Is AI Finally Reversing Eroom’s Law? What Insilico Medicine’s Latest Benchmark Means for Pharma R&D
?? ?? Thibault GEOUI ?? ??
Science CDO - Head of AI/ML for Drug R&D ??- Bridging Science ??, Data ??, and Technology (AI) ?? to Help Life Sciences Companies Bring Better Products ?? to Market Faster - Linkedin Pharma Top 1%
Since Insilico Medicine published their Press-Release "Insilico Medicine announces developmental candidate benchmarks and timelines for novel therapeutics discovered using generative AI" - Link here - I really wanted to write something about it, because I think (1) the results are impressive and (2) given the combination of hype and skepticism around the use of AI in Pharma R&D, their transparent communication and their use of both traditional (i.e.: scientific publications) and more modern (i.e.: social media) channels is the way to go!
So what do I think about insilico medicine's journey and result so far?
Pharma R&D has long been plagued by a paradox: despite scientific and technological advances, drug discovery has become slower, riskier, and more expensive. This is the essence of Eroom’s Law, a term coined by Jack Scannell to describe the decades-long trend of declining R&D productivity, in stark contrast to Moore’s Law in computing. While R&D spending has skyrocketed, the number of new drugs reaching the market per billion dollars invested has fallen dramatically (100x less today vs 1950s, adjusted by inflation)
The culprits? The need for drugs to outperform existing treatments, bloated R&D costs, increasing caution from regulators and a reliance on mechanism-driven approaches that often fail to translate into successful drugs.
But what if AI could finally turn the tide?
Insilico Medicine’s latest benchmark publication suggests we might be at an inflection point. Their AI-driven drug discovery engine has achieved something remarkable:
These numbers challenge long-standing skepticism about AI’s utility in drug discovery and could mark the beginning of a much-needed shift in Pharma R&D efficiency.
Breaking the Bottleneck: Why Insilico’s Results Matter
Pharma R&D operates under extreme inefficiencies:
?? Preclinical studies take too long. AI-driven approaches can cut preclinical timelines by half or more, as seen with Insilico’s QPCTL program, which reached Phase I trials in just 9 months.
?? Costs are unsustainable. The average cost of bringing a new drug to market is $2.6 billion (up to 6billion for the top 16 Pharma), much of it lost in failed programs. Insilico synthesized only 60–200 molecules per program (compared to thousands in traditional screening), reducing wasted effort and cost.
?? Failure rates are brutal. 90% of drugs fail in clinical trials, often due to unanticipated toxicity or poor efficacy. AI-driven workflows integrate predictive toxicology and molecular design earlier in the process, improving candidate quality.
AI Beyond the Hype: Real-World Validation
AI in Pharma has been met with skepticism. Many dismiss it as a glorified virtual screening tool, good for repurposing known drugs but incapable of true innovation.
Insilico’s results challenge this narrative. Their ISM001_055 (a TNIK inhibitor for IPF) progressed from AI-generated design to Phase IIa trials in under four years, demonstrating:
? De novo molecular design – not just repurposing existing drugs
? Novel mechanisms of action – expanding beyond well-trodden targets
? Clinical validation – proving AI-generated candidates can work in humans
By openly publishing methodologies in Nature Biotechnology, Insilico has provided transparent, peer-reviewed evidence—a crucial step in gaining industry trust.
Implications for Pharma’s Future
?? R&D Cost Restructuring
If AI-driven platforms can reliably reduce early-stage failure rates and timelines, companies can shift capital toward more exploratory research and parallel pipeline expansion, de-risking late-stage failures.
?? Hybrid R&D Models
Insilico’s partnership with Fosun Pharma is an example of how AI startups and established Pharma players can collaborate. Instead of replacing traditional Pharma infrastructure, AI augments and accelerates it.
?? Expanding AI’s Reach
Insilico is moving into chronic pain, obesity, and muscle-wasting diseases: therapeutic areas with high unmet need and complex biology. AI’s ability to model multifactorial diseases and design selective, tissue-targeted drugs could open entirely new frontiers.
?? Regulatory Challenges & Industry Adoption
Regulatory inertia remains a major roadblock. The FDA and EMA lack formal AI-drug approval frameworks, slowing adoption. However, Insilico’s pre-IND regulatory engagement sets a precedent for AI-driven drug development, paving the way for more AI-native drug programs.
A Tipping Point for AI in Pharma?
We’ve heard countless claims that AI will “revolutionize” drug discovery, yet Pharma R&D remains notoriously slow to change. Insilico Medicine’s latest benchmarks, however, provide compelling data-driven evidence that AI can compress timelines, reduce costs, and improve success rates, potentially reversing decades of declining productivity.
The big question now:
_______
Love ?? this article? Add your comments ??, re-share ? it with your network, follow me ?? for more posts like this, and DM me ?? if you want to deep dive on this topic!
Are there really any regulatory hurdles regarding this? Provided that most of it is applied at a pre-clinical stage, is there a true need for formal AI-drug approval frameworks? I mean, does it matter if a drug was designed via AI, provided that the requisites for Quality, Safety and Efficacy are met? (SMEs and biologicals alike) AI is already being incorporated in some tools to help design drug candidates or model their properties at a pre-clinical stage - the down side is wet biology will always have a heavier "weight of evidence" compared to computational predictions, so depends on how good are your models, and how are you applying them. But being able to filter out potentially bad DCs from the pipeline to begin with that's already a nice step forward to save resources, be it reagents, be it labor, be it time. This being said, some of these benchmarks are really impressive.
AI Advisor to Progressive C-Suite Leaders at Small Biotechs in Massachusetts | Specializes in AI Strategy, Tools Selection & Implementation
2 周Nice infographic! It's not a fix-all yet, but AI/ML helping to get 100% of candidates to IND is a stat you have to respect. Going in the right direction.