molearn is a generative neural network trainable with protein conformations. This repository contains the following Jupyter notebooks.
molearn analysis.ipynb: a tutorial showing in detail how to interact with a trained neural network, and how to analyse its performance. This notebook shows how analysis is carried out inmolearn.analysis.MolearnAnalysis.molearn_GUI.ipynb: a demonstration of a graphical user interface used to display a neural network latent space, and exploit it to generate protein conformations.minimal_example.ipynb: the minimal lines of code required to load a trained neural network and the data it was trained with, gather analysis data, and display them graphically.
If you use molearn in your work, please cite: V.K. Ramaswamy, S.C. Musson, C.G. Willcocks, M.T. Degiacomi (2021). Learning protein conformational space with convolutions and latent interpolations, Physical Review X 11