Welcome back to the second edition of Is it for Real? Today I want to continue our conversation about AI, this time in a space close to my heart: life sciences. Specifically, I will be highlighting a biotech startup that I believe has done the best job so far at integrating AI into its drug discovery pipeline.
So it AI in Life Science for real?
When people talk about “AI transforming drug discovery,” it often sounds like a pitch‑deck buzzword. So let’s start with a company that has been quietly building one of the most sophisticated AI‑native biotech platforms: Insitro.
At a high level, AI needs what biology has historically lacked: large, well‑labeled, and consistent datasets. Most experiments in life sciences are designed to answer a single narrow question about a single target. Insitro flips that script by designing experiments for machine learning from day one. In high‑throughput robotic labs, they run large‑scale tissue culture experiments and capture multi‑modal readouts. This data is then used for machine learning to understand trends.
But what does collecting such data really entail?
Insitro’s approach centers on four complementary streams, each paired with machine learning to extract insights:
Fixed Cell Imaging
Proprietary, large-scale spatial omics that combine basic cell morphology with protein and transcript readouts. Machine learning helps turn these complex images into interpretable features that support disease phenotyping and target discovery.Live Cell Imaging
By tracking living cells over time, Insitro adds a temporal dimension. Techniques like Quantitative Phase Contrast and their custom “insitroPaint” system allow them to follow disease progression and reversal. Machine learning layers on top, extracting new longitudinal phenotypes.Cellular Omics
Their omics platform has profiled more than 35 million human cells, creating a compendium of cell types, states, and disease effects. Combined with pooled CRISPR screens, machine learning can interrogate the impact of genetic variation with high accuracy.Activity Phenotyping
Using high-throughput electrophysiology and other functional assays, Insitro measures electrical activity from neuronal networks. These functional readouts give a direct view into neurological diseases like epilepsy and neurodegeneration, helping uncover novel therapeutic targets.
The idea is elegant. If you can model disease biology more accurately, you can pick better targets, design better drugs, and match them to the right patients. Insitro describes its platform as one that can de-risk the entire pharma value chain, from target discovery to clinical trial design.
But as much as I am a fan, many questions and doubts remain about the promises of this approach. Are Insitro’s models being built to answer specific biological questions, or are they primarily exploratory datasets without clear hypotheses? In vitro models of disease have well-known limitations, and simply generating more data does not automatically solve them. Another open question is whether Insitro is building proprietary machine learning algorithms tailored for biology, or if its platform ultimately relies on the same large language model advances that have dominated headlines in recent years.
The future is promising. While AI cannot yet tell us which experiments to run, platforms like Insitro are beginning to generate data at a scale and richness that were previously out of reach. The challenge now is not just collecting these data but proving that they can translate into real biological insights and, ultimately, new medicines.
Here is what I am into:
Who I am rooting for
The Premier League is back. While I can’t in good faith recommend rooting for my own team, Manchester United, I can recommend spending your weekend mornings tuned into the league. There’s a special kind of energy at this stage of the season when every club feels like a contender and every fan base dares to believe this could be their year. The football is fast, the narratives are fresh, and the optimism hasn’t yet been crushed by reality.
What taught me things I didn’t know
One of the Middle East’s most popular educational YouTube channels just released an episode on the genius of Friends. It dives into how the cast and characters were assembled, why the chemistry worked so well, and the cultural impact the show continues to have on young adults around the world. The episode is subtitled in English and well worth a watch.
What Stated the Obvious
The Atlantic recently published a piece titled “Actually, Slavery Was Very Bad” and it is a must-read. Clint Smith III, the author, writes with both clarity and urgency, making it feel less like a choice and more like a responsibility to confront this history directly.
What I’m Interested In
I’m fascinated by how new media is reshaping what we consume, with podcasts and YouTube channels rising in influence compared to traditional television. Few people can unpack this shift better than Pablo Torre, especially now as he builds on his new New York Times deal.
Thanks for joining me this week. I’d love to hear your thoughts. You can reach me at [email protected]
Yours truly,
Gad