Jan 2, 2019 · 30m · a16z
a16z Podcast | On the Genomics of Disease, From Science to Business
Gabriel Ott · 9m spoken Sonal Chokshi · 6m spoken Vijay Pande · 6m spoken Malinka Walaliyadde · 5m spoken
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 a16z podcast episode, Sonal Chokshi, Vijay Pandey, Gabriel Ott, and Malinka Walaliyadde discuss how modern machine learning and computational software are transforming genomics and medicine. They explore how AI enables early, non-invasive disease detection, shifts biological research from single-gene focus to high-dimensional systems biology, and navigates complex market dynamics and reimbursement structures.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 23.6% of the talking time here. How this is scored →
The host as informed peer 5.1 Guest teaching 4.3 Guest disagreement 1.5 The host pushing back 2.5
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 14:36 Gabriel dismisses consumer empowerment from risk prediction
Gabriel forcefully rejects the premise that predictive risk scores help consumers, arguing that learning one has a 30 percent lifetime risk of cancer is unempowering.
Hardest push from the host ▶ 14:50 Host defends consumer empowerment of risk testingSonal directly refuses Gabriel's framing, invoking Angelina Jolie's op-ed to argue that risk knowledge provides vital proactive agency for patients.
Biggest teaching moment ▶ 3:48 Correcting host premise on Moore's Law in genomicsWhen Sonal suggests that genomics lacks a Moore's Law equivalent, Vijay and Gabriel immediately correct her premise by explaining that genomic sequencing cost drops drastically outpace Moore's Law.
The host holds their own ▶ 24:05 Host outlines private insurance market failuresSonal demonstrates strong industry expertise by explaining how patient plan switching and high deductibles create a structural misalignment against paying for early diagnostics.
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 |
|---|---|---|---|---|---|---|
| From Single-Gene Focus to Systems Biology | 4 | 4 | 1 | 1 | Sonal sets the stage by asking why ML matters now and offers an intuitive alphabet and word combination analogy for multi-variable systems biology. Gabriel validates her framing while explaining how computational biology moves beyond single-gene focus like Huntington's disease. | |
| The Genomics Cost Curve and Hardware-Software Layers | 5 | 5 | 2 | 2 | Sonal asks about cost curve drivers, mistakenly assuming Moore's Law does not apply to genomics. Vijay and Gabriel correct her, pointing out that genomic sequencing costs beat Moore's Law by several orders of magnitude before Sonal synthesizes the hardware and software layer dynamics using a semiconductor comparison. | |
| Unlocking the Whole Genome via Blood Biopsies | 4 | 4 | 1 | 1 | Malinka and Gabriel explain how machine learning allows diagnostic tools to analyze the 99.99 percent of non-mutated dark genome across the whole blood sample. Sonal adds lighthearted humor by calling famous mutated genes the villains of genomics and summarizes how this reinforces the systemic approach. | |
| The Survival Impact of Early Cancer Detection | 3 | 6 | 1 | 1 | Sonal asks why cancer remains difficult to diagnose early. Gabriel educates the host by contrasting the low survival rates of late-stage therapies with the 80 to 97 percent survival rates enabled by early detection. | |
| Shifting From Symptomatic Medicine to Detection | 6 | 4 | 3 | 7 | Gabriel argues that predictive risk testing is unempowering to consumers who receive percentage odds. Sonal directly pushes back on Gabriel's stance, citing Angelina Jolie's famous decision and arguing that risk knowledge offers actionable personal empowerment. | |
| Illumina's Market Position and Platform Dynamics | 6 | 3 | 2 | 5 | Sonal challenges the assumption that market monopolies like Illumina are problematic if consumers benefit, sparking a discussion on platform risk versus software porting. Sonal then demonstrates market understanding by comparing Illumina's move into application software to hardware transitions in tech. | |
| Reimbursement Challenges in Genomic Diagnostics | 7 | 4 | 1 | 2 | Malinka explains how US insurance reimbursement hurdles stifle commercial adoption because payers demand short ROI windows. Sonal demonstrates deep healthcare market expertise by explaining how high-deductible plans and patient churn structurally disincentivize long-term coverage. | |
| Expanding Horizons: Proteomics, Agriculture, and 3D Genomics | 6 | 4 | 1 | 1 | Sonal proactively drives the discussion into proteomics and mass spectrometry while defining technical terms for the audience. She synthesizes Gabriel's explanation of 3D spatial genomic organization into a broader computational system metaphor. |
Statements from this episode (21)
Ott: Machine learning maturity is accelerating genomics research
“What's different now as opposed to what was happening in 2000 is finally the technology, the machine learning techniques, as well as the hardware supporting that has matured to a point where we don't have to try to manually figure this complicated system out b…”
Pandey: Sequencing cost drop from $3B enables commercial clinical genomics
“The other key part of what's now is that human genome project cost three billion dollars, and you know, that's a pretty hefty copay for someone to have to shell out. It's only now that genomics is starting to get actually to the point where you could imagine p…”
Ott: DNA sequencing cost reductions have outpaced Moore's Law by magnitudes
“And it's beat Moore's law by several magnitudes.”
Walaliyadde: One dominant company drove a decade of sequencing cost drops
“The dominant player in the sequencing machine space has almost entirely by itself been driving down the cost of sequencing over the last decade or so.”
Walaliyadde: Genomics splits into sequencing hardware like Intel and clinical applications
“And an easy way to think about genomics is there's a sequencing layer, which is the companies that make the sequencing machines like Illumina, which is like sort of Intel and chips. And then there's the application layer, which are companies that are using thi…”
Ott: Computational biology tools reduced DNA alignment time from days to minutes
“That step is called DNA alignment, and that used to take days. Now there are tools out there that can do the alignment process in five minutes.”
Ott: Well-studied cancer driver genes represent less than 1% of the genome
“The genes that I mentioned, we're really talking about less than one percent of the entire genome. And so there's this 99.99% of the genome that hasn't really been understood in terms of how they affect people's bodies and how these diseases come out from othe…”
Ott: Single-gene targeted approaches fail for early cancer detection
“What sometimes happens when you are very focused though, is you missed a bigger picture. And what we're realizing is with respect to things like detecting cancer early, that focused approach is not working.”
Pandey: Blood contains vast tumor genomic data for new liquid biopsy diagnostics
“Really, the other key advance here in the cancer space and genomic space is the fact that blood has so much genomic information, even from things like tumors, and that opens the door for these new technologies to come in.”
Pandey: Unlike traditional blood tests, machine learning diagnostics improve with data
“There's a learning aspect here of machine learning, which is intriguing, that as you get more data, you get better, and that's something that's really not like any other test, where, you know, a lipid blood test doesn't get better as you have more patients the…”
Ott: Early detection increases cancer survival rates to 97% versus late-stage treatments
“The best drugs that we have today to treat cancer called immunotherapies only give you about 30 to 40% chance of five-year survival in treatment. These are the best drugs that we have. Chemotherapy and radiation give you less than 20%. Chance of survival. On t…”
Ott: No single silver-bullet drug will ever cure all cancers
“With that in mind, I think a lot of us like to think that it's going to be the next greatest drug that's going to cure cancer. It's going to be that, you know, silver bullet. There isn't really going to be a silver bullet in my mind to treating a hundred diffe…”
Ott: Science has not solved which biological signatures best detect disease early
“Yes, there are signatures early on that indicate whether there is a tumor or not. But what hasn't been solved yet is exactly what are those signatures that allow us to detect those diseases? And what are the best signatures for us to detect them most accuratel…”
Walaliyadde: Medicine lacks effective methods for early cancer detection and active diagnosis
“Risk tests just tell you your chance of getting cancer. They don't actually tell you if you have cancer right now, and so if you're at higher risk, that probably means you want to get tested more frequently, but we don't have good methods of testing and diagno…”
Ott: Knowing a 30% lifetime cancer risk is not empowering without action
“It's not really empowering to the consumer to know that they're going to have a 30% chance of getting cancer sometime in their life.”
Ott: Illumina effectively holds a justified monopoly in DNA sequencing hardware
“Illumina effectively is a monopoly in the space of DNA sequencing, and in a lot of ways, rightly so.”
Walaliyadde: The magical continuous diagnostic machine of the future is your bathroom
“I actually think the magical diagnostic machine of the future is your bathroom.”
Walaliyadde: Illumina expanded into the applications layer via Verinata, Grail, and Helix
“Lumina has started realizing that as well and is itself building applications. They bought a company called Verinata, which is a NIPT company... You know, they have a cancer company called Grail... There's a company called Helix. So they have started participa…”
Walaliyadde: Single-payer systems like the UK will drive higher genomic test adoption
“In fact, I think in single-payer systems like the UK or outside, we might see much more adoption of some of these genomic tests.”