Alphafold and friends

This course material is originally designed for a synchronous in person or online course but can vastly be followed asynchronously, at the desired pace and in the absence of an instructor.

This tutorial material was created by Jasper Zuallaert (VIB-UGent), with the help of Alexander Botzki (VIB) and Kenneth Hoste (UGent). For questions and remarks, feel free to contact jasper.zuallaert@vib-ugent.be.

Jump to the content

Overview

At the end of 2020, results for the CASP14 (Critical Assessment for protein Structure Prediction) were released. Just like in 2018, DeepMind Technologies won the competition with their deep (machine) learning based AlphaFold prediction models. However, the leap in prediction accuracy now raised the performance to a level that had several big names in the field consider the protein structure prediction task as ‘solved’.

Architectural details, code and trained AlphaFold models were released by DeepMind in 2021. Given the high computational cost of deep learning algorithms, specialized hardware and software are required. Online solutions are available but come with considerable disadvantages. Therefore, the Flemish Supercomputer Center (VSC) provides high performance computing facilities, on which AlphaFold is installed and fully operational. This course gives a solid introduction on how AlphaFold can be easily and swiftly accessed using the HPC.

Audience

This course material is for students, postdocs, and researchers, from any scientific environment (academia, facilities, companies, etc.) interested in structure prediction using AlphaFold.

Competencies

At the end of the course, participants should be able to:

  • run protein structure prediction (monomer, multi-mer) using AlpahFold on the Tier-2 at the HPC UGent
  • compare the resulting predictions in their structural visualiser of choice
  • download existing structure predictions from AlphaFold DB

Learning outcomes

At the end of the course, the participants are expected to:

– interpret the output results of the respective AlphaFold monomer and multi-mer run

– list alternative methods for structure prediction next to AlphaFold

– explain the caveat of the use of AlphaFold for structure prediction of monomers as well as complexes

Prerequisites

Knowledge / competencies

Knowledge of the basic molecular biology is presumed. No bioinformatics competencies are required.

Technical

Command line knowledge and basic bioinformatics are an asset.
Jump here for a refresher of command line use and the OpenOnDemand web interface of the HPC:

Time required

5 hour

Additional information

Please note that participation in VIB courses is subject to our general conditions.

For more information, please contact training@vib.be.

This work is licensed under a Creative Commons Attribution 4.0 International License.

Course Content

Expand All
AlphaFold
RoseTTAFold
Exercises
Solutions