New AI algorithm searches 10 sextillion drug candidates

A study from Uppsala University, published in Nature Communications, reveals a new computational method capable of searching through an astonishing 10 sextillion (10²²) potential drug molecules. The approach, developed in collaboration with Karolinska Institute and Stockholm University, has the potential to significantly speed up the costly and time-consuming drug development process.

Harnessing the power of supercomputers

The research team, led by Professor Jens Carlsson from Uppsala University, utilised advanced computer algorithms to model drug molecules and identify promising candidates. Traditional drug development often involves screening large libraries of chemical compounds, which is both expensive and frequently yields limited results. In contrast, this new method employs fragment-based drug design to pinpoint tiny molecules that can bind effectively to target proteins.

“We use the computer models to search through databases containing billions of molecules. This method will be able to speed up the costly drug development process,” said Professor Carlsson.

Targeting the OGG1 enzyme for anti-inflammatory effects

The study focused on the enzyme OGG1, a protein responsible for repairing DNA damage within cells. By identifying molecules that bind to OGG1, the researchers aimed to influence the enzyme’s activity and achieve an anti-inflammatory effect. Laboratory tests on over a hundred designed molecules demonstrated their potential, as many successfully inhibited the enzyme and exhibited anti-inflammatory properties.

“It’s amazing that we can now design molecules and show that they actually work exactly as we hoped. The same strategy will work for many other proteins and diseases,” Professor Carlsson added.

The research utilised fragment-based drug design, beginning with simple molecular fragments and gradually building them into complex drug molecules. This approach is likened to assembling a jigsaw puzzle, starting with a single piece and meticulously adding to it until the drug molecule perfectly fits the target protein.

Screening a vast number of drug candidates is vital for discovering effective treatments. AI has proven to be an invaluable tool in this process, allowing researchers to efficiently navigate large pools of molecules and identify those with the highest potential for selectively targeting specific diseases.

The team partnered with a specialised company that manufactures molecules on-demand. Using custom-built computer programs and supercomputing power, they screened billions of available molecules to find suitable candidates. The researchers then pushed the boundaries of this technology, with PhD student Andreas Luttens developing a program capable of generating all possible molecules, expanding the search to a staggering 10 sextillion options.

“With our strategy, we can search through sextillions of drug molecules very efficiently. In the near future, we will be able to test all potential drug molecules in our computer models – a breakthrough that has great potential,” said Carlsson.

Future challenges for medicinal chemists

While the computational approach offers remarkable opportunities, challenges remain in translating digital discoveries into real-world drugs. Many of the theoretically designed molecules may not yet be producible with current chemical methods.

“We’ll need to develop new methods in the future in order to successfully develop the molecules that computations can design so quickly,” Professor Carlsson concluded.

This pioneering study not only demonstrates the immense potential of computational chemistry but also sets the stage for a new era of drug development, where researchers can efficiently explore vast chemical spaces and accelerate the creation of innovative treatments.

This study was published in Nature Communications.