May 19, 2023 · 46m · no-priors
No Priors Ep. 6 | With Daphne Koller from Insitro
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of No Priors, Insitro founder and CEO Daphne Koller joins Sarah Guo and Elad Gil to discuss how machine learning and high-throughput cellular biology are transforming drug discovery, navigating cross-disciplinary team cultures, and building mission-driven platforms to improve human health.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 16% of the talking time here. How this is scored →
The hosts as informed peer 4.7 Guest teaching 5.4 Guest disagreement 0.7 The hosts pushing back 0.2
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 39:40 Chiding tech arrogance in life sciences
Daphne directly cautions tech professionals who enter biology assuming machine learning will easily solve everything without respecting biological complexity.
Hardest push from the hosts ▶ 30:10 Challenging development timeline limits via COVID accelerationElad challenges standard pharma timelines by citing rapid COVID drug approvals and questioning whether timeline bottlenecks are mostly regulatory inertia.
Biggest teaching moment ▶ 18:42 Translating iPSC biology into A/B testing terminologyDaphne gives a clear, foundational breakdown of reprogramming adult cells into pluripotent stem cells and framing variant evaluation as biological A/B testing.
The host holds their own ▶ 30:10 Elad citing HRD pathways and regulatory safetyismElad demonstrates substantial biotech knowledge by discussing HRD synthetic lethality mutations and referencing Paul Janssen's analysis of clinical risk frameworks.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Bridging Computer Science and Biology at Stanford | 3 | 4 | 0 | 0 | Sarah introduces Daphne's background and asks about her foundational textbook on probabilistic graphical models. Daphne explains the historical pendulum shift from graphical models to deep learning and the emerging return to causal interpretability. | |
| Career Trajectory: Coursera, Calico, and Founding Insitro | 3 | 4 | 1 | 0 | Elad asks about Daphne's pivot from Stanford to Coursera, Calico, and founding Insitro. Daphne recounts resigning her endowed chair at Stanford and why Calico's single-biology focus prompted her to build a platform company. | |
| Insights from Calico and Limitations of Aging Research | 4 | 5 | 1 | 0 | Elad inquires into specific takeaways from Calico that influenced Insitro. Daphne highlights the severe data bottlenecks in human longitudinal aging research and the opportunity to apply ML to richer biological datasets. | |
| Tackling Drug Failure Rates via Target Identification | 6 | 6 | 1 | 0 | Elad outlines the entire pharma development pipeline and cost structure. Daphne sharpens the focus, explaining that 95% of drug programs fail primarily because of picking the wrong target or patient population rather than molecular chemistry or trial execution. | |
| Insitro's Dual Strategy: Human Data and High-Content Wet Lab Screening | 4 | 6 | 0 | 0 | Sarah asks how target identification can be trained without end-stage clinical trial labels. Daphne details Insitro's dual strategy combining human genetic 'experiments of nature' with high-content wet-lab cellular perturbation data. | |
| Therapeutic Focus Areas and iPSC Cellular Modeling | 3 | 7 | 0 | 0 | Sarah asks Daphne to explain how lab neurons are created for a non-biology audience. Daphne walks through the step-by-step process of reprogramming adult cells into iPSCs and running in vitro A/B tests on disease mutations. | |
| Navigating Biological Complexity: Single Cells, Hepatocytes, and Organoids | 5 | 5 | 0 | 0 | Sarah asks how to capture multicellular complexity beyond single cells, including organoid models. Daphne explains using hepatocytes under stress and pragmatically prioritizing diseases that manifest in single-cell lineages. | |
| Business Model Strategy: Platform Engines, Asset Owners, and Partnerships | 5 | 5 | 0 | 0 | Elad asks about Insitro's commercial strategy regarding in-house development versus partnering with pharma giants like BMS. Daphne explains how platform engine companies can avoid the empty-cupboard risk and partner on existing assets. | |
| The Power of Machine Learning in Clinical Biomarkers | 6 | 6 | 0 | 0 | Elad asks about ML opportunities in clinical biomarker development. Daphne emphasizes that biomarker-guided trials double success rates and illustrates this with the Herceptin precision oncology case study. | |
| Accelerating Drug Timelines, Regulatory Realities, and Biology's Clock | 7 | 6 | 2 | 2 | Elad cites HRD drugs, regulatory safetyism, and rapid COVID trials to question whether drug timelines are purely regulatory constraints. Daphne pushes back, clarifying that acute viral infections differ from slow chronic diseases like Alzheimer's where biology dictates time. | |
| Cultural and Mindset Clashes Between Engineers and Biologists | 5 | 6 | 1 | 0 | Elad and Daphne discuss cultural differences between deterministic engineering and messy biological systems. Daphne shares an anecdote about technician onion breath affecting cell survival and contrasts pattern-seeking engineers with anomaly-focused scientists. | |
| Operationalizing Cross-Disciplinary Collaboration and Insitro's Core Values | 6 | 6 | 2 | 1 | Elad shares how Color operationalized cross-functional scrums between software engineers and genetic variant scientists. Daphne explains where agile applies versus where cell differentiation takes fixed time, while warning against tech arrogance. | |
| High-Impact Frontiers: Digital Bio, Climate, Energy, and Education | 4 | 4 | 1 | 0 | Sarah asks what high-impact areas founders should tackle and how to handle tough business sectors. Daphne outlines frontiers across climate, agriculture, and edtech while encouraging mission-driven problem selection. |
Statements from this episode (20)
Koller: AI requires synthesizing deep learning with causal and interpretable models
“What I think we're starting to see right now is a the pendulum starting to swing back in the sense that there is a greater understanding that you really need a bit of both. You need that hugely powerful pattern recognition that we get from deep learning, but y…”
Koller: Longitudinal Human Aging Data Is Bottlenecked by Decades-Long Timelines
“Data in aging and specifically human aging is really hard to get because human aging is a very long process. And in order to get data on the longitudinal trajectory of human aging today, one needed to start collecting data, you know, 2030 years ago, and the co…”
Koller: ML in biopharma is like general computing, transformative everywhere
“My analogy is that it's not like x-ray crystallography. It's like computers. You're going to use it everywhere, and it's going to be transformative everywhere.”
Koller: 95% of all drug discovery programs fail
“If you look at the actual numbers of what makes drug discovery so expensive, it is the fact that 95% of drug programs fail. They just do not succeed.”
Koller: Most drug failures stem from wrong target selection, not trial design
“It's a place where most programs fail is because we're just not modulating the right thing. It's the wrong target in the wrong indication or the wrong patient population.”
Koller: Drug discovery lacks clinical outcome ML training data until trials
“Because when you think about it, it's the one area where you really don't have the right type of training data, at least not obviously, because the question you're asking yourself is, if I make this therapeutic intervention in this patient, what is it going to…”
Koller: Insitro aims to replace untranslatable animal models using human cellular models
“Which ultimately what we're looking to do is to replace the sort of untranslatable animal models with something that is much more driven from human biology.”
Koller: Early Alzheimer's drug approvals will increase available brain MRI data
“I think there will be even more now with the approval of some of the earliest Alzheimer's drugs, because it's going to be part of the process by which people are either selected to receive the drug or not, depending on whether their brain MRI shows certain cer…”
Koller: Insitro expands into metabolism and oncology due to standard-of-care data
“The other areas that we've gone into are metabolism and oncology, because again, those are areas where relevant, disease relevant data, That is high content, that is unbiased and truly informative about the disease state is collected quite abundantly as part o…”
Koller: Insitro prioritizes single-cell diseases before complex organoid models
“There's plenty of things that we can do today where the where the disease does manifest sufficiently in a single cell lineage, and so we go after those first, and we defer some of the other ones to a later stage, because technologies such as organoids, for exa…”
Koller: In three years new technologies may unlock more disease models
“Maybe in three years we'll have another tranche of diseases that are unlocked by the technological tidal wave that we're all riding.”
Koller: Partnering on existing drug assets can shave two to five years
“And that can usually say shave off, you know, two, three, maybe even five years from the development of a program because you've already made the drug. Sometimes you've already put it in people. You've shown that it's safe. You have a good biomarker for when i…”
Koller: Drugs with biomarkers are twice as likely to succeed clinically
“There's you know, there's research that shows that drugs that have a biomarker are about twice as likely to be successful in the clinic as ones that do not.”
Koller: Drugs backed by human genetics are twice as likely to succeed
“There's also data that show that drugs that have support in human genetics are twice as likely to succeed as ones that do not”
Koller: Broad patient targeting to maximize revenue causes drug trial failures
“That is one of the big things that causes drugs to fail is that you are trying to apply it more broadly. If I'm being cynical, sometimes it's supposed to maximize the revenues that you can get from a drug versus trying to figure out exactly in which patients i…”
Koller: Research-grade biomarkers frequently fail in standard clinical care settings
“You can have the most beautiful, sophisticated biomarker that works in a very carefully designed research environment, and it's not going to work in the wild as part of the standard of care.”
Koller: Computational methods cannot compress slow biological disease progression timelines
“And so ultimately there's only so much that you can speed up biology in certain cases because biology takes as long as it takes.”
Koller: Engineers look for patterns; scientists look for outliers
“When you show an engineer or computer scientist a bunch of dots
Usually the natural inclination is to try and find the pattern, the thing that explains as many of the points as you can, because that is the thing around which you will engineer your system.
If y…”
Koller: Insitro uses agile for platform biology, but not drug discovery projects
“We have product managers who do scrums and do you know, these agile planning processes, and we apply that also to our platform development, even on the biology side. But at the same time, you know, drug discovery projects, which are years long, you don't do sc…”
Koller: Tech workers entering biotech often disrespect biological challenges and create friction
“There's a lot of tech people who come in To life sciences, and it's like, we have that cell verbal. We are the smartest. We're machine learning. We're going to solve everything. And they don't respect the challenges of the other discipline. They sometimes don'…”