Development of machine-learning based prediction methods for inhibitors of HTRA1

Abstract

A wide array of activities in bioinformatics involves prediction of patterns and classification in biological data. Biological databanks, with their behemoth size, necessitate computer intervention and automation in this classification process. Currently, support vector machines (SVMs) are the computer programs with best prediction performance. SVMs optimise the margin separating two classes for better generalisation on unseen data. HTRA1, a 50 kDa secreted protein, a member of a family of serine proteases called “High Temperature Requirement A”. The family includes other members namely: HTRA2, HTRA3, and HTRA4. All these proteins show a nonspecific protease activity while the exact role of these HTRAs is yet unknown. HTRA1 comprises a signalling peptide, a Kazal-like protease inhibitor domain, an IGF (Insulin like Growth Factor) binding domain, a PDZ domain, and a conserved serine protease domain. The protein has shown a role in osteoarthritis, Alzheimer's disease and age-related macular degeneration, to suggest a few studies. Changes in expression of the HTRA1 gene or changes in activity of the enzyme are usually responsible for such conditions. The protein has also shown a role in chemotherapy-induced cytotoxicity in gastric, ovarian and other similar cancers. These studies are suggestive of HTRA1 as a novel therapeutic target for multiple diseases and conditions. A specific inhibitor for this serine protease would be of paramount importance in further studies to elucidate the normal function of HTRA1 & its deregulation in the development and progression of human disease. It could potentially lead to the development of novel and effective clinical interventions.