Workshop on biomolecular simulations

Mercedes Alfonso-Prieto, Assistant Professor, Forschungszentrum Jülich, Germany

Proton transfer (PT) is one the most common processes in biology. Studying PT at the molecular level is a challenging task, due to the intrinsic complexity of proteins, including their functional dynamics and the presence of several titratable residues. Therefore, combination of experimental and computational techniques is needed to obtain a more comprehensive picture. Recent methodological and software advancements have enabled the efficient application of quantum mechanics/molecular mechanics (QM/MM) simulations to (large) biological systems. This multiscale approach allows treating the region of the protein where PT takes place at the QM level, while including the effect of the rest of the protein and its environment at the MM level. In this talk, I will show some examples of QM/MM simulations applied to PT processes in different systems, from enzymes to membrane proteins. Such hybrid simulations have provided atomistic insights complementary to experiments.

Miłosz Wieczór, Postdoctoral fellow, IRB Barcelona

We present a new paradigm for the parametrization of nonbonded interactions in molecular force fields directly from quantum chemical data. The proposed machine-learned force field (ML-FF) accurately reproduces well-validated quantum chemical interaction energies, as well as allows for the systematic incorporation of experimental data such as measured affinity constants. Contrary to many recently published general ML-FFs, our force field ensures stability in molecular dynamics simulations thanks to a combination of regularization and active learning, and permits gradual upgrades should issues arise in previously untested systems. Meanwhile, a simple embedding scheme allows the parameters learned from minimal methyl-nucleobase systems to be employed in simulations of large biomacromolecules at virtually no computational overhead. We also present several key directions for the future evolution of our force field, both in terms of accuracy and applicability.

Simon Olsson, Assistant Professor, Chalmers University of Technology

Long-time-scale and enhanced sampling molecular dynamics simulations reveal the approximations of classical molecular dynamics force fields disagreeing with experimental data. However, under ideal conditions, experiments and simulations observe the same process — consequently, by optimally balancing experimental and simulation data, we may build models extracting helpful information from both. In this talk, I will overview efforts in this direction and talk about Augmented Markov models, and their recent extension dynamic Augmented Markov models.