Towards Better Receptor-Ligand Prioritization: How Machine Learning on Protein-Protein Interaction Data Can Provide Insight Into Receptor-Ligand Pairs

The prediction of receptor-ligand pairs is an active area of biomedical and computational research. Oddly, the application of machine learning techniques to this problem is a relatively under-exploited approach. Here we seek to understand how the application of least squares support vector machines (LS-SVM) to this problem can improve receptor-ligand predictions. Over the past decade, the amount of protein-protein interaction (PPI) data available has exploded into a plethora of various databases derived from various wet-lab techniques. Here we use PPI data to predict receptor ligand pairings using LSSVM. Our results suggest that this approach provides a meaningful prioritization of the receptor-ligand pairs.