STFC welcomes exciting breakthrough in the fight against disease
02 Dec 2020
No
- Marion O'Sullivan

 

 

At this week's CASP14 conference Google DeepMind announced a major breakthrough from its Artificial Intelligence programme, Alphafold2.

No
Model of a protein complex

​​

 

 This largely solves the protein folding problem for single proteins – a tremendous scientific achievement overcoming a grand challenge that has been sought for over five decades and that could shape the future fight against diseases like cancer.

For the first time, the CASP conference (Critical Assessment of Techniques for Protein Structure Prediction) is being hosted by the Science and Technology Facilities Council (STFC) through its staff within the Collaborative Computational Project for Electron Cryo-Microscopy (CCP-EM) and STFC Video Conferencing. 

Tom Burnley from STFC's Scientific Computing Department, and who leads CCP-EM, said, “It was tremendously exciting to be hosting the CASP conference this year thanks to the support of STFC video conferencing.  The big reveal of the AlphaFold2 results was truly breath-taking and a major advance for the field of structural biology.  For me, the most exciting thing is this will to help us solve many more advanced questions to fully understand the molecular basis of life: how do simple proteins form larger functional complexes, how do proteins move and how do proteins interact in their natural cellular environment.  We can predict what an individual looks like but how does it interact with others, how does it dance and where does it go to work?

“It should also be emphasised the critical role open science played in this development - without open sequence and structural databases the AI algorithms would not have the knowledge to learn from.  STFC has played a vital role here in providing sustained access to state-of-the-art instrumentation at its Daresbury Laboratory and at the Diamond Light Source, and through software support from CCP4 and CCP-EM to generate the reservoir of structural knowledge AlphaFold could mine.  Furthermore, during CASP14 the CCP4 program AMPLE (developed in collaboration with Rigdon Group, Liverpool) was also used to validate the predictive power of the AlphaFold structures."

Ronan Keegan, part of the CCP4 team in STFC Scientific Computing, explains, "The CCP4 team, in a collaboration with the Rigden Lab Bioinformatics group at the University of Liverpool and the Max Planck institute in Tubingen, Germany, assessed the accuracy of the predicted protein models from all groups involved in the CASP competition, including AlphaFold2. In particular, we examined their suitability for solving what's known as the "Phase problem" in X-ray crystallography, a difficult challenge to overcome when eliciting the structure of proteins experimentally using facilities like the Diamond Light Source.

“To succeed in solving this problem, a predicted model needs to have less than 2 Angstroms deviation from the true protein structure over the length of its fold. For almost all of the targets we looked at, the AlphaFold models were well within this threshold, often being less than 1 Angstrom. Making use of the "AMPLE" software, jointly developed by CCP4 and the Rigden group, these models and many of those created by the other CASP competitors could be enhanced to make the solution to the Phase problem even more tractable."

The Collaborative Computational Projects CCP-EM and CCP4 are not directly involved in the Alphafold programme but they play an important role in developing software and providing knowledge, access to vital instrumentation, and infrastructure support.

More information:

CCP-EM is supported by UK Research and Innovation through STFC and the Medical Research Council.

CCP4 is supported by UK Research and Innovation through STFC and the Biotechnology and Biological Sciences Research Council 

The CCPs are part of CoSeC

CASP14 Press release

CASP14 Conference

 



Contact: O'Sullivan, Marion (STFC,RAL,SC)