Data-driven drug discovery
Story originally published in BioTalent Canada’s Close-up on the bio-economy: Western Canada
For Vancouver-based start-up Variational AI, the future of drug discovery is most definitely digital. Its innovative machine-learning platform generates novel and optimized molecules — with the potential to eventually cut preclinical drug discovery times from years to months. The challenge lies in proving that potential to chemists, biologists and clinicians.
Q: What makes Variational AI’s solution different from conventional drug discovery platforms?
HANDOL KIM, CO-FOUNDER AND CEO: We use a generative artificial intelligence (AI) platform to discover novel small model therapeutics in a fraction of the time it takes with status quo techniques. Our AI can search the entirety of chemical space and rapidly generate new druglike molecules with optimized properties such as potency, synthesizability and selectivity. What makes us unique is that we can do this with less data than other AI-for-drugdiscovery approaches.
Q: What are the implications of your approach for the life sciences field?
HK: Our ultimate goal is to redefine the unit economics of drug development and create therapeutics for unmet medical needs faster while improving patient outcomes across a broad range of disease areas. There’s a big advantage to adopting and mastering these technologies because they’ll deliver outsized market share to the winners. Life sciences is being disrupted by digitalization just like other industries. AI and machine learning are becoming integral and strategic differentiators. We in Canada need to adopt computationally and digitally based drug discovery methods to maintain our competitive status. Global leaders already have this focus.
Q: Does the newness of your approach create HR challenges?
HK: It’s hard to find people who understand both the chemical and computational domains. Experts in one area are sometimes suspicious or dismissive of the other. Many chemists who have been around drug discovery for decades are understandably skeptical. I mean, there is still no approved AI-discovered drug and developing drugs is extremely difficult. I’d say, too, the Canadian educational system is structured so that by the time someone’s finished their PhD they’ve optimized and focused their skillset and cheminformatics and AI are not seen as essential — yet. At the same time, machine learning researchers are often naïve about how difficult drug discovery is, especially with respect to the quality and amount of training data needed. We’re seeing indications this may be starting to change and we couldn’t be happier.
Q: What occupations do you need the most in the short term, and where do you look for talent?
HK: We need cheminformaticians, computational chemists, medicinal chemists and synthetic chemists. We use the talent placement agency Mitacs for grad students and post-docs. We also search LinkedIn for candidates with biopharmaceutical experience or PhDs in organic or synthetic chemistry. Life Sciences BC and the Pharmaceutical BioScience Society (PBSS) job board are great for finding people with very specific skills. But word of mouth and networking have the highest impact. We use our own professional network, our board, our observers and the local ecosystem. It’s easiest to hire where there’s already a critical mass of talent — Boston, the San Francisco Bay Area, Montreal and, of course, Vancouver! We prefer to hire from within Canada, but we’ll hire the right person from anywhere.
Q: Does the cost of U.S. talent affect your HR plans?
HK: It’s the cost of doing business. Uprooting someone from Boston to work in Vancouver is expensive and they won’t be productive for six months. I’d rather pay someone 30 percent more to stay where they are and contribute right away. This is possible because we’re primarily computational. Before the pandemic, our technical team wasn’t so keen on remote work. But we found we could often be more productive working remotely than having everyone on site. The pandemic solidified the idea that a decentralized team can work well.
Company profile: Variational AI
Since September 2019, Variational AI’s team of experienced AI/machine learning and business specialists has been collaborating with biopharmaceutical partners to apply generative AI to drug discovery and bring new therapeutics to market.