onepot ai raises $13m to help make chemical drug creation easier

The Genesis of Onepot AI: Addressing a Critical Bottleneck in Drug Discovery

Daniil Boiko and Andrei Tyrin were both driven by a shared source of frustration, which ultimately sparked the creation of Onepot AI.

Boiko explained to TechCrunch that frequently, promising avenues in drug discovery were hindered not by biological complexities, but by the challenges inherent in synthesis. This process, the creation of novel molecules through chemical reactions, can be likened to assembling a complex puzzle from individual components.

The Hurdles of Molecular Creation

The creation of these fundamental molecular building blocks proves to be a remarkably difficult undertaking.

Boiko, currently pursuing a PhD in machine learning applied to chemistry at Carnegie Mellon University – following prior degrees in organic chemistry from a Russian institution – observed that drug researchers often abandoned potentially valuable concepts simply because the necessary chemical molecules appeared too difficult to produce.

Consequently, these compounds were never subjected to testing, hindering potential breakthroughs.

Tyrin, with a computer science background from MIT, discovered a similar lag while working on computational pipelines for drug discovery. He noted that while models could generate innovative ideas rapidly, laboratory execution often lagged significantly, taking months to catch up.

A Shift in Focus and Geopolitical Considerations

Both founders recognized a disparity in investment, with substantial resources directed towards molecular design while the complexities of actual molecule creation were largely overlooked.

Boiko further highlighted a growing geopolitical concern: the increasing vulnerability of global supply chains and the resurgence of competitive dynamics between the U.S. and China.

“A need became apparent,” Boiko stated, “to rebuild small molecule synthesis capabilities within the United States from the ground up.”

Introducing Onepot and Phil, the AI Chemist

This realization led Boiko and Tyrin to establish Onepot, centered around the small-molecule synthesis laboratory, POT-1.

They also developed Phil, an AI-powered organic chemist, designed to accelerate experimental analysis and streamline the compound synthesis process for their initial commercial clients – biotech and pharmaceutical companies.

On Wednesday, Onepot officially launched, announcing $13 million in funding encompassing pre-seed and seed rounds, with Fifty Years leading the seed investment.

The Current Landscape of Molecular Synthesis

Tyrin described the conventional approach to molecular synthesis: companies either maintain extensive in-house chemistry teams or outsource to contract research organizations located abroad.

Creating even a single compound through traditional methods can consume months of a chemist’s time and incur costs reaching thousands of dollars.

This process is characterized by extensive trial and error, involving the study of various compounds, data collection on biological activity and pharmacokinetics, and iterative experimentation. “The primary constraint isn’t the testing of compounds, but their initial production,” Tyrin emphasized. “Our goal is to reduce this timeframe to mere days.”

Onepot’s Streamlined Approach

Onepot’s offering is straightforward: the company maintains a catalogue of synthesizable molecules.

Clients select desired compounds, and Onepot’s technology synthesizes and delivers them – either as dry compounds or in solution – for customer experimentation.

The core of Onepot’s innovation lies in its backend, where Boiko and Tyrin meticulously analyze the intricacies of chemical synthesis to identify effective molecular combinations.

They have established a laboratory environment where Large Language Model (LLM) agents can access and learn from “molecule recipes,” enabling them to predict successful and unsuccessful compound building strategies.

Data-Driven Insights and Reproducibility

“During laboratory experiments, we meticulously record every detail of the process,” Tyrin explained, including temperature variations and the precise sequence of ingredient additions.

This comprehensive data capture ensures reproducibility, even years after the initial experiment.

Furthermore, their AI agents derive insights from real-world experimental data, rather than relying solely on information extracted from the internet.

Funding and Future Expansion

Boiko described the fundraising process as “demanding,” noting that their connection with their lead investor originated through a referral. “An anticipated brief meeting evolved into an extended whiteboard session focused on the industrialization of synthesis,” Boiko recounted.

Additional investors include Khosla Ventures, Speedinvest, OpenAI co-founder Wojciech Zaremba, and Google’s Chief Scientist Jeff Dean.

The newly acquired capital will fund the establishment of a second laboratory in San Francisco, enabling the company to serve a larger customer base and expand its team and compound discovery engine.

Onepot views WuXi AppTec and Enamine as key competitors in the service landscape.

A Vision for Accelerated Drug Discovery

Ultimately, Boiko and Tyrin aspire to accelerate drug discovery by at least a factor of two and to redefine the boundaries of what is considered achievable in chemical synthesis.

“We aren’t simply accelerating drug discovery; we are broadening the scope of possibilities for both drugs and materials,” Boiko stated. “The breakthrough drug that remains undiscovered may be within reach, awaiting our identification.”