PROSTATA: a framework for protein stability assessment using transformers.

MOTIVATION: Accurate prediction of change in protein stability due to point mutations is an attractive goal that remains unachieved. Despite the high interest in this area, little consideration has been given to the transformer architecture, which is dominant in many fields of machine learning. RESULTS: In this work, we introduce PROSTATA, a predictive model built in a knowledge-transfer fashion on a new curated dataset. PROSTATA demonstrates advantage over existing solutions based on neural networks. We show that the large improvement margin is due to both the architecture of the model and the quality of the new training dataset. This work opens up opportunities to develop new lightweight and accurate models for protein stability assessment. AVAILABILITY AND IMPLEMENTATION: PROSTATA is available at https://github.com/AIRI-Institute/PROSTATA and https://prostata.airi.net.

nablaDFT: Large-Scale Conformational Energy and Hamiltonian Prediction benchmark and dataset.

Electronic wave function calculation is a fundamental task of computational quantum chemistry. Knowledge of the wave function parameters allows one to compute physical and chemical properties of molecules and materials. Unfortunately, it is infeasible to compute the wave functions analytically even for simple molecules. Classical quantum chemistry approaches such as the Hartree-Fock method or density functional theory (DFT) allow to compute an approximation of the wave function but are very computationally expensive. One way to lower the computational complexity is to use machine learning models that can provide sufficiently good approximations at a much lower computational cost. In this work we: (1) introduce a new curated large-scale dataset of electron structures of drug-like molecules, (2) establish a novel benchmark for the estimation of molecular properties in the multi-molecule setting, and (3) evaluate a wide range of methods with this benchmark. We show that the accuracy of recently developed machine learning models deteriorates significantly when switching from the single-molecule to the multi-molecule setting. We also show that these models lack generalization over different chemistry classes. In addition, we provide experimental evidence that larger datasets lead to better ML models in the field of quantum chemistry.

Charge Transfer Complexes of 1,3,6-Trinitro-9,10-phenanthrenequinone with Polycyclic Aromatic Compounds.

Understanding the interactions of organic donor and acceptor molecules in binary associates is crucial for design and control of their functions. Herein, we carried out a theoretical study on the properties of charge transfer complexes of 1,3,6-trinitro-9,10-phenanthrenequinone (PQ) with 23 aromatic π-electron donors. Density functional theory (DFT) was employed to obtain geometries, frontier orbital energy levels and amounts of charge transfer in the ground and first excited states. For the most effective donors, namely, dibenzotetrathiafulvalene, pentacene, tetrathiafulvalene, 5,10-dimethylphenazine, and tetramethyl-p-phenylenediamine, the amount of charge transfer in the ground state was shown to be 0.134-0.240 e-. Further, a novel charge transfer complex of PQ with anthracene was isolated in crystalline form and its molecular and crystal structure elucidated by single-crystal synchrotron X-ray diffraction.

SEMA: Antigen B-cell conformational epitope prediction using deep transfer learning.

One of the primary tasks in vaccine design and development of immunotherapeutic drugs is to predict conformational B-cell epitopes corresponding to primary antibody binding sites within the antigen tertiary structure. To date, multiple approaches have been developed to address this issue. However, for a wide range of antigens their accuracy is limited. In this paper, we applied the transfer learning approach using pretrained deep learning models to develop a model that predicts conformational B-cell epitopes based on the primary antigen sequence and tertiary structure. A pretrained protein language model, ESM-1v, and an inverse folding model, ESM-IF1, were fine-tuned to quantitatively predict antibody-antigen interaction features and distinguish between epitope and non-epitope residues. The resulting model called SEMA demonstrated the best performance on an independent test set with ROC AUC of 0.76 compared to peer-reviewed tools. We show that SEMA can quantitatively rank the immunodominant regions within the SARS-CoV-2 RBD domain. SEMA is available at https://github.com/AIRI-Institute/SEMAi and the web-interface http://sema.airi.net.