Google’s AI solves calculations in hours to create medicines that need days

Google is something more than the most used web search engine in Spain; is a tech and science giant that is so big that it had to create a parent company, Alphabet, to cover everything that is done.

For example, the development of Artificial Intelligence has been one of the top priorities in recent years, especially applied to medicine; is where he has achieved achievements in cancer screening.

Another project that has gained a lot of strength is that of DeepMind, which rose to fame after beating the best masters of Go, as well as learning chess and playing video games.

Google a for the secret of life

But from the start Google made it clear that the future of DeepMind went far beyond becoming a gaming master; It has been demonstrated today, with the announcement that its AlphaFold system has managed to solve one of the big problems facing the scientific community: protein folding.

The chemical composition of a protein does not explain everything; the most important part is how it folds on itself, creating microstructures, because if it doesn’t fold properly, it won’t work as it should in a biological system. This is especially vital for the development of new medicines. It doesn’t matter if the composition is correct: if it is not applied in the way it should, it will be useless.

That’s something science has known for 50 years, but so far I haven’t been able to do much respect, other than try all the combinations for ‘brute force’. Thus were born projects such as Folding @ Home, which take advantage of the unused power of our computers and consoles to process the protein fold to find the right solution. But even so, it just takes too long, several months.

The achievement of DeepMind

What the DeepMind team has accomplished with AlphaFold is predicting protein folding with great accuracy; specifically, it is able to know how it will fold with a margin of error of just the diameter of an atom. This process can take days or even just hours, so it is possible to get almost immediate results that can then be verified by researchers.

Although it is not an absolutely accurate system, it is enough to let scientists have to subtract the work. To understand the extent to which this advance can change everything, is the example of a team of German researchers spent a decade intending to find out the structure of a protein; once I got help from DeepMind, in just half an hour they managed to find the final structure.

A decade of work finished in half an hour; and that is just the beginning. In the future, it is expected that this human intervention will not be necessary, and that computers will be able to achieve the necessary accuracy to pronounce the folding of any protein; effectively, developing all its secrets and opening the doors to new developments.