[17a-F206-1~12] 12.6 Nanobiotechnology

[17a-F206-8] Automated analyses and knowledge derivations of fragment molecular orbital calculations

〇Yuji Mochizuki^1,2, Yuto Komeiji³, Mayu Fujimoto¹, Sona Saitou¹, Jun Iijima¹, Hideo Doi¹, Tsuyoshi Iyama¹, Akira Okusawa⁴, Takeshi Makimura⁴, Takaya Nakanishi⁴, Kaori Fukuzawa⁵, Shigenori Tanaka⁶ (1.Rikkyo Univ., 2.Univ. Tokyo, 3.AIST, 4.Knowledge Communication Co Ltd, 5.Hoshi Univ., 6.Kobe Univ.)

Keywords:Fragment Molecular Orbital (FMO) calculations, Machine Learning

The inter-fragment interaction energies (IFIEs) are of the most useful quantity to grasp the nature of interactions among amino acid residues in the target protein calculated by the fragment molecular orbital (FMO) method. When the number of sample structure grows, the manual understanding of IFIE should become hard. Thus, we have been developing the machine learning-assisted automatic characterization with the MS Azure environment. Furthermore, Google’s TensorFlow as a representative deep-learning tool has been used to analyze the IFIE-map which is a two-dimensional visualization of IFIE values. In this presentation, we will present a summary of these researches and address the future directions.