For molecular biology to understand how a protein folds on itself in a leap that can be compared to that of the first man on the moon. The final form it assumes, in fact, conditions the properties of the protein from a physiological and pathological point of view In 1969, the same year as Apollo 11, an American molecular biologist said that the duration of the universe would not be sufficient to find the correct one between all possible combinations into which a protein can fold. It was thought that it would take decades to achieve this, said Venki Ramakrishnan, 2009 Nobel Prize in Chemistry.
Thanks to artificial intelligence (AI), on the other hand, it has already become possible today to model in 3D with an accuracy of more than 90% the folding of a protein starting from a one-dimensional sequence of amino acids, the building blocks of it. It will facilitate the creation of new drugs and the understanding of details that still elude us in the biology and structure of components essential to life such as proteins. Over the years, various bioinformatics methods have developed to predict how bending occurs, but they had wide margins of error, explains Daniela Corda, director of the Department of Biomedical Sciences at Cnr. This new means is important for the possibility of accelerating the development of drugs, bypassing a long and laborious phase that until now passed through the purification and crystallization of the protein to be studied.
DeepMind, a London-based company linked to Google through the holding Alphabet, announced that it had solved the dilemma that had lasted for over 50 years thanks to AlphaFold, its AI system that participated in a specific competition, Casp (Critical Assessment of Structure Prediction), in which various computational models for a quarter of a century have been confronting and challenging each other to solve the mystery of how proteins fold. But why is it so important to understand how proteins fold? The DNA that contains the genome of all living beings in order to function must be translated into long strands of amino acids which then form proteins, which to become active fold several times into particular structures of various shapes, continues Corda. Both the interaction with other proteins and the formation of molecular complexes, essential for the creation of complex cellular structures, depend on the way of folding.
Neural network system
AlphaFold is not a normal computer program but structured with connections that simulate the human brain. To arrive at the extraordinary result obtained with an infinitesimal margin of error, equal to the diameter of an atom, illustrates Alessio Bechini, professor of Bioinformatics in the course of Biomedical Engineering at the University of Pisa, the great increase in the computing power obtained was decisive. in the last few years. In fact, a tool like AlphaFold requires enormous computing power, unimaginable only ten years ago.
One more weapon
still uncertain whether DeepMind will share the technology with the scientific community. Demis Hassabis, co-founder and chief executive of the company, said that he plans to publish the details, but not before 2021. It may not be in time to use this new weapon to fight Sars-CoV-2, but certainly a tool that will help to counteract possible future pandemics that could take us by surprise such as Covid-19 with rapid and effective responses. When the user will be available without a doubt, concludes the CNR researcher. a huge step forward in finding new drugs.
December 1, 2020 (change December 1, 2020 | 22:53)