Title: Building QSAR/QSPR models
QSAR/QSPR is a modeling approach for estimating a chemical's properties or biological activity based on a broad range of real and theoretical descriptors. The premise is that chemicals with similar structure or molecular properties tend to interact with a biological system in a similar manner, therefore it should be feasible to build a predictive model for toxicity or other properties of interest utilizing inductive reasoning or concept learning.
Many challenges and decisions arise when building QSAR/QSPR models, including the SAR paradox, choice of validation methods, bias in data and feature selection, the use of hybrid feature vectors, the multitude of machine learning strategies available, and the all too frequent scarce data scenarios. Our interest and work is in accelerating data-driven research through effective QSxR practices and the development of integrated, feature rich, software tools that support rational chemical design.
J. Gormley — CTO, IOMICS Corporation
Gormley is an industry consultant with 20 years experience developing software-based research tools. He obtained a B.S. in Computer Science from the University of Maryland and an ALM in (Molecular) Biology from Harvard University.
D. Corkill — Senior Research Scientist, UMass Amherst and CSO, IOMICS Corporation
Corkill has a long history of AI research in blackboard and multi-agent systems, collaborating software entities, planning and control, information fusion, and knowledge discovery. He received his Ph.D. in Computer Science from the University of Massachusetts Amherst in 1983.