Cheminformatics and Machine learning

Bioinformatics and cheminformatics are scientific disciplines that have evolved in recent decades at the interface between chemistry, biology and computer science. In many areas of chemistry, biology and materials science, the huge amount of data and information produced by robotic high-throughput assays and micro-array technologies can only be processed and analyzed using computational techniques. Predictive informatics methods employ statistical techniques to mine this data for hidden correlations and to retrieve molecules, genes or patterns with specific properties or desirable biological activities from large databases. Furthermore, many of these problems are so complex that novel approaches utilizing solutions based on informatics methods are necessary. Informatics methods have been especially valuable in drug design and development, but are also increasingly employed in several other disciplines which are governed by the science of complex systems. This calls for an inter-disciplinary approach to the study of these problems.

Currently active research programs include:

• Drug and polymer design through quantitative structure-activity/property relationship modeling (QSAR/QSPR), statistical learning algorithms and target-based chemogenomics;
• Protein and DNA bioinformatics using a combination of structure-based methods and fundamental molecular-level descriptors;
• Applications of Deep Learning in Cheminformatics;
• Chemical and biological networks: It is only recently, with the advent of public repositories of information and availability of high-throughput assays and computational resources, that analysis of the properties of large chemical and biological networks, such as protein-protein and protein-drug interaction networks, has become possible. A holistic view of drug design, incorporating multiple sources of information: genomic, proteomic, metabolomic and transcriptomic, is still very much in its infancy, but is the need of the hour.