Nobel Laureates’ AI methods a boost to Uppsala chemists’ research

“We’ve identified molecules that could be developed into drugs to treat schizophrenia and psychosis,” says Jens Carlsson, professor of computational biochemistry.

This year’s Nobel Prize in Chemistry goes to discoveries that show how new proteins can be built and how their structures can be predicted with the aid of artificial intelligence (AI). In 2003, one of the prizewinners David Baker, an American, successfully designed a new protein with an innovative 3D shape. His lab has subsequently shown how you can develop proteins that can be used as drugs, vaccines, nanomaterials and minimal sensors.

The other two prizewinners – Demis Hassabis and John Jumper from Google DeepMind, UK – presented the AI model AlphaFold in 2020. Thanks to this model, you can now predict the complex structures of proteins with a high degree of accuracy. The same AI tool is also being used at Jens Carlsson’s lab at the Department of Cell and Molecular Biology.

“Through experiments, researchers have determined about 200,000 protein structures over the past 60 years. AlphaFold is now complementing this work by creating models for an additional 200 million proteins, which will be a fantastic asset for us,” says Jens Carlsson.

Creating molecules that can become drugs

The next step is to figure out how the proteins work using their structures. To determine whether a protein can bind a particular drug molecule, it’s very useful to know its structure. Jens Carlsson’s many years of research on computer simulations and receptor proteins is a valuable asset in this field. A protein is a chain built from 20 different amino acids that folds into a three-dimensional structure, and the structure depends on the order in which the amino acids are linked.

This summer, Jens Carlsson’s research group published a study describing a model of the trace amine–associated receptor’s (TAAR1) unknown three-dimensional structure that they had created with the aid of AlphaFold.

“When we compared our models created by AlphaFold with experiments, the structures generated by AI turned out to be extremely good. Our research colleagues at Karolinska Institutet have also shown that the drug molecules we created using the model are promising starting points for the development of drugs to treat schizophrenia,” says Jens Carlsson.

Three-dimensional molecules depicted on computer

The next step is meeting with investors to see how far the project can progress. The hope is to start a company that can improve the molecules and ultimately develop a new drug. However, when Jens Carlsson and the other computational chemists at SciLifeLab conduct their basic research, computers are their main tools.

On one of the computer screens, doctoral student Alejandro Díaz-Holguín shows spiral-shaped, three-dimensional models of the trace amine-associated receptor 1 (TAAR1) in blue and purple. In between, you can glimpse atoms connected by bonds as small 0.0000000001 meter, which together form a molecule that binds to the receptor.

What do you want to achieve in your research?

“Our goal is to find molecules that can be used in experiments to switch receptors on and off. When you get an adrenaline rush, it’s the adrenaline molecule that activates a receptor. We want to find such molecules for all the hundreds of receptors found in the body, and we have no idea yet what many of them actually do. Since 30 to 40 percent of all drugs bind to receptors, they are very important for our health. Based on this knowledge, we could develop drugs to treat many diseases,” says Jens Carlsson.

Anneli Björkman