how AI, ML, & BIG DATA are disrupting drug discovery & development

…or so we try.

May 2022

by Marta Gaia Zanchi

This month I had the pleasure to converse on the topic at the 2022 LSX Healthtech Leaders conference, as a participant to a panel skillfully moderated by Steve Carney (Editor, Drug Discovery Today) and in the company of Martin-Immanuel Bittner (CEO, Arctoris), Mark Davies (SVP of Informatics and Data, Benevolent), and Neil Sahota (United Nations Artificial Intelligence Advisor and Interim CEO ACSILabs).

first, sincerest thanks

to everyone who helped me prepare for this panel: my colleagues Sebastian and Ferran; Matthew Li; our friends at Insight Partners, Kindred Capital, The Venture Collective, and other investors across two continents; and numerous KOLs at Bayer, Novartis, Johnson & Johnson. And of course, sincerest thanks to the 128 startups focused on drug discovery, plus the 63 startups focused on clinical trials, that considered us for their funding needs, in the process educating us on the challenges they see from the front lines of innovation.

we were not short of topics to discuss,

as we put our heads together in advance of the panel in order to zoom in on a few key objectives for the day. The panelists’ full list of interests and questions for debate included:

• The role in cost cutting of new drug spend and development times — how big of an impact can AI have in time and cost saving?
• Should biopharma companies be planning for a future in where AI is routinely used in discovery?
• AI collaboration to target, identify, and diagnose faster — how does this collaboration look like?
• Gaining insights from clinical trial data — should everyone be taking a data science approach to clinical trials, and how?
• What’s the role of data for AI-driven drug discovery? Is high quality data in drug discovery scarce? Are public datasets the best source for AI model training, and if not — is access to data still a bottleneck? And if so, how to address that?
• What are the latest trends in AI drug discovery? Some examples: increasing interest in supporting rapid iterations/ accelerated DMTA cycles in molecule design; synergy between AI and wet lab operations.
• Company culture: biology and engineering come with very different mindsets —can we all agree on the importance of building teams that effectively transcend disciplines?
• Health Tech or Biotech? Do pure drug discovery software platforms exist, or are they (should they) be the new generation of biopharma startups — developing their own assets to maximize Value and ROI? Is AI simply what the mouse model was for biopharma in the previous century?
• Working the development pipeline: What development stages of clinical trials are next for disruption? Trial design, recruitment, monitoring & retention, expanded access, …anything else? Is the old concept of clinical trials simply starting to show its inadequacy?
• What is the role of CROs in the revolution? How is their role changing?
• Innovation rather than automation: introducing AI capabilities into the drug discovery process also requires business process (and capability) changes — what are they? Rather than automate/replicate what we already have, we also need to explore way to find new, optimal ways that may not have existed before.
• “It takes a village:” For many industries and even agencies like the United Nations, they are developing ecosystem partnerships to accelerate their work and amplify the potential beneficial outcomes. Have AI biotech companies (especially startups) finally adopted an ecosystem model by leveraging CROs, academic researchers, pharma (co-development), etc? Or are we not there yet?
• Biocomputational VCs in Europe: how many and which VCs have the right expertise at the intersection of Tx and tech? What’s more important? (From a founders’ perspective!)
• Payer’s perspective: With the increasing number of new drugs entering the market, especially complex drugs like gene therapies and cell therapies, what is going to happen on the reimbursement side? Do small molecules have an advantage in this regard?
• Where in the drug discovery process can AI bring the most compelling cost savings? Lead discovery, optimization, …
• Do drug discovery companies have a relevant moat against new entrants? How important it is to have differential datasets vs the best in class algorithms vs … what else?

In the end, the panel converged on just a few topics that were top-of-mind priorities for speakers and attendees alike:

1. Will access to high quality data prove to be the rate limiting step in the process?
2. Innovate rather than automate? Do we have the opportunity to explore uncharted water rather than just to get to an expected solution more quickly?
3. Where do you see this most profitably placed currently and how do you envisage its positioning in the future to achieve most significantly its ability to disrupt the Pharmaceutical industry? How might this synergize or otherwise with current set ups in discovery for example?
4. How will current company culture adapt to ensure optimum positioning of AI within/across teams?
5. How might incorporation of AI affect future ability to diagnose and stratify patients and impact the clinical trial process?
6. Will the AI part of biotech in terms of platforms drive the development of new startups in the future? Will their future derive from the “tech” rather than the “bio”?
7. What is the role of CROs in an AI future?

to watch the full panel discussion,

find the video recording on YouTube at this link.

The conversation gave me a chance to share a few points that I am biased to believe deeply, and wanted to summarize here.

biology and engineering come with very different mindsets, almost as different as startups and corporations: innovation at the nexus of the two is only possible by fostering a culture of collaboration between members of multidisciplinary teams, and by creating strong partnership early.

The winners will be the startups that can partner with Pharma/Biotech early and collaboratively.

Startups with multidisciplinary teams who bridge domains of knowledge will be at an advantage. Not just startups, but Pharma/Biotech corporations as well will extract the most value from innovations in compute if they are able to foster a culture of collaboration (as opposed to, competition) between their in-vitro and in-vivo team. There will be room for both startups that sell their tech (pure-play software, generally packaged in SaaS business models), and startups that develop their assets*; either way, collaboration with Pharma/Biotech is key.

*this is a controversial topic, with most VCs believing that startups developing their own assets will be the ones to maximize Value + ROI for their investors (but will also require the most time and capital!).

Except for disease indications where there is very little data, there is diminished return for startups to invest in data, and best-in-class algorithms are no guarantee of success either.

What will separate winners from the rest will be the ability to realize maximum utility from the algorithms, by working in a close iterative loop with the teams who are focused one step downstream, in the process of testing the hypotheses generated by these novel computational approaches — whether they be new “hits” or leads ready for inclusion in a clinical trial funnel. This will require a “360” approach of early discovery through to translational analyses that can be reduced to practice in the clinical setting.

As far as trends, lots of cool companies are doing rapid DMTA using ML + automation supporting rapid iterations/ accelerated cycles in molecule design.

I hope we will see more synergy between AI and wet lab operations, more AI in cell therapy (manufacturing), developing on the cloud, and tech ops blurring lines in the CGT field. Also, CROs are getting wise! This will be an interesting space to watch.

If we don’t fix the clinical trial process, we will not be able to reap but a small fraction of the value generated by innovation applying AI to the process of drug discovery & development.

Most clinical trials fail due to flawed design coupled with poor recruitment, monitoring, and retention strategies. In a perfect world where all the innovation applied to the process of drug discovery is successful — in everything from designing chemical libraries to lead generation and screening — we are still going to see 90% of the trials these new leads enter fail, for reasons unrelated to their true potential for safety and efficacy in certain indications, and instead related to poor clinical trial design and execution. Somehow I feel the ratio of AI startups that we see focused on drug discovery versus clinical trials (2:1, as mentioned above) should be flipped.

Are you a preseed/seed stage startup focused on helping Pharma/Biotech with their drug development and clinical trials challenges? Apply for funding by visiting https://www.nina.capital. We look forward to hearing from you!

Marta-Gaia