CASTELO: clustered atom subtypes aided lead optimization-a combined machine learning and molecular modeling method.

BACKGROUND: Drug discovery is a multi-stage process that comprises two costly major steps: pre-clinical research and clinical trials. Among its stages, lead optimization easily consumes more than half of the pre-clinical budget. We propose a combined machine learning and molecular modeling approach that partially automates lead optimization workflow in silico, providing suggestions for modification hot spots. RESULTS: The initial data collection is achieved with physics-based molecular dynamics simulation. Contact matrices are calculated as the preliminary features extracted from the simulations. To take advantage of the temporal information from the simulations, we enhanced contact matrices data with temporal dynamism representation, which are then modeled with unsupervised convolutional variational autoencoder (CVAE). Finally, conventional and CVAE-based clustering methods are compared with metrics to rank the submolecular structures and propose potential candidates for lead optimization. CONCLUSION: With no need for extensive structure-activity data, our method provides new hints for drug modification hotspots which can be used to improve drug potency and reduce the lead optimization time. It can potentially become a valuable tool for medicinal chemists.

Artificial Intelligence for Accelerating Time Integrations in Multiscale Modeling.

We developed a novel data-driven Artificial Intelligence-enhanced Adaptive Time Stepping algorithm (AI-ATS) that can adapt timestep sizes to underlying biophysical dynamics. We demonstrated its values in solving a complex biophysical problem, at multiple spatiotemporal scales, that describes platelet dynamics in shear blood flow. In order to achieve a significant speedup of this computationally demanding problem, we integrated a framework of novel AI algorithms into the solution of the platelet dynamics equations. Our framework involves recurrent neural network-based autoencoders by the Long Short-Term Memory and the Gated Recurrent Units as the first step for memorizing the dynamic states in long-term dependencies for the input time series, followed by two fully-connected neural networks to optimize timestep sizes and step jumps. The computational efficiency of our AI-ATS is underscored by assessing the accuracy and speed of a multiscale simulation of the platelet with the standard time stepping algorithm (STS). By adapting the timestep size, our AI-ATS guides the omission of multiple redundant time steps without sacrificing significant accuracy of the dynamics. Compared to the STS, our AI-ATS achieved a reduction of 40% unnecessary calculations while bounding the errors of mechanical and thermodynamic properties to 3%.

Online Machine Learning for Accelerating Molecular Dynamics Modeling of Cells.

We developed a biomechanics-informed online learning framework to learn the dynamics with ground truth generated with multiscale modeling simulation. It was built on Summit-like supercomputers, which were also used to benchmark and validate our framework on one physiologically significant modeling of deformable biological cells. We generalized the century-old equation of Jeffery orbits to a new equation of motion with additional parameters to account for the flow conditions and the cell deformability. Using simulation data at particle-based resolutions for flowing cells and the learned parameters from our framework, we validated the new equation by the motions, mostly rotations, of a human platelet in shear blood flow at various shear stresses and platelet deformability. Our online framework, which surrogates redundant computations in the conventional multiscale modeling by solutions of our learned equation, accelerates the conventional modeling by three orders of magnitude without visible loss of accuracy.

Dynamics-Based Peptide-MHC Binding Optimization by a Convolutional Variational Autoencoder: A Use-Case Model for CASTELO.

An unsolved challenge in the development of antigen-specific immunotherapies is determining the optimal antigens to target. Comprehension of antigen-major histocompatibility complex (MHC) binding is paramount toward achieving this goal. Here, we apply CASTELO, a combined machine learning-molecular dynamics (ML-MD) approach, to identify per-residue antigen binding contributions and then design novel antigens of increased MHC-II binding affinity for a type 1 diabetes-implicated system. We build upon a small-molecule lead optimization algorithm by training a convolutional variational autoencoder (CVAE) on MD trajectories of 48 different systems across four antigens and four HLA serotypes. We develop several new machine learning metrics including a structure-based anchor residue classification model as well as cluster comparison scores. ML-MD predictions agree well with experimental binding results and free energy perturbation-predicted binding affinities. Moreover, ML-MD metrics are independent of traditional MD stability metrics such as contact area and root-mean-square fluctuations (RMSF), which do not reflect binding affinity data. Our work supports the role of structure-based deep learning techniques in antigen-specific immunotherapy design.