The world of AI-powered drug discovery is expanding as the capabilities of machine learning grow. An approach that seemed unthinkable a few years ago is to simulate the complicated interaction of two interlocking molecules — but that’s exactly what drug designers need to know, and exactly what Charm Therapeutics aims to do with its DragonFold platform.
Proteins do just about everything it’s worth doing in your body, and are the most common drug targets. And to create an effect, you first have to understand that target, specifically how the chain of amino acids that make up the protein “folds” under different conditions.
In the recent past, this was often done with complex, time-consuming X-ray crystallography, but recently machine learning models such as AlphaFold and RoseTTAFold have been shown to be able to produce just as good results, but in seconds instead of weeks or months.
The next challenge is that even if we know how a protein folds in the most common conditions, we don’t know how it can interact with other proteins, let alone new molecules specially made to bind to them. When a protein meets a compatible binding agent or ligand, it can transform completely, as small changes can cascade and reconfigure the entire structure – in life, this leads to things like a protein opening a passageway to a cell or exposing a new surface that activates other proteins, and so on.
“That’s really where we innovated: we built DragonFold, the industry’s first protein and ligand folding algorithm,” said Laskh Aithani, CEO and co-founder of Charm Therapeutics.
“Designing drugs that bind very tightly and selectively to the disease-causing protein of interest (ie avoiding binding to other similar proteins necessary for normal human functioning) is of paramount importance,” he explained. “This is easiest if one knows exactly how these drugs bind to the protein (the exact 3D shape of the ligand that is bound to the disease-causing protein). This makes it possible to make precise modifications to the ligand, so that the can bind more tightly and selectively.”
You can see a depiction of this situation at the top of the article: The small green molecule and the purple protein fit together in a very specific way that isn’t necessarily intuitive or easy to predict. Effective and efficient simulation of this process helps screen billions of molecules, similar to previous processes that identified drug candidates, but goes further and reduces the need to experimentally check whether they interact as expected.
To achieve this, Aithani enlisted David Baker, designer of the RoseTTAFold algorithm among others and head of an influential lab at the University of Washington, as its co-founder. Baker is known in academia and industry as one of the leading researchers in the field, and he has published numerous articles on the subject.
Shortly after algorithms were shown to be able to predict protein structures based on their sequence, Baker determined that they could also “hallucinate” novel proteins that worked as expected in vitro. He is clearly leading the way here. And he won a $3 million Breakthrough award in 2020—certainly to a tech co-founder. Aithani also proudly noted the presence of DeepMind veteran Sergey Bartunov as director of AI and former pharmaceutical research leader Sarah Skerratt as head of drug discovery.
The $50 million A round was led by F-Prime Capital and OrbiMed, with participation from General Catalyst, Khosla Ventures, Braavos and Axial. While such large amounts are not uncommon for software startup, Charm does not stop at developing the ability to characterize these protein-ligand interactions.
The company’s early funding was used to build the model, but now they’re moving on to the next step: positively identifying effective drugs.
“We have the first version [of the model] done, and that has been validated in silico,” said Aithani. “We will validate it experimentally in the coming quarters. Note that the ‘product’ will be primarily for internal use to help our own scientists discover potential drugs to which we own 100% of the rights.”
Normally, the testing process involves wet-lab screening of thousands and thousands of candidate molecules, but if it works as advertised, DragonFold should cut that number down massively. That means a relatively small lab on a relatively small budget could potentially use a drug for which a major pharmaceutical company invested perhaps hundreds of millions a few years ago.
Given the profit profile of a new drug, it’s no surprise that the company has attracted this kind of investment: a few tens of millions is a drop in the ocean compared to the R&D budget of a major biotech research company. All it takes is one slap and they laugh. It’ll be a while, but discovering AI drugs also shortens timelines – so expect to hear about their first candidates sooner rather than later.