RESUMO
Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper first an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), is developed that learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, interpolating data points in the learned latent space are selected that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-ß1-42 (Aß42) monomer, our deep learning model provided a comprehensive sampling of Aß42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. The proposed approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.
RESUMO
Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper we first introduced an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), which learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, we selected interpolating data points in the learned latent space that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-ß 1-42 (Aß42) monomer, our deep learning model provided a comprehensive sampling of Aß42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. This approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.
RESUMO
Understanding ligand binding kinetics and thermodynamics, which involves investigating the free, transient, and final complex conformations, is important in fundamental studies and applications for chemical and biomedical systems. Examining the important but transient ligand-protein-bound conformations, in addition to experimentally determined structures, also provides a more accurate estimation for drug efficacy and selectivity. Moreover, obtaining the entire picture of the free energy landscape during ligand binding/unbinding processes is critical in understanding binding mechanisms. Here, we present a Binding Kinetics Toolkit (BKiT) that includes several utilities to analyze trajectories and compute a free energy and kinetics profile. BKiT uses principal component space to generate approximated unbinding or conformational transition coordinates for accurately describing and easily visualizing the molecular motions. We implemented a new partitioning approach to assign indexes along the approximated coordinates that can be used as milestones or microstates. The program can generate input files to run many short classical molecular dynamics simulations and uses milestoning theory to construct the free energy profile and estimate binding residence time. We first validated the method with a host-guest system, aspirin unbinding from ß-cyclodextrin, and then applied the protocol to pyrazolourea compounds and cyclin-dependent kinase 8 and cyclin C complexes, a kinase system of pharmacological interest. Overall, our approaches yielded good agreement with published results and suggest ligand design strategies. The computed unbinding free energy landscape also provides a more complete picture of ligand-receptor binding barriers and stable local minima for deepening our understanding of molecular recognition. BKiT is easy to use and has extensible features for future expansion of utilities for postanalysis and calculations.