"Deepmind has just shown that AI is invaluable in understanding how life works"

30 November 2020

The director of BSC Life Sciences department, Alfonso Valencia, is part of the jury that has just verified the amazing results of DeepMind, Google's sister company, when it comes to predicting the structure of proteins, one of the most complex problems in current biology.

"Deepmind has just demonstrated that Artificial Intelligence is invaluable in understanding how life works and will play a key role in the fight against diseases and other great challenges for humanity", says Alfonso Valencia, Icrea professor and director of the Department of Life Sciences of the Barcelona Supercomputing Center (BSC).

Valencia, with a long career in bioinformatics and with very relevant articles on how to study the structure of proteins, is part of the CASP-14 jury, an international competition with a long tradition before which DeepMind has shown the results of its AlphaFold program, dedicated to the prediction of protein structures.

Predicting the structures of proteins from their DNA sequence is an extremely complex problem and solving it is scientifically very relevant, since most biological processes depend on these structures and knowing them is enormously useful, for example, to understand diseases and be able to fight them.

“The jury has been completely astonished by the results of DeepMind. His program solves one of the challenges of biology of the last 50 years. It is comparable to knowing a second genetic code”, states the director of Life Sciences of BSC.

DeepMind, which became popular for its AI programs capable of beating world chess or go champions, has shown itself capable of solving these problems with a program based on Machine Learning and Deep Learning.

"Artificial intelligence allows biology (like most fields of science at the moment) to make a leap of magnitude based on correlating large amounts of data," explains Valencia.

However, although the potential of using AI for scientific research is enormous and appreciated by the community, the results obtained through these technologies are not results supported by the scientific method. "They allow us to obtain results, but they do not explain the processes by which they have been produced," says Valencia. "Understanding how the DNA of a protein results in a certain structure is a question that we still cannot answer," he says.

Alfonso Valencia is the author of one of the most cited articles among students of protein structure [1], in which he begins to draw the method by which the relationship between the amino acid sequences of a protein and its structure and functions can be deciphered.

Valencia is part of the jury of CASP Critical Assessment of Techniques for Protein Structure Prediction, an international competition between research groups to evaluate protein structure prediction methods that takes place every two years with the aim of advancing in this field of biology. This year, together with Alfonso Valencia, BSC researchers Alba Lepore, Camila Pontes, Victoria Isabel Ruiz and Edoardo Milanetti participated in the CASP jury.

 


  1. “Correlated mutations and residue contacts in proteins”. Ulrike Göbel, Chris Sander, Reinhard Schneider, Alfonso Valencia. Proteins , April 1994. https://doi.org/10.1002/prot.340180402