AI Sifts Through a Mind-Boggling 10 Sextillion Molecules to Find New Medicines

Uppsala University researchers develop computational method that drastically cuts time and cost of drug discovery

Finding a needle in a haystack seems simple compared to what pharmaceutical researchers face when searching for new drugs. Scientists at Uppsala University have developed a computational approach that can evaluate an almost unfathomable number of potential drug molecules—10 sextillion (that’s a 1 followed by 22 zeros)—to identify the most promising candidates for development.

The study, published in Nature Communications this February, demonstrates how virtual fragment screening can identify molecules that inhibit OGG1, an enzyme involved in DNA repair that plays a role in inflammation and cancer.

“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,” says Jens Carlsson, professor of Computational Chemistry at Uppsala University and one of the study’s authors.

Building Drugs Like a Jigsaw Puzzle

Traditional drug discovery often relies on high-throughput screening—physically testing hundreds of thousands of compounds against a target protein—which is both expensive and often unsuccessful. The researchers instead used fragment-based drug design, starting with tiny molecular fragments and building them into more complex drug molecules.

“It’s like doing a jigsaw puzzle. We start with one piece of the puzzle and then gradually build up a drug molecule by adding new pieces. In the end, we have a drug molecule that fits the target protein perfectly,” explains Carlsson.

The team began by computationally screening 14 million fragment-like compounds against the OGG1 enzyme. After identifying promising fragments, they conducted further searches among billions of readily synthesizable compounds, followed by laboratory testing of the most promising candidates.

This approach successfully identified molecules that inhibited OGG1 and demonstrated anti-inflammatory and anti-cancer effects in cellular tests. The crystal structures of the protein-fragment complexes confirmed that the compounds bound to OGG1 exactly as predicted by the computational models.

From Billions to Sextillions

While the initial searches were confined to commercially available molecules, the researchers later expanded their horizons dramatically. PhD student Andreas Luttens developed a new computer program capable of generating and evaluating all possible molecules that could theoretically exist within certain chemical constraints.

This expanded the search space to approximately 10 sextillion potential molecules—more than the number of stars in the observable universe. The team demonstrated that their computational method could efficiently navigate this vast chemical space to identify the most promising candidates.

“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,” says Carlsson.

Advantages of Virtual Fragment Screening

The researchers compared their fragment-based approach with traditional screening methods and found it to be significantly more efficient. Their analysis showed that by using fragment-based design, they could evaluate all theoretically possible fragment-like molecules and their relevant elaborations by docking just two billion compounds in two steps.

In contrast, selecting two billion compounds randomly from the vast lead-like chemical space would be unlikely to contain any relevant candidates for a specific target.

The team also tested their approach on three additional protein targets linked to cancer or inflammation (SMYD3, NUDT5, and PHIP) and found that in each case, they could efficiently identify promising compounds for further development.

A New Era for Drug Discovery

While computational methods can now rapidly identify promising drug candidates, translating these virtual discoveries into real-world medicines brings new challenges.

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

Nevertheless, this approach represents a significant advancement in drug discovery methodology. The ability to virtually screen vast chemical libraries and focus resources on the most promising candidates could dramatically reduce the time and cost of bringing new medicines to patients.

The researchers’ OGG1 inhibitors displayed better physicochemical and pharmacokinetic properties than previously identified compounds, demonstrating the real-world benefits of this computational approach.

The study—conducted in collaboration with Karolinska Institutet and Stockholm University—highlights how computational methods are transforming drug discovery from a largely experimental process to one increasingly guided by sophisticated algorithms and models.

As computing power continues to increase and these methods are refined, the ability to efficiently navigate the vast landscape of possible drug molecules will likely lead to faster development of more effective medicines for a wide range of diseases.