Step by step one goes very far
Project context and summary :
We have developed a computer tool, named InDeep, that relies on 3D fully convolutional neural networks to predict functional binding sites at the surface of proteins. These functional binding sites can take two forms, either a epitope binding site (location of a protein-protein interaction), or a druggable binding site (location for the binding of a future drug). Presently, the tool is already used on campus in several structural biology and drug discovery projects to support the identification of chemical probes with therapeutic purposes. This includes SARS-CoV2 projects in collaboration with Fabrice Agou, Félix Rey and Marc Delarue. In its present form, InDeep relies on GPU calculations and has to be used within a Linux Shell in command line and although a pymol plugin has been also developed, it requires some prerequired installations. This impedes its usage by the largest audience especially by the community for whom it was designed, namely biologists and chemists. The purpose of this project would be to design a web interface assisted by a GPU cluster on campus to allow the use of InDeep even for non-computer specialists. This would also be the opportunity to add intuitive functionalities (3D structure visualization, metrics, cross-references to well established databases) to assist the user in his/her attempt to identify pertinent functional binding sites.Related team publications :
Sorry. You must be logged in to view this form.