„AI Has Incredible Potential in Biology"

| By Leon Kirschgens & Jacqueline Plaster (transcribed)
Professor Blank, head of the MIX-UP project, is convinced: IT and biology will no longer be separate fields in the future. Artificial intelligence is becoming a significant pillar in basic research.

A few weeks ago, the Google company DeepMind achieved a sensational success. The AI software Alphafold managed to predict the structure of a protein with surprising accuracy. Will AI also help us with the plastic problem? Professor Lars Blank investigated that matter.

It's unbelievable: for decades, we biologists have been cutting our teeth on the question of how the structure of a protein comes about, i.e. how the process of protein folding works. After all, one small protein can have more diversity than all the atoms in the universe put together. And then the Google subsidiary "DeepMind" comes around the corner and solves the problem of protein folding within a competition as if it were nothing more than a computer game. DeepMind had been known for this until then - as an unbeatable opponent in PC games like "StarCraft II". So now the highly developed artificial intelligence has dedicated itself to the broad field of biology - great! We scientists can only benefit from this. The implications will be enormous for biology in the coming years, that much is certain. Certainly we will see advances in drug development.


But as great as the hype is - quite rightly - I have to put the expectations for the field of plastics into perspective: In the fight against plastics, we are not facing an immediate revolution with artificial intelligence. Because knowing the structure of a protein only helps us to a limited extent. Thanks to DeepMind, we can now go through the steps of DNA sequence, amino acid sequence and now protein structure very quickly. But what really interests us is the function of the protein, i.e. which structure of an enzyme can cleave which bond and how quickly.

Yet A Great Benefit

Nevertheless, AI makes our work easier, which is why Google's recent success is also an important step for us scientists in the fight against plastic. Now we are able to develop the complex enzymes, i.e. those that have to be able to hydrolyse polymer bonds, faster and in a more targeted way. Crystallising proteins is a time-consuming and not always successful intermediate step in determining their structures. Now, ideally, it is no longer necessary. In practice, this could help to develop highly efficient enzymes that break down plastics to their basic building blocks under natural conditions within a few hours - until now it has taken much longer, sometimes weeks, months and in the environment even years or decades.


Despite all this, first-principle approaches will still be needed to answer many other complex questions that are important for the development of enzymes. In that context, it will also be exciting to understand how DeepMind's first-principle approaches work. Because what really concerns us is the function of proteins, i.e. in the case of enzymes, the catalytic performance. These and many other questions cannot be answered with big data and sophisticated AI, or the data volumes are simply still lacking. This shows: AI will not replace us scientists, but will be available as a valuable tool.

 

everwave: We will keep you in the loop about this interesting topic...!