Title: Computational Analysis and Integration of Large-Scale Biological Data with Deep Learning Approaches
Presenter: Dr. Tunca Doğan from EMBI – EBI & METU
Date: August 2, 2018
Abstract: Machine learning and data mining techniques are frequently employed to make sense of large-scale and noisy biological/biomedical data accumulated in public servers. A key subject in this endeavour is the prediction of the properties of proteins such as their functions and interactions. Recently, deep learning (DL) based methods have outperformed the conventional machine learning algorithms in the fields of computer vision, natural language processing and artificial intelligence; which brought attention to their application to the biological data. In this talk, I’m going to explain the DL-based probabilistic computational methods we have recently developed in our research center (KanSiL, Graduate School of Informatics, ODTU); first, to predict the functions of the uncharacterised proteins (i.e., DEEPred); and second, to identify novel interacting drug candidate molecules for all potential targets in the human proteome (i.e., DEEPscreen) to serve the purposes of drug discovery and repositioning, together with the aim of biomedical data integration. Apart from the benefits of employing novel DL approaches, I’ll also mention the limitations of DL-based techniques when applied on the biological data, to explain why deep learning alone cannot solve every problem related to bioinformatics.