DeepMind, Alphabet’s AI, has published the 3D structure of (almost) all existing proteins. That’s why it’s important

He was born just four years ago. And since then he worked tirelessly, reaching today the incredible result he has revealed the molecular structure of about 200 million proteins, almost everything known to science. He is the protagonist of this business AlphaFoldartificial intelligence system developed by Deepmind of Alphabet (or the large Google family) and from European Laboratory of Molecular Biology (Embl), whose database has grown from 350 thousand molecules a year ago, exactly, to 200 million today. A huge database that can be consulted for free by all scientists in the world and that makes a fundamental contribution to the fields of pharmacologyfrom biologyfrom medicine and more: “Just in the last year,” he commented on the matter Sameer Velankarhead of the European Embl-Ebi Protein Databank team, “over a thousand scientific articles have been published on a wide range of research topics using AlphaFold’s facilities. And this is just the impact of one million predictions: imagine what that 200 million could be , all open in our database.”


The future as seen by Annalisa Barla, the teacher who teaches machine learning

by Emanuele Capone

To understand the scope of AlphaFold’s work and how it works, we must first remember what proteins are and how they are made. It is about molecules that are one of the main “ingredients” of all living forms, consisting of individual units – amino acids – arranged three dimensional in the place. Although they exist only twenty amino acidstheir possible spatial arrangements are many and each arrangement corresponds to a different protein and to each protein different biological characteristics and functions. Knowledge of the three-dimensional structure of a protein is crucial, because in addition to providing us with information about its operation, it also gives us instructions on how to modify, block or adjust it. It’s a bit like having twenty building blocks that we can put together in a million different configurations to make an infinite number of different objects, and AlphaFold has just given us the instruction book to make two hundred million of those objects.


Conscious deep learning is still a utopia

by Andrea Daniele Signorelli

Until the refinement of artificial intelligence algorithms, revealing the structure of a protein was extremely complicated: it was basically done by trying to observe the molecules under a microscope or X-rays. AlphaFoldInstead, it works in a completely different way, by exploitation machine learning, bioinformatics and structural biology techniques: his “brain” has studied the structure of thousands of known proteins and learned to predict the shape of others from this. And he has learned it very well: currently, given a list of amino acids as input, he is able to predict the three-dimensional structure of a protein with an accuracy comparable in two-thirds of the time to that of microscope and X-ray experiments. And it is just the beginning, say DeepMind: “AlphaFold is a glimpse into the future and what could be possible by applying the computational techniques of artificial intelligence to biology. Just as mathematics is the perfect descriptive language for physics, we believe that artificial intelligence is the right technique to tackle the dynamic complexity of biology. We believe that we are pioneers in the field of ‘digital biology’ and look forward to these tools that will help us understand the fundamental mechanisms of LIFE.”

Artificial Intelligence

Cloud and artificial intelligence: IBM wants to revolutionize materials chemistry

by Andrea Nepori, Bruno Ruffilli

Leave a Comment