Ep172: Mostafa Ronaghi on Studying Live Cells at Scale

Episode Summary

Executive Summary: Mustafa Ranagi traces his path from wartime Iran to Sweden, Stanford, Illumina, and now Selenome, arguing that genomics alone is insufficient to explain disease. He says cell-level, multimodal measurement of live cells at scale is the missing link for building better AI models, understanding mechanism of action, and ultimately enabling a virtual cell and more predictive drug discovery.

Main Topics: Childhood in wartime Iran and early scientific formation (Priority: 5/5): Ranagi describes growing up in Tehran during the Iran-Iraq war, moving to a quieter village for safety, and developing strong math/science interests through school lab work and hospital exposure. Discovery of pyrosequencing and entrepreneurial mindset (Priority: 5/5): In Sweden, during graduate work, he combined enzymatic chemistry with DNA sequencing to invent pyrosequencing, first observing the signal in a restroom during an improvised experiment; he also emphasizes his long-standing desire to found companies. From academic innovation to Illumina’s rise (Priority: 5/5): Ranagi explains how his sequencing-on-chip work and startup efforts led to multiple acquisitions by Illumina, where he helped drive the company’s sequencing roadmap from $200,000 genomes toward sub-$1,000 sequencing. Grail, liquid biopsy, and the shift from genome to phenotype (Priority: 4/5): He recounts work at Illumina on prenatal screening and cancer detection from cell-free DNA, including discovering cancer signals in pregnant women, and notes that these efforts highlighted the need for better links between genotype and phenotype. Why Selenome focuses on live-cell, multimodal phenotyping (Priority: 5/5): Selenome is designed to measure the same single live cell across morphology, surface proteins, secretions, interactions, and transcriptome-related outputs, because those combined signals are needed to interpret genetic variation and disease. AI, foundation models, and the ‘virtual cell’ vision (Priority: 5/5): Ranagi argues that effective AI for biology requires high-quality multimodal data from the same cell, and he believes Selenome can become a category-defining data engine for foundation models and eventually a virtual cell.

Key Arguments: Genomics is powerful but not sufficiently actionable on its own; most variants still lack clear clinical meaning. Cell biology needs multimodal, same-cell measurement to connect genotype to phenotype with less noise. Single-cell data are only useful at scale when enough cells are measured in context, not as isolated snapshots. Existing transcriptome/proteome tools are informative but incomplete because low-abundance regulatory signals are hard to detect. A foundation model for biology will require integrated measurements from cells, perturbations, and downstream behaviors. The best biological analog for Selenome is Waymo rather than Tesla: fewer but richer sensor modalities can outperform huge volumes of noisy data. Drug discovery and mechanism-of-action studies will improve if platforms can test perturbations and read out cellular responses in one system.

Data Points: Years of war exposure: 8 years - Ranagi says he experienced bombing throughout the Iran-Iraq war after the revolution. Age at revolution: 10 years old - He says the revolution occurred when he was 10 and the war began a year later. Family size: 5 kids - He describes his family as lower middle class with five children. Village population: 2,000 people - He moved from Tehran to a mountainous village for safety during the war. Tehran population: 10 million - He contrasts city life in Tehran with village life. Practical school schedule: 6-day school week - He notes Iranian schools ran six days a week, with one day for practical work. Swedish exam timeline: 6 months - He learned Swedish and took the university-entry exam after about six months. Graduate sequencing throughput: 10 samples per run / 32 samples per run - He compares early Pharmacia and ABI sequencers during graduate school. Pyrosequencing signal development time: About 6 months - He says it took six months of optimization to reach single-base resolution. Sequencing-roadmap cost goal: $200,000 to $1,000 - Illumina’s 10-year program targeted a major cost reduction in sequencing. Illumina initial genome cost: $200,000 per genome - He cites this as the cost level when he joined Illumina. Illumina run output: 5G per run - He says Illumina could do 5G per run in 2008. Sequencing price point achieved: Sub-thousand-dollar sequencing - He says Illumina demonstrated this within three years and reached the market after five years. Cancer sensitivity improvement: 2 orders of magnitude - He says Illumina’s cancer-detection work increased sensitivity enough to detect stage 1 disease. Company funding: $150 million Series B - Selenome raised this amount in January 2024. Selenome staffing/engagement: More than 200 groups - He says he has interacted with over 200 interested groups. Platform data output: Petabyte level per run - He describes raw data generation from Selenome experiments. Compressed data output: A few hundred GB - He says on-device GPU processing reduces data before cloud upload. Cell throughput: Tens of thousands to hundreds of thousands of cells per run - He describes platform scale for multimodal studies. Transcriptome detection floor: Around 200 transcripts - He argues that current transcriptome sequencing struggles below this level. Selenome company count: 9 companies - He says Selenome is the ninth company he has started. Device count: 18 different devices - He says Selenome is the 18th device launch in his career. Virtual cell timeline: 5 to 10 years - He estimates a virtual cell model could be feasible in this window.

Pivotal Quotes: "the new marriage for me is basically cell biology and I kind of felt that, you know, oh, the cells actually, they talk to you." — Mustafa Ranagi: He explains why he moved from genomics into live-cell biology and phenotyping. "if you would actually have multimodal measurements on the same cell" — Mustafa Ranagi: He describes the core technical requirement behind Selenome’s platform. "I think that this is going to be a category defining technology for the field." — Mustafa Ranagi: He predicts Selenome will become a foundational instrument class for cell biology, akin to NGS for sequencing.

Implications: The conversation suggests the next leap in biotech will come from measuring live cells more completely, not just sequencing DNA. If Selenome succeeds, it could reshape drug discovery, mechanism-of-action studies, and AI biology with richer, less noisy data.