Context and problematic

Finding new patentable drugs is increasingly difficult and carrying out such research manually is starting to show its limits. The use of AI suggest lines of research, and this could greatly speed-up the creation of new drugs.

Goals

Based on the many medical publications and clinical trials, to detect correlations between drugs in order to create new combinations of more effective molecules.
Moreover, take advantage of this project to provide the company with an efficient, centralized publication search API, and not a limited one, in terms of number of requests.

Our intervention

2 Data Scientist, 1 Data Engineer​

• Data extraction, cleaning, then indexing and “using” this data with the Azure indexer
• Creation of a graph database through relationships between molecules and publications (available with the Azure indexer)
• Creation of a visual application that displays sub-parts of this graph database

Results

APIs available to efficiently access all medical publications and clinical trials !
A data visualization application which helps technical teams look for interesting publications and molecules, but also view the links between them.

Technical environment

Azure – Api Rest – CosmosDb – Gremlin – Python – Networkx – Plotly – Dash