This is a guest post from Leon Eyrich Jessen, a postdoctoral researcher in the Immunoinformatics and Machine Learning Group at the Technical University of Denmark.
The aim of this post is to illustrates how deep learning is successfully being applied to model key molecular interactions in the human immune system. Molecular interactions are highly context dependent and therefore non-linear. Deep learning is a powerful tool to capture non-linearity and has therefore proven invaluable and highly successful. In particular in modelling the molecular interaction between the Major Histocompability Complex type I (MHCI) and peptides (The state-of-the-art model netMHCpan identifies 96.5% of natural peptides at a very high specificity of 98.5%).