Sep 25, 2023 · 21m · a16z
Digital Biology with insitro's Daphne Koller
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
Daphne Koller, Founder and CEO of insitro, discusses with a16z's Vijay Pande how combining machine learning with human-derived biological data creates a new era of 'Digital Biology' that accelerates drug discovery and treats complex human diseases.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 2.1 Guest teaching 3.0 Guest disagreement 0.3 The host pushing back 0.0
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 21:00 Playful challenge on funding
In a very friendly interview, Daphne playfully turns the table on host Vijay by asking if he will fund her prospective climate project.
Hardest push from the host ▶ 8:08 Steering double clickThe interview lacks serious host pushback, but Vijay briefly interrupts to steer the conversation back to today's applications.
Biggest teaching moment ▶ 7:24 Correcting the data pipeline sequenceDaphne gently corrects the host's premise by explaining that AI is needed directly inside the experimental hardware before constructing latent spaces.
The host holds their own ▶ 9:41 Explaining legacy ML limitations in drug designVijay demonstrates strong industry knowledge by explaining why legacy ML required 100 active drugs and why foundation models represent a major breakthrough.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Why Life Sciences and Bridging AI with Biology | 1 | 2 | 0 | 0 | The host asks a standard open-ended interview question regarding why the guest chose life sciences. Daphne articulates her rationale around leverage and bridging ML with biomedical data without any conflict or pushback. | |
| Insitro's Data Factory and Generating Data on Spec | 2 | 3 | 0 | 0 | Vijay demonstrates baseline context by referencing Insitro's POSH paper. Daphne details the platform and explains generating data on spec using stem cells and pooled optical screening. | |
| Creating Latent Spaces for Biological Modalities | 3 | 4 | 1 | 0 | The host introduces the concept of latent spaces in biology. Daphne gently re-frames the topic by stepping back to clarify that AI is required at the collection and instrument level before latent spaces can even be constructed. | |
| Systematic Recipes and Long-Term Vision for Therapeutics | 4 | 3 | 0 | 0 | Vijay shares an insightful observation about how foundation models solve the cold-start problem of needing 100 active compounds in legacy ML. Daphne agrees and details Insitro's systematic recipe for human-derived therapeutics. | |
| Building a Cross-Functional Culture at Insitro | 1 | 2 | 0 | 0 | Vijay asks a straightforward organizational question about bridging cultures between ML scientists and biologists. Daphne outlines her strategy of hiring cross-disciplinary translators and enforcing collaborative company values. | |
| Respecting Atoms: Robotics and Physical World AI | 2 | 3 | 0 | 0 | Vijay asks about transitioning AI from digital bits to physical atoms. Daphne explains the subtle physical variability in biological experiments, such as individual technician habits, justifying Insitro's heavy use of robotics. | |
| The Convergence of AI and Biology: The Digital Biology Era | 2 | 4 | 1 | 0 | Daphne synthesizes historical scientific eras leading to digital biology. She playfully puts Vijay on the spot by asking if he will fund her potential climate biology venture. |
Statements from this episode (15)
Koller: Biological datasets became large enough for meaningful ML around 2016
“What brought me back to this field back in 2016 post Coursera was the realization that we can now finally, for the first time, measure biology at scale, both at the cellular level, sometimes at subcellular level, and at the organism level via ways of quantitat…”
Koller: ML lagged in life sciences due to lack of cross-disciplinary talent
“It wasn't having much of an impact in the life sciences, and I believe one of the main reasons for that is because there's so very few people who actually have the language of both disciplines and are able to bring them together.”
Koller: insitro operates a unique data factory generating biological data on spec
“So one of the things that we have at Insitro that is truly unique is we have a data factory. We have put together the tools that have been developed by people who are taking pluripotent stem cells, which are cells from you or me or anyone in this audience and …”
Koller: Arrayed cell screening creates environmental noise across separate wells
“One of the really challenging things about cells is because they're live, if you put different cells in different wells, then they each have a slightly different environment, and you get subtle differences, and it's really hard to reconcile.”
Koller: Insitro runs genome-wide CRISPR screens in two weeks
“When they're all in a pool, you eliminate all of those artifacts, and all of a sudden you have the ability to measure a genome-wide CRISPR screen, basically, so, 20,000 genes in the genome, all modifying the same cellular background in the same dish with a dif…”
Koller: Every part of insitro's technology stack is intrinsically AI-enabled
“It's impossible to run this instrument without AI being built into it, because you can't even segment the cells. You can't call the barcodes. I mean, all of it is an AI-enabled architecture. Every part of our technology stack is intrinsically AI-enabled.”
Koller: Human radiologists see only a small percentage of MRI data
“In MRI data, your radiologist doesn't see more than like a small percentage of what's there in your radiology images”
Koller: Insitro will deliver AI-discovered medicines to patients by 2030
“The hope is that by the end of the Of this decade, we will have built this process, we will have run through it a number of times, we will have delivered some medicines to patients in our first tranche of indications, but then we will have learned enough from …”
Koller: insitro conducts all discovery work in human-derived systems
“And this notion of, you know, we can cure lots of mice is, is something that really drove our discovery strategy at in situ, which is all of our work is done in human and human derived systems.”
Koller: Machine learning and life scientists speak mutually incomprehensible languages
“You take your average, you know, machine learning scientists and your average life scientists, even if they're very well intentioned, you put them into the room together, they might as well be talking kind Swahili to each other.”
Koller: Human technician variance is a primary signal in biological ML
“When you do biological experiments, one of the strongest signals when you apply machine learning to it is what was the technician who actually did the experiments? You could read that very clearly off the cells because they behave a little bit differently.”
Koller: Insitro builds robots to eliminate human technician variability
“Which is one of the reasons why we spend so much of our time building robots, because they do the same thing over and over again.”
Koller: AI's next impact frontier is interacting with the physical world
“The next frontier of what a, of the impact that AI can have is when AI starts to touch the physical world.”
Daphne Koller: AI and quantitative biology are merging into 'digital biology'
“I think this time that we're living is the time when those last two disciplines are actually going to merge, and they're giving us an era of what I think of as digital biology, which is the ability to measure biology at unprecedented stability and scale, inter…”