Harnessing tech to revolutionize drug discovery
For 50 years, pharma drug discovery has been pursued in the same way, through trials in traditional wet lab environments. Now a new breed of companies is working with artificial intelligence and machine learning to try to revolutionize that process.
Laksh Aithani previously worked with Exscientia, one of the first companies to use these technologies to predict molecular properties. Today, as co-founder of Charm Therapeutics, he is taking AI into a new space: predicting how a protein and a small molecule bind together.
Aithani believes this mystery will be solved within five to ten years. “Once you do that, you can essentially identify a hit against any target,” he says. “The next part of the puzzle, which is potentially much harder, is how do you optimize that hit into a lead, and then finally into a drug?”
At Congruence, CEO Clarissa Desjardins and her team are also using AI and machine learning, to target mutated proteins that cause conditions such as genetic obesity and Parkinson’s disease.
Their mission will require cracking the mysterious movement patterns of small molecules and the proteins themselves. “We really don’t understand what those forces are. We can simulate some of them, not all of them, and that’s where AI and machine learning are going to help,” says Desjardins.
Hungry tech feeds on old and new data
Working with AI requires use of large data sets, including the Protein Data Bank, which has been amassing protein ligand structures for around 50 years. There are now more than 200,000 structures in the collection – yet even that is not enough, says Aithani.
“We’ve realized that for the projects we work on, it can be very helpful to generate our own data,” he says. “That’s why we’ve built an in-house crystallography capability.
“We’ve generated over 200 proprietary structures, which sounds like a small amount – but when it’s really focused data on a single protein, it makes a huge impact on the accuracy of the model.”
Desjardins also uses publicly-available databases for Congruence’s work on protein thermodynamics. However, she notes that many databases disappear abruptly, as the academics who launched them retire or move to another field.
“We have a long way to go until we have really robust databases that we can train sophisticated models on,” she says. “But for an individual protein, we can generate our own data.”
“We’ve generated over 200 proprietary structures, which sounds like a small amount – but when it’s really focused data on a single protein, it makes a huge impact”
Laksh Aithani, CEO and Co-founder, Charm Therapeutics
Capabilities beyond human understanding
Technology is essential to overcome human limitations in comprehending the circuitry of a cell. While humans may never be able to predict how a protein and a small molecule bind together, this is AI’s real value in drug discovery, rather than merely replacing human scientists’ existing activity.
Desjardins notes how biotech industry is still “completely embryonic” in its use of AI and machine learning. She cites a recent review of generative AI in generating small molecules in silico, which found only 55 papers. “This is a drop in the ocean; we are just at the beginning of this exercise,” she says.
AI/ML platform face difficulties in valuation
Investors struggle to value AI/ML drug discovery platforms due to a lack of comparatives and methods to evaluate their effectiveness. As a result, they often focus on the efficiencies these tools bring to the development of new drug candidates. Aithani notes, “Right now, whether you like it or not, if you’re going to raise funds from biotech specialist investors, you’re going to be classified as a biotech company. Platform companies are finding it pretty difficult given the path of interest rates.”
Recognizing these limitations, companies like Congruence have adopted a strategy of building their pipeline and demonstrating the efficiencies gained in the development of new drug candidates using their platform technologies. Desjardins remarks, “[Today], we’re not investing the majority of our R&D dollars to protect this platform. Maybe someday we can go back when we’ve earned the right to then fully invest in all of the potential of the platform.”
Mission to design molecules in silico
Both Aithani and Desjardins foresee the ability to design molecules in silico that bind to selected proteins, making the toughest part of the drug discovery process much more efficient.
Aithani believes pharma’s use of AI in molecular properties will become similar to its use of contract research organizations for synthetic chemistry today.
“As soon as the AI becomes good enough to start massively cutting down on the molecules you need to make and test, that’s when you’re going to start getting real spend on software,” he says.
“As soon as the AI becomes good enough to start massively cutting down on the molecules you need to make and test, that’s when you’re going to start getting real spend on software”
Laksh Aithani, CEO and Co-founder, Charm Therapeutics
For now, innovators are grappling with key conundrums, such as whether protein binding is driven by enthalpy or entropy. “These limitations will be overcome, and there will be a day where we can discover small molecules in silico,” says Desjardins. “And I think the implications of that are mind-boggling.
“I’d be happy if in my lifetime, we could go from choosing a protein to finding a molecule that hits on it – and it seems like every week and every month, there’s progress being made on that front.”
“There will be a day where we can discover small molecules in silico – and I think the implications of that are mind-boggling”
Clarissa Desjardins, CEO, Congruence Therapeutics
