Recapping Insights from a recent webinar by Endpoints News "AI faces its moment of truth"

Today I would like to briefly recap insights from a recent webinar by Endpoints News “AI Faces Its Moment of Truth”. 

This webinar brought together three top experts in the field -- Dr. Andrew Hopkins, CEO, Exscientia; Dr. Daphne Koller, CEO & Founder, Insitro; and Sanjiv Patel, President & CEO, Relay Therapeutics. They discussed the current state of AI-driven drug discovery, challenges and opportunities over the next 5-10 years, and they tried to broadly define what “AI-first” research really means, versus what it is pictured by hype media.

Key takeaways: 

• AI is useful for all three key stages of drug discovery: 

1. identifying “the right” biology to go after -- the right target in the right indication, for what therapeutic hypothesis, etc.
2. Having the right target, AI can help identify the right molecule for the right target patient profile 
3. Taking that molecule to clinical development, drawing on the pieces of information from the first two steps. Design clinical trials in a way that aligns with your patient population. There is a lot of room for AI/machine learning in patients' biomarker development -- both for deciding what to include in the trials, as well as for deciding on the right endpoints for efficacy and toxicity. 

• The above steps are interconnected and AI helps use insights generated at each step to improve each other step. For example, building advanced cell models using AI, which is helpful for target discovery, appears also useful for lead optimization towards a particular cellular phenotype. Or, such models are further useful -- for discovering biomarkers for clinical development, therefore, for better patient selection, etc. 
• Besides those three areas, there are lots of other applications for AI in the pharmaceutical value chain -- manufacturing (pharma 4.0), marketing, etc. (Check out the analytical report “The Landscape of Artificial Intelligence (AI) In Pharmaceutical R&D” to get a focused picture of what the current market of AI-driven drug discovery companies is).
• The major role of AI in drug discovery is not so much about the efficiency of drug discovery (speed/cost), but it is about being able to do things with AI which are not possible to do without AI. That’s where real impact on society comes into play. And efficiency is more like a desirable byproduct. For example, shaving off 6 months of drug development time is not really a major thing for the industry, because it is not 6 extra months that drives the cost up. It is the extremely low success rate of drug candidates in clinical trials which is crucial for cost and business overall. If AI can let us understand biology to the extent that we are able to increase success rates of drug discovery in clinical trials -- that would be a real industry transformation. 
• Some of the key advances over the last 5-10 years in AI-driven drug discovery come from progress in both algorithms and novel data. Things like knowledge graphs, deep learning, and generative AI are some of the key creative examples in the space. 
• Overall, the role of AI in drug discovery can be represented by a spectrum of use cases distributed between two grand goals: on the one hand, to make existing drug discovery processes faster and cheaper (efficiency problem), and on the other hand -- solve hard problems which were unsolvable without AI (innovations problem). The whole spectrum of pharma AI companies, in terms of their value proposition, falls somewhere between those two goals (or both). 
• While there are all sorts of business models in the AI for drug discovery space, including a variety of computation-only companies, the future, undoubtedly, belongs to those companies, that combine cutting-edge AI research with cutting-edge experimental tools/capabilities, biotech companies built as “AI-first” organizations, such as Insitro, Exscientia, and Relay Therapeutics. (read also “A New Breed of Biotechs is Taking a Lead”)
• Computational tools, such as AI, augment, not replace humans. Human-in-the-loop remains an essential concept for modern-day AI-driven drug discovery, i.e. the best effect can be achieved when drug hunting skills of human professionals are matched with cutting-edge AI tech. 
• We need to be thoughtful of what the role of AI is today, to avoid the inflation of the promise of what it can really do. “End-to-end” AI drug discovery -- a term often seen in media reports or company claims -- does not mean that there is a button on a computer and if you press the button, a drug is going to be packaged, ready for use, in let’s say three years from now. It is not how it works, and it will not be in the foreseeable future. AI is a tool, used by smart people, to substantially improve the research process, redefine it, and make it more integrated between stages; it is a great partnership between machines and humans. But we need to calibrate our promises and what language journalists use, and call things what they are, to avoid overheating the industry and another “AI winter”.  Let’s be clear, the impact of AI is transformative, but it is just not supposed to substitute humans any time soon and the R&D will remain a highly collaborative process with lots of efficiency and insights derived from AI tools and platforms, and creatively processed by human experts. 
• AI-driven transformation of pharma is not about adding AI tools to existing pharma research processes. But it is about transforming those processes altogether, building a new human-in-the-loop AI-centric process. 
• It is not really about what is “the first AI-discovered drug” and who did it first, it is about how to apply AI to ALL drug discovery to make the whole industry more efficient. 

What to expect over the next 5-10 years for AI in drug discovery? 

• Image recognition was not created originally for drug discovery, but this idea was then applied in pharma successfully. Similarly, many other AI-powered tech ideas from other industries will be applied in the pharma space. We will see what is working and what is not. 
• Data is king, we will have lots of new processes and technologies to get more quality data in higher volumes, this will improve AI performance as well. 
• Cultural shift. AI will not replace humans in pharma R&D, but humans will need to become more flexible in being able to combine AI and experimental tech. 
• Talent shortage for “double profession” specialists, skilled both in AI and biology/chemistry, maybe a challenge in the coming years.
• AI-driven methods can help redefine existing disease taxonomies into new taxonomies, more relevant to the underlying biology so that fewer data can be used to get meaningful results. 
• AI will potentially have the greatest positive impact on those therapeutic areas, which are more difficult to be modeled by traditional tools, for example, complex polygenic diseases like Alzheimer’s, Multiple Sclerosis, Bipolar Disorder, Depression, and other neural disorders. Mise models for neural disorders are almost ineffective since mise just doesn't develop those disorders. So understanding those diseases requires large-scale modeling, for example, using genetic data, and this is where AI modeling can be most useful. 
• A company of the future is an AI-first company with an inherently AI-centric and patient-centric research model, which combines all the processes, from target discovery to clinical development, in one end-to-end research process. 
• Achieving truly personalized medicine is only possible by also changing the underlying economic structure of the pharmaceutical business. 

BONUS PERK

Intelligent Health UK 2022, 06-07 April 2022, London, UK (promo code below)

In partnership with The NHS AI Lab, NIHR, The Health Innovation Network, AHSN Network and DigitalHealth.London, Intelligent Health UK will once again bring together the UK health community to focus on what matters the most, to break down the barriers between tech and healthcare to #SaveLivesWithAI.

As the official media partner for the event, BiopharmaTrend is offering you up to 20% discount on tickets -- use promo code BPT20 to claim the discount when registering for the event. 

---

Welcome to my newsletter "Where Technology Meets Biology". Here I am sharing notable news, trend observations, biotech startup picks, industry analyses, and interviews with pharma KOLs. Feel free to reach out to info@biopharmatrend.com for consulting or sponsorship opportunities. Check www.BiopharmaTrend.com or learn more about myself on www.andriibuvailo.com.

Enjoying the newsletter? Help make it better by sharing it with your colleagues and friends. Please, like/share/comment to help increase organic reach, so that more people may read it.

-- Andrii