Daphne Koller, founder and CEO of AI-driven drug discovery company insitro, announced in July 2026 that her company had welcomed an Israel-based team from CombinAbleAI as part of an acquisition aimed at launching insitro’s newly integrated TherML platform. The platform incorporates CombinAbleAI’s physics-informed optimization engine, which has been pre-trained on more than 100,000 molecular dynamics simulations to accurately predict protein structure and flexibility — a capability directly relevant to identifying and designing new drug candidates targeting specific proteins involved in disease.
The acquisition builds on insitro’s existing POSH platform, which the company validated in a peer-reviewed paper published in Nature Communications, demonstrating that self-supervised machine learning models trained on unbiased cellular morphology images — essentially, how cells physically look and behave under a microscope — can reconstruct meaningful information about gene function and causal biological relationships without being explicitly told in advance what patterns to look for. insitro has applied this general approach specifically to programs focused on metabolic disease and neuroscience, two therapeutic areas where identifying the true underlying biological drivers of disease has historically been a slow, expensive bottleneck in drug development.
A Public Voice for Realistic AI Expectations in Medicine
Despite leading a company built entirely around applying AI to one of medicine’s hardest problems, Koller has been notably public about tempering expectations for what AI can realistically achieve in drug discovery in the near term. In a widely discussed guest essay, Koller argued directly against treating AI as a “magic wand” capable of instantly solving drug discovery’s fundamental challenges, emphasizing instead that meaningful progress requires patient, rigorous application of machine learning to specific, well-defined biological questions rather than expecting sweeping, immediate breakthroughs. She has separately described the current moment in AI-powered biology as reaching “escape velocity” for identifying causal disease targets — a more measured claim about accelerating progress on a hard problem, rather than a claim that the problem has been solved.
Koller co-founded online education company Coursera in 2012 alongside fellow Stanford computer science professor Andrew Ng, following an academic career that included a MacArthur Fellowship, often referred to informally as a “genius grant,” awarded in recognition of her research contributions to probabilistic graphical models and machine learning. She founded insitro in 2018 specifically to apply modern machine learning techniques directly to drug discovery and development, positioning the company at the intersection of two fields — computational AI research and experimental biology — that have historically operated with largely separate methods, timelines, and cultures.
Building a Genuinely Interdisciplinary Company
insitro’s core operating model depends on integrating large-scale experimental biology — generating the kind of high-quality, systematically collected cellular and genetic data that machine learning models actually need to produce reliable results — directly alongside computational model development, rather than treating AI as a layer applied on top of data generated elsewhere by traditional pharmaceutical research methods.
Bridging Two Historically Separate Scientific Cultures
insitro’s operating model requires reconciling two fields that have traditionally operated on very different timelines and standards of evidence: computational machine learning research, which typically iterates quickly on digital data, and experimental biology, which depends on slower, more expensive laboratory work to generate the physical data those models are trained on and validated against. Koller has structured insitro specifically to run both functions inside a single organization rather than as separate academic and industry silos, a structural choice she has described as necessary for AI models in biology to be genuinely useful rather than producing plausible-looking predictions that fail to hold up under real experimental validation.
insitro has raised substantial venture funding since its 2018 founding from investors including Andreessen Horowitz and several major pharmaceutical companies pursuing direct partnerships to apply the platform against their own drug pipelines, reflecting a business model that blends traditional biotech venture funding with pharma-industry collaboration revenue rather than depending solely on eventual drug approvals for financial viability. Koller’s continued academic-style public writing about AI’s real limitations in biology, even while running a venture-backed company with strong commercial incentives to overstate its own platform’s capabilities, has been noted by several industry observers as an unusually candid posture for a well-funded startup CEO in a hype-prone sector.
What This Means for Philippine Founders
The Philippines’ biotech and pharmaceutical research capacity remains comparatively limited relative to more established regional hubs, but Koller’s own public emphasis on patient, rigorously validated AI applications rather than overhyped promises offers a useful model for Filipino biotech founders and researchers navigating a global funding environment where AI-for-biology claims are increasingly common and not always well substantiated. Philippine public health researchers focused on diseases with significant regional burden, including several tropical and infectious diseases with comparatively limited existing drug development investment from major global pharmaceutical companies, should also watch whether AI-driven drug discovery platforms like insitro’s eventually lower the cost floor enough to make previously uneconomical, regionally specific drug development programs newly viable.
Share this article