Predicting Protein Polypharmacy with Graph Machine Learning

Motivation

The journey of drug discovery and making it accessible to patients can span a decade, during which it is marked by numerous instances of unsuccessful outcomes. Digital advancements have revolutionized the process of developing drugs, resulting in the generation of vast amounts of data. A crucial aspect of the biomedical data utilized in drug discovery is its interconnected nature. This type of data structure can be represented as a graph, a concept widely employed in various fields of biology to model the intricate interactions between biological entities at different levels. On a molecular scale, proteins and other biomolecules can be depicted as graphs that capture the spatial and structural relationships among their molecules. By representing molecules as graphs and leveraging machine learning techniques, researchers can gain insights into the structure, function, and interactions of molecules. Graph machine learning enables the prediction of molecular properties, the design of novel drug candidates, and the analysis of drug-target interactions. It accelerates the drug discovery process by reducing the reliance on costly and time-consuming experimental screenings, leading to more efficient and effective drug development. The poster will highlight the underlying theory, limitations and outlook in this area.