“`html
Innovative Approach to Drug Design
A new British startup is convinced that artificial intelligence holds the key to accelerating the traditionally slow and intricate process of drug design. Potential Sciences aims to bypass conventional laboratory methods by creating large-scale chemical datasets grounded in fundamental physics principles. By simulating interactions between atoms, small molecules, and ultimately complex amino acids and proteins, the company believes it can train AI models and predictive systems to identify highly specific drugs for particular diseases, thereby circumventing costly and time-consuming laboratory experimentation.
“Our goal is to ensure that molecules unlikely to succeed in clinical settings fail quickly,” explains Hatem Helal, founder and CEO of Potential Sciences. “This requires better predictors of how compounds will behave when interacting with human biology.” The startup, which launched on 5 November, has reportedly secured initial funding in the tens of millions from SovAI (the UK government’s AI venture capital fund) and Twin Path Ventures, among other investors. The founders chose not to disclose the exact figures or valuation.
Potential for Accelerated Treatment Development
If Potential Sciences achieves its anticipated outcomes, it could pave the way for the expedited development of new treatments across a broad spectrum of diseases. By reducing the cost associated with experimental procedures, Helal suggests that the company’s approach might also render it economically viable for pharmaceutical companies to explore medications for rare diseases that previously lacked treatment options. “We aspire to enable more personalised therapies for illnesses; the concept that medications can be unique molecular combinations tailored to the singular biology of each individual,” he elaborates.
The crux of the challenge lies in whether Potential Sciences can accurately predict the ability of candidate drugs to bind with disease-related proteins, which would necessitate a significant leap forward in technology. “I am confident there will be improvements,” states Peter Wolynes, a chemistry professor at Rice University. However, he cautions, “currently, protein engineering is at the level of the Romans, not the Renaissance.”
AI’s Role in Scientific Discovery
Prominent figures in the sector, including Nobel Prize-winning chemist Demis Hassabis for his work at Google DeepMind, have long asserted that advancements in AI will lead to groundbreaking scientific discoveries, similar to the recent transformations witnessed in mathematics. Startups like Potential Sciences hope that drug discovery will be one of the first scientific frontiers to be revolutionised by this technology.
At present, a substantial number of AI-driven drug discovery labs develop models based on a singular public database of three-dimensional atomic structures (the Protein Data Bank), which researchers have expanded over decades with new experimental results. According to Helal, this dataset remains insufficiently large to train AI models advanced enough to reliably expedite the search for viable drugs. “Out of all the atoms in the universe, there are around 1060 ways to assemble them into molecules that could potentially be administered to a human,” he notes. “Consider drug discovery as a search problem; it is about how swiftly we can explore and refine that space to identify promising molecules for a specific disease.”
Creating Comprehensive Datasets
To tackle this issue, Potential Sciences is striving to develop much larger datasets, composed of synthetic data generated by simulating interactions among various chemical compounds. Theoretically, with these more extensive datasets, the startup will be able to train far more sophisticated AI models and predictive systems.
Academics are optimistic about the simulation-based approach embraced by Potential Sciences. “If we could accurately calculate the interactions between all molecules within a cell and the fluctuations in their structure, we could predict where molecules would bind, which would undoubtedly revolutionise drug development,” proposes Alan Fersht, an emeritus professor of chemistry at the University of Cambridge, specialising in protein engineering. “I am optimistic and believe this will eventually happen.”
Government Support and Competitive Landscape
By securing investment from SovAI, Potential Sciences joins other AI-driven drug discovery firms such as Basecamp Research, Isomorphic Labs, and the materials discovery startup CuspAI. The UK government established SovAI earlier this year with the mandate to invest in British startups capable of becoming essential components within the global AI supply chain, thereby granting the country crucial geopolitical influence. The fund has approximately $800 million at its disposal and the capability to provide its portfolio companies access to the UK’s national AI supercomputer.
“The physical approach (the chemistry physics) differentiates Potential Sciences from other companies,” states Joséphine Kant, venture capital director at SovAI, who has joined the startup’s board as part of the investment. “By granting them access to computational resources, they can perform this work at a scale that would be very challenging for a small startup to achieve.” However, the market for AI-driven drug discovery is highly competitive. Even with the backing of the UK government, Potential Sciences will contend with numerous entrepreneurs and academics aiming to address the same challenges from various perspectives.
Challenges of Drug Development
The ability to more swiftly identify promising drugs does not eliminate the costs and time involved in testing their safety and efficacy. Consequently, new treatments will still take considerable time to reach the market, regardless of advancements in AI. “The cost of designing a drug pales in comparison to determining what effects it will have on mice, severely ill patients, or healthy individuals,” clarifies Wolynes. “In some respects, this is not the limiting factor.”
As the field progresses, startups like Potential Sciences must also contend with the potential misuse of technologies capable of identifying effective treatments to develop highly dangerous substances. “If we can design tools that perfectly minimise toxicity, it stands to reason that some might attempt to maximise it,” Helal states.
Potential Sciences does not yet have all the answers. The startup has yet to define how it will generate revenue from the various technologies it is developing or decide whether it will make its models publicly accessible, given the potential for misuse. Nevertheless, Helal believes there are few issues more pressing than attempting to solve with AI.
“There have been many waves of promising advancements that have not had a significant impact on drug discovery,” Helal highlights. “It is a fundamental and highly important problem that we must strive to address.”
“`