This repository contains various scripts to tackle some of the structural biology problems I ran into while completing my doctorate. This is slowly being updated as I go through my work and beautify everything.
- DEERNet: A Colab notebook that reproduces most of the training process and functionality of the DEERNet neural network. The reproduction is carried out in Jax.
- Fit activity data: A Colab notebook that fits some activity data using a Michaelis-Menten model in Python using a standard nonlinear curve fit.
- MDDS: A Colab notebook and set of accompanying scripts for reproduction of the Molecular Dynamics of Dummy Spin Labels program available on the CHARMM-GUI web server. Instead of using molecular dynamics force fields, which are costly and can take up to 24 hours for the full 5 nanosecond simulation, this implementation using the No U-Turn Sampler, a Hamiltonian Monte Carlo sampler that samples the available space more effectively than classical random-walk Monte Carlo.
- CM: A series of scripts that does comparative modeling of a protein into a new conformation. This is intended to be used when the protein of interest is only known in a single conformation, and a different conformation adopted by a homolog is desired. The alignment is carried out using TM-Align and the sequence of interest is threaded using this alignment onto the target structure using the Rosetta application partial_thread. Then, RosettaCM (more specifically HybridizeMover, implemented in RosettaScripts) is applied to optimize backbone geometry, close loops, and add side chains. This script uses multiprocessing, but for best results, you should run on a computing cluster and generate hundreds to thousands of models.
- TM-PCA: Scripts to align and interpret a set of models. Alignment is carried out using TM-Align, while principal component analysis is carried out using Scikit-Learn. These scripts make extensive use of BioPython.
- Report residue movement: Script to align two conformations of the same protein and report on the movement of each alpha carbon. Alignment is carried out using TM-Align. This script uses BioPython.
- AAT_OmegaFold: A series of protein structural models generated with OmegaFold as part of a public demonstration (see this Twitter thread).