AlphaFold and friends on the HPC

AlphaFold
Introduction to AlphaFold
2 Topics
DeepMind
Critical Assessment of protein Structure Prediction (CASP)
The AlphaFold pipeline
2 Topics
Overview of the architecture
AlphaFold training details
Online accessibility
3 Topics
Publications, GitHub code and database
Official AlphaFold colab
ColabFold
AlphaFold on the HPC
10 Topics
Requirements to use the HPC
Basic UNIX commands
Prepare directories and files
Setting up the job script
AlphaFold outputs
Computational limits
Basic workflow example (ssh)
Basic workflow example (web interface)
Troubleshooting
Extra: AlphaFold-Multimer
VIBFold on the HPC
3 Topics
VIBFold: an adapted AlphaFold script
Instructions for running VIBFold
Overview of VIBFold outputs
AlphaPulldown on the HPC
2 Topics
Introduction
Instructions for running AlphaPulldown
RoseTTAFold
The RoseTTAFold pipeline
Online Accessiblity
Exercises
VIB Training Session (AlphaFold)
10 Topics
HPC: A first AlphaFold run
HPC: AlphaFold-Multimer
HPC: Generate PAE/PLDDT/MSA plots
HPC: Analyze the predicted SarsCoV2-VHH.E complex
HPC: Run AlphaPulldown on the HPC
ChimeraX: Display AlphaFold predictions and error estimates using ChimeraX
AFDB: Analyze the prediction for USP7
Colab: ColabFold on Google Colab
ESM: Run ESMFold via web interface
(old) HPC: Analyze the predicted SarsCoV1-VHH.72 complex
Solutions
VIB Training Session (AlphaFold)
3 Topics
Analyze the prediction for USP7 on AlphaFoldDB (solutions)
Analyze the predicted SARS-CoV-2 + VHH-E complex (solutions)
(old) Analyze the predicted SARS-CoV-1 + VHH-72 complex (solutions)
Previous Topic
Next Topic

Official AlphaFold colab

AlphaFold and friends on the HPC Online accessibility Official AlphaFold colab

In addition to all code and models being released, DeepMind also released a Google Colab notebook (with a slightly less accurate model version, due to the host medium). This allows the end user to run AlphaFold without any installation or hardware requirements. However, this approach comes with its limitations.

Google Colab

Google is free Jupyter notebook environment that allows to run python code in the cloud. It is hosted on Google Cloud, and connects to your Google Drive account.

Example of running code in Google Colab.

It was designed mainly for educational purposes. Access is free but restricted, meaning that notebook sessions are limited in time, and the allocated time diminishes in case of frequent use. To run AlphaFold, specialized GPU hardware is allocated in the form of a 12 GB or 16 GB GPU, allowing to predict structures for sequences of up to approximately 1000 and 1400 amino acids respectively. At the same time, you are not guaranteed to always have this specialized hardware at your disposal.

Official AlphaFold colab

The official online notebook can be found at https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb.

An illustration of how to run the Google Colab notebook, and of the outputs that are produced.

Another limitation of this notebook is that apart from the information shown above, there are no further files downloaded. Only one predicted structure is returned, in addition to the pLDDT and the predicted aligned error.

Previous Topic
Back to Lesson
Next Topic