained with the most recent, cutting-edge methods based on direct learning of flexibility data conducted with sophisticated algorithms. PredyFlexy can be accessed at http://www.dsimb.inserm.fr/dsimb_tools/predyflexy/.

An introduction to NMR-based approaches for measuring protein dynamics

Proteins are inherently flexible at ambient temperature. At equilibrium, they are characterized by a set of conformations that undergo continuous exchange within a hierarchy of spatial and temporal scales ranging from nanometers to micrometers and femtoseconds to hours. Dynamic properties of proteins are essential for describing the structural bases of their biological functions including catalysis, binding, regulation and cellular structure. Nuclear magnetic resonance (NMR) spectroscopy represents a powerful technique for measuring these essential features of proteins. Here we provide an introduction to NMR-based approaches for studying protein dynamics, highlighting eight distinct methods with recent examples, contextualized within a common experimental and analytical framework. The selected methods are (1) Real-time NMR, (2) Exchange spectroscopy, (3) Lineshape analysis, (4) CPMG relaxation dispersion, (5) Rotating frame relaxation dispersion, (6) Nuclear spin relaxation, (7) Residual dipolar coupling, (8) Paramagnetic relaxation enhancement. This article is part of a Special Issue entitled: Protein Dynamics: Experimental and Computational Approaches.