[14p-P18-1~19] 12.6 Nanobiotechnology

[14p-P18-7] Machine-learning assisted interaction analyses of amino acid residues in Ubiquitin

〇Yuji Mochizuki^1,2, Mayu Fujimoto¹, Yuto Komeiji³, Iijima Jun¹, Sona Saitou¹, Hideo Doi¹, Akira Okusawa⁴, Takeshi Makimura⁴, Takaya Nakanishi⁴, Kaori Fukuzawa^2,5, Shigenori Tanaka⁶ (1.Rikkyo Univ., 2.Univ. Tokyo, 3.AIST, 4.Knowledge Communication Co. Ltd., 5.Nihon Univ., 6.Kobe Univ.)

Keywords:Machine Learning, Fragment Molecular Orbital Method, Interaction Analyses

In our poster presentation at the 2016 Spring JAP meeting, we introduced a machine learning-assisted automatic analysis (supplied by MS’s Azure environment) of residue-residue interaction energies obtained by the fragment molecular orbital (FMO) calculations as a set of inter fragment interaction energies (IFIEs), where Chignolin consisting of 10 residues was employed as the first example. As a successive study, Ubiquitin with 76 residues has been analyzed with a pre-conditioning of clustering for the IFIE entries as many as 2775. The visualized IFIEs like maps have been subjected to the deep learning with Google’s TensorFlow as well. These results will be presented as a poster.