Molecular Informatics Team (https://aip.riken.jp/labs/goalorient_tech/mol_inf/) at RIKEN AIP
Speaker 1: Koji Tsuda (10min)
Title: Overview of Molecular Informatics Team
In molecular informatics team, we develop computational methods for analyzing and developing molecules and materials. We focus on machine learning driven design of new materials and proteins, and functional analysis of protein dynamics using the rigidity theory. I introduce our activities briefly in this talk.
Speaker 2: Masato Sumita (45min)
Title: Informatics driven research for advanced materials based on physicochemistry
The development of nano-technology widely realized us the effectiveness of controlling materials at the atomistic level. Many studies at the atomistic level are accumulated in physicochemical theories. However, the atomistic-level control of materials using these theories is not easy because of complexity of the atomistic arrangement of materials. Machine learning is useful to deal with this complexity. In this seminar, we will show some successful examples where machine learning compensates physicochemical theories to search or design materials.
Speaker 3: Adnan Sljoka (45min)
Title: Rigidity theory of frameworks and graphs with applications to protein structure validations and functional analysis
Mathematical Rigidity theory stands at the nexus of combinatorics, geometry, graph theory and algorithms. Rigidity theory is concerned with the rigidity and flexibility of structures that are defined by geometric constraints (fixed lengths, directions etc.) on a set of points, line segments, polygons, bodies, atoms etc. This theory has rich historical roots which date back to Euler (1766) and to Maxwell’s studies of mechanical linkages in the 19th century. Over the last few decades, interest and mathematical developments in rigidity theory have blossomed rapidly, which is motivated by applications in many areas of science, engineering and design, where geometric constraints serve as suitable mathematical models for an assortment of man-made structures (e.g. robots, mechanisms, sensor networks, meta-materials, and Computer-Aided-Design software) or natural materials (e.g. biomolecules, proteins, and crystals). In this talk I will introduce basic concepts in rigidity theory setting up a combinatorial characterization of rigidity of generic frameworks and graphs. Macromolecules such as proteins are well suited for analysis using rigidity theory, whose functions is delicately controlled by interplay of rigidity and flexibility. I will briefly highlight some of my advances in this area, such as development of a first workable method for validation of NMR protein structures (Nature Communications 2020) which has larger implications in protein structure prediction, fast computational predictions of protein flexibility and dynamics that have provided critical clues and breakthroughs in biological mysterious like enzyme catalysis (Science 2017, JACS 2019), cell signalling (Nature Communications 2018, Cell 2021), antibody-antigen recognition (Frontiers in Immunology 2018), Covid-19 protein motions etc. We also briefly discuss, time permitting, some recent work where we seek to combine rigidity theory techniques with search algorithms / reinforcement learning to better understand complicated disordered protein dynamics events that are linked to degenerative diseases.