SI2-SSE: High-Performance Software for Large-Scale Modeling of Binding Equilibria

点击翻译按钮获取中文摘要

Living organisms are regulated by specific interactions among proteins and other macromolecules. Cancer and genetic diseases are due to altered interactions caused by mutations. Cellular organelles and viruses spontaneously build themselves from the ordered assembly of component molecules. While a complete understanding remains elusive, these and many other phenomena fundamentally rest on the ability of molecules to recognize and bind specifically to other molecules. The general principles of molecular recognition are being employed to design new drugs, chemical catalysts, and advanced materials. Computer models offer an important mean to disentangle and analyze molecular interactions and to produce quantitative predictions. The main objective of the project is to develop novel algorithms to model molecular recognition processes at atomic resolution on modern parallel computer architectures. This research seeks to increase the speed of the calculations and expand hardware support to enable the screening of larger sets of drug candidates and the study of multiple protein mutations under various conditions. The increased accuracy and availability of modeling software technologies by a larger community will lead to new ideas and research approaches, and ultimately to new discoveries in medicine, chemistry and material science. This effort will contribute to the establishment of in silico means to evaluate environmental and clinical claims. For example computational evidences on toxicity of substances can inform public policy in the same way that, for instance, atmospheric models are currently used for global climate projections.The project targets the Binding Energy Distribution Analysis Method (BEDAM, for short), an accurate model of molecular binding, currently limited by computational performance. Outcomes of this research include deployment the BEDAM model for the first time on General Purpose Graphical Processing Units (GPGPU) and Many Integrated Core (MIC) massively parallel architectures. To this end, the mathematical formulation of the model will be tuned to best utilize the features of these modern computing architectures. Specialized recursive computational geometry algorithms will be developed to extract from the parallel hardware near optimal performance. Robust automated tools for the processing of molecular models and their analysis will be put in place for large scale applications. Accessible user interfaces will be implemented to ensure wide applicability and adoption of the software. These software applications will be distributed under an open source license to promote sharing and community contributions. Student research assistants form an integral part of the research team. While contributing meaningfully to scientific research, students from challenging socioeconomic backgrounds will acquire computer programming and software maintenance skills useful to enter the high tech job market.