"TTDock: a docking method based on tensor train decompositions"
Zheltkov D.A., Oferkin I.V., Katkova E.V., Sulimov A.V., Sulimov V.B., Tyrtyshnikov E.E.

A new docking method based on tensor train decomposition is proposed. This method allows one to find the position of the energy global minimum for the ligand-protein system with a high probability. The proposed method is compared with one of the best genetic algorithm docking program SOL. According to the testing results, the docking method based on tensor train decompositions is up to 10 times faster, whereas the energy global optimum is reached with the same probability.

Keywords: tensor train decomposition, cross interpolation method, global optimization, docking, computer drug design

Zheltkov D.A., e-mail: dmitry.zheltkov@gmail.com – Moscow State University, Faculty of Computational Mathematics and Cybernetics; Leninskiye Gory 1-52, Moscow, 119991, Russia;
Oferkin I.V., e-mail: io@dimonta.com;   Katkova E.V., e-mail: katkova@dimonta.com;   Sulimov A.V., e-mail: sulimovv@mail.ru – Dimonta Company; ulitsa Nagornaya 15, Moscow, 117186, Russia;
Sulimov V.B., e-mail: vladimir.sulimov@gmail.com – Research Computing Center, Lomonosov Moscow State University; Leninskiye Gory, Moscow, 119991, Russia;
Tyrtyshnikov E.E., e-mail: eugene.tyrtyshnikov@gmail.com – Institute of Numerical Mathematics, Russian Academy of Sciences; ulitsa Gubkina 8, Moscow, 119333, Russia