Implementation of COMP 561 final project at McGill University, taught by Prof. Mathieu Blanchette.
We investigate graph convolutional networks (GCN) for error correction under large time step molecular dynamics.
Below is the simulated annealing of the carbohydrate-recognition domain of the human langerin protein (PDB 3P5G) with implicit solvation. We conducted this experiment under the canonical ensemble by decreasing temperature from 500K to 300K, over the course of 5 nanoseconds. Expectedly, the structure becomes increasingly compact as temperature cools. Data collected from this simulation was then used to develop the GCN.