This project deals with the development and use of advanced computer simulation methods having primary application area in biomolecular and materials modeling. The molecular simulations methodologies, based on the fundamental physical principles and theory, are essential for computer modeling to be predictive, moving us to the state when computer modeling and simulations can be used as a guide in the development of new compounds and materials with desirable properties and safe for the environment. Specific aims of the project are the following: Specific aims of the project are the following: 1) Development of methodology of the systematic multiscale modeling based on machine learning and artificial neural networks potentials 2) Use of the developed methodology to build coarse-grained lipid models fully transferable over wide concentration range for many-component mixtures. 3) Modeling of peptides adsorption on TiO2 4) Modeling of solubility and crystallization of polymorphic drugs in different solvents. Addressing to these problems requires extensive use of various computational chemistry software, particularly classical molecular dynamics, mesoscale (coarse-grained) dynamics, Monte-Carlo simulations, neural networks software. All these computations need access to high performance computing resources in order to reach the scientific goals.