Albany 2015:Book of Abstracts
June 9-13 2015
©Adenine Press (2012)
Computational Analysis and Prediction of Proteins that undergo domain swapping
Available three-dimensional structures of proteins suggest that proteins can engage in limited extent of aggregation through a peculiar form of 'hand-shake' strategy that is defined as 'domain swapping'. Few proteins also form unlimited aggregation through domain swapping leading to diseases and hence this phenomenon acquires biomedical importance. Starting from a few hundred protein structures where domain swapping is evident, analytical tools were developed to automatically recognise the hinge regions and sequence and structural analyses were performed to study amino acid preferences at 'hinge' regions. Such subtle amino acid preferences were next encoded into a machine learning algorithm, driven by Random Forest for its training, to predict proteins that are likely to be engaged in domain swapping from mere sequence information. Surprisingly, a significant fraction of gene products in the entire human genome show predictive tendencies to be in domain-swapped form. Further, meta-analysis of such domain-swapped proteins show their involvement in neurodegenerative diseases and cancer.
National Centre for Biological Sciences (TIFR)