ExaBio

Towards exascale computational bio for genome to function. An HPC workflow pipleline to infer protein structure from gene sequence, then use structural and bioinformatics tooks to infer function.

Data

This section discusses the data for this project.

Compute

We intend to use various ORNL compute resources such as Summit, CADES, DGX, Raptor etc.

Software

Here are the links/refs to the software that we intend to use as part of this project.

OpenMM: https://github.com/openmm/openmm

AMBER version of ExaFold (CPU only), this one is completely functional: https://github.com/BSDExabio/openFoldTools
^ repo also includes script (and instructions) for generating customizable basic restraints from exsisting PDB files with BioPython.

ExaFold (in progress, uses OpenMM instead of AMBER): https://github.com/jkwoods/ExaFold

DeepMetaPsiCov: Uses other codes to get info to use a DNN (Deep Neural Network) to predict contacts for the protein folding step Uses Pytorch-- let's see if inference will just run out of the box on GPUs! Uses CCMPred and other codes that are referenced in the doc https://github.com/psipred/DeepMetaPSICOV

Protein-protein rigid docking (we are in touch with the dev and want to offload this to GPU): https://github.com/brianjimenez/lightdock

The PAW workflow manager: https://github.com/pawtools/nopaw

https://github.com/pawtools/db-launcher

DNSS secondary structure prediction: https://github.com/multicom-toolbox/DNSS

Workflows

This section describes end-to-end workflows that will be enacted as part of this project.

End-to-end stages, workflow 1

1. MSA
2. SSpred
3. CCMpred
4. Build Restrained System
5. ExaFold
6. Run MD to test stability
7. PPI interactions: rigid docking, MD, co-evo hot spots

Meta: coupled learning/ analysis