Structure Prediction and Mutagenesis with Alphafold for Drug Target Discovery

Introduction: Novel drug target discovery is one of the first steps of the drug development pipeline. Protein NMR and X-ray crystallography have yielded a golden era of target identification by allowing scientists to characterize the three-dimensional structure of potential targets. Structure determination using empiric methods carries a non-negligible cost that often prohibits the characterization of novel drug targets that have yet to be extensively studied. AlphaFold is a machine learning algorithm which has been shown to produce protein structure predictions with unprecedented accuracy from amino acid sequence alone. Here, we seek to characterize the feasibility of NTNG1 as a drug target in invasive glioma.

Methods: We employ structure prediction with AlphaFold as well as molecular dynamics simulation with NAMD 2.14 to characterize a possible protein-protein interaction between NTNG1 and TSP-1. AlphaFold predictions were obtained via a non-templated approach using the Google Colaboratory notebook. Molecular dynamics simulations were run on our in-house JumperOne Server. We employ 100 nanosecond simulations of both WT and mutant TSP-1 binding NTNG1 in a water box at physiologic ion concentration.

Results: AlphaFold predictions show high homology between different models, suggesting NTNG1 and TSP-1 interaction is likely. Molecular dynamics simulations reveal an aspartate cluster in NTNG1 that has high affinity for TSP-1. Molecular dynamics simulation of mutant TSP-1 reveals a much more mobile complex (RMSD ~15Å at 100 ns) than WT TSP-1 in complex with NTNG1 (RMSD ~7Å at 100 ns).

Conclusion : Here, we describe a potential protein-protein interaction between NTNG1 and TSP-1 which we characterize via machine learning protein structure prediction and molecular dynamics simulation. Our study serves as a proof of concept for the use of AlphaFold to determine mutations of particular importance in electrostatic interactions between protein domains. Future work aims to determine localization of NTNG1 and TSP-1 in highly infiltrative glioma.