Design a site like this with
Get started

Interviewing Innovators: A Liquid Biopsy with Dr. Oguzhan Atay, CEO, BillionToOne

What if we could replace every biopsy with a blood test?  …a liquid biopsy

Systems have accelerated.  It’s now possible to positively impact millions in short time frame.

Access is key to our mission. We released our first diagnostic three months ago, and it is 100% reimbursable by Medicaid as well as other major insurance providers.

Innovator: Dr. Oguzhan Atay

Expertise: Quantitative genetics and machine learning

Idea: High-resolution liquid biopsy


Phi Beta Kappa, Princeton, BS Molecular Biology and minors in Physics, Computer Science, and Applied Mathematics (summa cum laude)

PhD Systems Biology, Stanford, Dissertation topic: Switch-like transitions modularize complex biological network 

About Dr. Oguzhan Atay:

Dr. Oguzhan Atay believes that diagnostics can be made much more powerful, accurate, and affordable if they were engineered to be quantitative. At BillionToOne, Dr. Atay and his co-founder, Dr. Tsao, developed a patent-pending molecular counter platform that increases the resolution of molecular diagnostics by over a 1000-fold. Dr. Atay has led BillionToOne through Y Combinator and raised a total of 17.5M in funding including from venture capital funds and investors who previously invested in tech companies such as SpaceX, Box, Spotify, Palantir, Braintree, and biotech companies such as Counsyl, WebMD, and Omada Health. Dr. Atay received his PhD from Stanford, where his work was published on highly regarded peer-reviewed journals including on the Cover of Cell Systems.

“We looked to solve the underlying problem of amplification bias and noise.  If we were able to remove the noise, then biological and diagnostic mysteries could be explained using math.” – Oguzhan Atay, CEO, BillionToOne

What have you learned about biological systems and humanity?

I went from optimizing for the next 100 years to optimizing for the next 10 years.

Systems have accelerated.  It’s now possible to positively impact millions in short time frame.

My plan was to enter academia after Stanford.  The goal for academia has been to build a body of knowledge for generations to come, to drive forward understanding without the confines of investor expectations.  However, the system has changed, both in and outside of academia.  Now the real place to build extensible systems and amplify our capabilities is at the intersection of science, technology and capital.

As you have worked on making a liquid biopsy (blood tests replacing biopsies) a reality, what have you learned about science, ie the intersection of biology and math?

Diagnostics can be made much more powerful, accurate, and accessible if they were engineered to be quantitative.

Marrying computer science and systems biology enables us to deliver affordable and accessible products to millions, perhaps billions.  Reducing the cost of healthcare is possible!  This is our mission.

Why are you the person to bring us this innovation?  Who are you at your core?  What about your background/upbringing led you to do this?

My co-founder and I met at a Caltech sponsored summer camp, and then we were roommates at Princeton.  We enrolled in one of the country’s first Integrated Science programs at Princeton. 

With this integrated science background, David and I were able to use deep knowledge in multiple fields to solve problems.  Normally a machine learning expert will wait for data output and then optimize.  And biochemists continue to build their experiments without understanding how the generated data will be analyzed. 

Our breakthrough needed the same person to have deep expertise in multiple areas. For instance, upon exiting Princeton, I had offers for multiple PhD paths: MIT for Physics, Cambridge for Applied Math, and Stanford for Biology.  My co-founder, David Tsao, was a spacecraft engineer, a physicist, a software engineer, and a bioengineering PhD. We used ideas from different disciplines to solve problems in biology.  We were able to deliver a massive phase shift improvement, one thousand times the previous diagnostic resolution.

Growing up, I tested high in math and science and learned by reading encyclopedias.  I was usually 3-4 years ahead of what was being taught in school. I was always a high achiever. I tested #1 in University Entrance Exam in Turkey among 1.5M students and attended Princeton.

Let’s dive deep into your innovation. What are you aiming to do that has not been done?

Together with our scientists, my co-founder, David, and I have developed QCT: Quantitative Counting Templates that can accurately count the DNA molecules using next generation sequencing. 

What do you mean by next gen sequencing?

Next-generation sequencing is the high-throughput methods in which letters that make up our DNA, our genome, is read. It required teams of scientists and billions of dollars to sequence the first human genome at the turn of the century. This cost has plummeted at a rate faster than Moore’s law, enabling a revolution that has the capacity to surpass even semiconductors’ effect on computing. Currently, it costs only ~$1,000 to sequence all of three billion letters that make up our genome. This has already changed clinical practice from diagnosing rare disorders to prenatal testing.

However, many of the other, potentially transformative clinical applications require us to not only ‘read the letters’ but quantify their levels. Traditional methods for next-generation sequencing add a significant level of noise into the input level.  Currently, biologists don’t design assays for machine learning capabilities.  Machine learning incrementally works to get more signal out of the noise.  We looked at how to remove the noise. If you remove the noise, then you have far more control over and thus visibility into the signal. 

How do you remove the noise?

Current approach to biological innovation is trial and error lab development of diagnostics or tests, also named assays, This development is qualitative, expensive, and it is hard to know if the assay will work at the design stage.

We asked ourselves, what is the problem?  One of the main problems is the amplification bias.

Before sequencing the DNA, you need to amplify a gene by 1 million to 1 billion times.  The input (original gene sequence) and output (amplified gene sequence, the one we actually measure) are only loosely correlated, due to the fact that an exponential amplification process adds 20% to 200% noise into the system. Due to this noise, scientists do not know what the input actually was, and are forced into costly trial and error development cycles. 

We looked to solve the underlying problem of amplification bias and noise.  If we were able to remove the noise, then biological and diagnostic mysteries could be explained using math.

We added specifically designed artificial DNA molecules into the sample that corresponds to the gene of interest.  Then when the genes are amplified, we can know exactly by how much, and therefore we can measure and remove the amplification noise, and we are left with pure signal.  The real breakthrough here is to insert a set of artificial DNA molecule as a group of “markers” within the machine learning process, to optimize for the capabilities of machine learning, rather than using sophisticated machine learning algorithms after the noisy data are generated.  We modified our biochemical process to empower our machine learning.  This led to a 1000X improvement in diagnostic resolution.

By removing the noise, we made the DNA molecules countable.  Then every molecular diagnostic is possible, and far more affordable.  Much of what is now done with biopsies may in the future be done using blood tests. 

For example, today we are shipping CLIA Approved Unity Screen, which is a prenatal blood test for the 5 most tested inherited genetic disorders.  Before our test, first, mothers would get screened for genetic carriers of diseases like cystic fibrosis, spinal muscular atrophy, sickle cell disease, and thalassemias.  10-15% are carriers, which create a lot of anxiety for them during a sensitive period.  Then, the requirement is to screen the father, which is more problematic as 40% of births in the US are by single mothers.  Finally, if both test positive, then an amniocentesis is performed, but many parents opt out due to perceived risks of the test and delays associated with the workflow. With our test, we quantify the DNA molecules of the fetus that are floating around in the mothers’ blood. We take a blood sample from the mother, and find the result with an accuracy rate of 99%+. In this way, only truly positive pregnancies go through a confirmatory amniocentesis before making any decisions. 

Our test is less expensive than the original blood test a mother would take, which means these tests can be available for a much wider population.  We worked to bring our test to the market in such a way that it is 100% reimbursable by Medicaid.  Our focus in on access: making a difference for everyone.

While UNITY is the only screen that test both mother and fetus for these disorders, other, similar tests exist for Down Syndrome, as Down Syndrome is a much easier disorder to screen for. We will add Down Syndrome and other chromosomal disorders to our test next year, which we plan to call UNITY Complete.  We aim to be the most reliable blood test for all prenatal testing.

Next in our pipeline is cancer diagnostics. 

We can figure out if a cancer treatment is working or not. We can quantify the amount of tumor DNA in the blood stream using a cell free DNA test, on a regular basis, to see how the cancer treatment is working. Today patients wait for tumors to either grow or recede.  This testing can be done to optimize treatment and also to test drugs in development as they are being used, vs long lag periods.  This is not possible today due to the inability to quantify the cell-free tumor DNA in blood. 

What is your stretch goal?  Where could your invention take us 5 years forward?

With predictive engineering-based designs and machine learning, our diagnostic development cycles are greatly amplified, so what I do know is we can achieve far more than even I am able to imagine. 

We believe we can transform all molecular diagnostics.  By integrating biological science with machine learning, a new product team of 2-3 people can develop a new diagnostic product every 4-6 months.  This is the fastest growing area in diagnostics.

We want to reduce costs for all diagnostics and related R&D.  You can know sooner if a drug therapy is actually working, in testing or in real life.

Where can we learn more?
Podcast with Dr. Oguzhan Atay

BillionToOne: /

Important Disclosures

Current relationship, holdings, no representations: I have no financial relationships with BillionToOne or their employees. I have no holdings in BillionToOne.  This blog contains no investment advice, no warranties of accuracy, no duty to update, and no offer of securities.  This blog contains many forward-looking statements.


Published by Emmy Sobieski

I have 24 years experience analyzing new technologies, interviewing founders, and learning why they are the people driving innovation. This blog is my way of sharing insights of those on the cutting edge of technology.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: