The Francis Crick Institute has announced a partnership with Deep Mind, a pioneer in artificial intelligence software affiliated with Google. This cooperation is the first of its kind - and the only one so far - that DeepMind concludes with a research institute outside the company's borders.
According to a press report published by the Francis Crick Institute on July 6, this collaboration aims to advance our understanding of biology, especially in the areas of protein design and genomics.
The partnership, which was concluded for several years, includes the establishment of a DeepMind research laboratory within the Francis Crick Institute, to combine the expertise of the Institute's scientists with DeepMind's expertise in artificial intelligence in order to study the mechanisms of health and disease.
Today’s artificial intelligence is already having a direct impact on our lives. We see examples of it in self-driving technologies and in tracking headphones.
The idea of establishing the Francis Crick Institute began in 2007, in response to a report by policy advisor David Coxey, who was appointed by the Queen of Britain to lead a team developing the biological sciences and structuring a new approach to medical research in the United Kingdom.
By the summer of 2011, the scientific vision for the project - and the unique foundations for a state-of-the-art building design - was nearing completion. The project is named Francis Crick, after the British scientist who contributed to the understanding of the structure of the "double helix" (DNA). The institute was officially inaugurated by the Queen in November 2016, and it was fully operational by spring 2017.
The institute - headed by Paul Nurse, winner of the Nobel Prize in 2001 - has been working since its inception to attract the best scientific minds in the field of biomedicine globally. Despite its modernity, the institute ranked fourth in the world according to the Nature Index classification of research institutes in the field of biosciences in 2022.
On the other hand, DeepMind was established as an independent company in 2010 in the United Kingdom. She joined Google in 2014. The company aims to develop machine learning algorithms in a way that simulates - or outperforms - human intelligence, hence the name of artificial intelligence.
In 2016, an artificial intelligence program developed by DeepMind, known as "Alpha Go", was able to defeat Lee Sedol, the Korean player and one of the professionals of the ancient game "Go", which is the most complex board game in the world. .
AlphaGo beat the player with the second most international victories in this game. The program won him in 4 games to one. This is a shock to AI experts with what this intelligence may achieve in the near future.
Artificial Intelligence in Biology
Demis Hassabis, CEO and co-founder of DeepMind, realized that the AI algorithms developed by his company had the potential to solve fundamental problems. Knowing the three-dimensional structure of proteins has been one such dilemma that has stymied researchers for 50 years.
Scientists use complex and very expensive techniques to learn the exact shape of each protein inside the body and how it works, and then design drugs for any defect that may affect it. However, discovering the structural form of one of these proteins requires years of hard work.
So, DeepMind decided to tackle this problem. And I worked on designing a "machine learning" algorithm that can learn from the structure of known-shaped proteins - which scientists have discovered before - in predicting the structure of unknown proteins.
Indeed, in 2020 and 2021, the company succeeded in designing an artificial intelligence program known as "Alpha Fold2" that could predict the three-dimensional shape of proteins with high accuracy, close to scientific results. This was the most complex that DeepMind had ever succeeded in so far.
A partnership that pushes the boundaries of science
Knowing the three-dimensional shape of cellular mechanisms - such as proteins and others - is the basis for understanding their function. It is also the cornerstone of new drug design.
In fact, the shift from games to science was the main reason Hassabis set up his company; Company scientists are now trying to develop AI algorithms to understand fundamental dilemmas in science.6
However, the problems facing the biologist are still very complex. Although AlphaFold2 was successful in predicting the structure of individual protein units, it was not as successful for protein complexes composed of a number of these single units.
Therefore, the collaboration of AI scientists with laboratory scientists has entered a new phase in order to understand the complex foundations that play a fundamental role in health and disease.
This is confirmed by the recent partnership between the two giants of medical research and artificial intelligence. Which may promise to open a new field in medical research in the next few years.