The Incredible, Sci-Fi-Like Future of Blood Testing
One of the easiest ways to check your health is with a drop of blood. At a wellness screening a few months ago, a technician pricked my finger, squeezed out a drop of blood, and loaded it into a machine about the size of a small microwave oven. Within five minutes, I knew my cholesterol numbers. My numbers fortunately weren’t too high. If they had been, this simple blood test would have helped my doctor decide whether to put me on a lipid-lowering drug.
The basic blood tests performed at doctor’s offices and wellness fairs play a big role in managing our health and preventing common diseases like diabetes and heart disease. Several newer blood tests, which detect fragments of DNA circulating in our bloodstream, are more technologically sophisticated, including newer prenatal tests and cancer tests. But much more could be learned about our health from a drop of blood, if only we had the right tests — and if those tests were as fast and inexpensive as a cholesterol test.
Several biotech companies are attempting to develop sophisticated, but inexpensive blood tests that could be taken each time you visit your doctor. Perhaps the most well-known company is the now-disgraced Theranos, which was founded by one-time Silicon Valley favorite Elizabeth Holmes after she dropped out of college at Stanford University. Theranos claimed to have a device that could perform hundreds of lab tests with a single drop of blood. But the technology never worked, and Theranos faked its results. Holmes is now awaiting trial for fraud.
Other companies are working from a more solid scientific foundation. SomaLogic, based in Boulder, Colorado, has developed a platform called the SomaScan, which works by measuring thousands of proteins in a blood sample, and finding patterns that relate to health traits. The company already makes a few tests that are designed to pick up early signs of liver disease, heart disease, diabetes, and a few other conditions.
One of the newer companies to enter the field is Nautilus Biotechnology, founded by Stanford professor Parag Mallick and tech entrepreneur Sujal Patel. Nautilus just announced a $76 million investment from investors that include Paul Allen and Jeff Bezos. Like SomaLogic, Nautilus aims to develop an inexpensive test that taps into a rich store of health information carried by the thousands of proteins circulating in our blood.
What will it take for tests like these to become something that you would routinely take at a doctor’s office or wellness fair? The main challenges to be solved fall under three themes: protein chemistry, automation, and software. When these three elements are in place, more sophisticated blood tests will be poised to become a reality.
Working With Proteins
Your blood is packed with information carried by proteins. That is the central premise behind the technologies built by Nautilus and others. While DNA carries the genetic information you’re born with, proteins change continually and reflect the state of your health. Proteins circulating in your blood vary throughout the day, as well as over weeks, months, and years. They vary with your diet, your fitness level, your stage of pregnancy, and your cancer status.
Most routine blood tests that rely on proteins measure only one or a few at a time. Measuring hundreds or even thousands of circulating proteins would provide a much richer picture of your current health. In the jargon of the protein field, the totality of your proteins is called the proteome (just like the collection of all your genes is your genome). What companies like Nautilus and SomaLogic aim to do is to measure proteomes — in a way that is cheap, fast, and reliable.
However, there is a reason why the term “genome” is discussed much more often than “proteome” in biotech: Compared to DNA, working with proteins is hard. DNA is a chemically uniform double helix, with its sequence of A’s, C’s, G’s, and T’s packed neatly inside. This is what makes DNA a good information storage molecule — like a hard drive, it needs to be read out consistently by the same hardware, regardless of the information it carries. Proteins, on the other hand, do much of the chemical work in the cell, and thus have an enormous range of shapes, sizes, and chemical properties.
In spite of the challenges, technologies designed to read out proteomes have steadily improved over the past few decades. But the technologies often involve expensive equipment that is hard to use — equipment that will never end up in a doctor’s office, much less a wellness fair. That’s why blood testing companies are focused on automation and miniaturization.
Shrinking the Lab
My recent cholesterol test was performed on a small machine that automated everything. The technician put a drop of blood on a slide, loaded it into the machine, and pressed start. To be successful, proteome blood tests will need to be just as easy.
That’s why biotech companies invest heavily in so-called lab-on-a-chip technologies, which use microfluidic devices to manipulate extremely small volumes — millionths or billionths of a liter. Microfluidic devices are easily automated and can perform a very broad range of lab tests on a device that’s about the size of a dime. These devices can take a drop of blood, process it, split it up, and perform a dozen or more different tests.
A complementary technology consists of miniature slides arrayed with tens of thousands of probes, which is the foundation of the SomaLogic platform. In a SomaScan, thousands of probes bind proteins in a small blood sample, and the results are read out by imaging the slide. These miniaturized, automated technologies quickly yield a lot of data, which leads to another problem: how to make sense of the deep, high-dimensional proteomics data.
Modeling Your Health
A doctor doesn’t need software to interpret your cholesterol test. But newer blood testing technologies rely heavily on software tools from start to finish. Software is needed to process the data that comes off the testing device, to manage the secure storage and transfer of the very large files that hold the data, and to interpret the results so that they’re understandable by a human physician. The need for extensive computation is the price you pay for the much richer data set that comes from the proteome circulating in your blood.
To process and manage large data files from blood proteomes, biotech companies can rely on advances in other fields, like cloud computing. The bigger challenge for these companies is to understand what the data say about someone’s health. That understanding depends on figuring out the best way to assess patterns across a person’s proteome.
To tackle that problem, companies are turning to machine learning and mathematical modeling. Researchers with SomaLogic analyzed 85 million protein measurements from nearly 17,000 blood samples, using mathematical models to capture the relationship between blood levels of hundreds of proteins and specific health traits. Patterns of protein levels of the study volunteers reflected, among other things, their lean body mass, how much alcohol they drank, how often they exercised, and how well their kidneys functioned.
Nautilus co-founder Parag Mallick and his Stanford research group built a computer model of growing tumors using information gleaned from the blood proteome. Cancer cells generate a lot of debris, shedding proteins as they grow and metastasize. Those discarded proteins circulate in the blood, and Mallick and his group used protein data to predict the state of tumors — how fast they grow, how much oxygen they consume, and how effectively they recruit new blood vessels to feed further growth. Without reliable software, none of these tumor traits could be assessed from a mere blood test.
There is much more to learn about your health from a drop of blood than the few simple numbers that we typically get at an annual check-up. Most of today’s blood tests rely on relatively simple lab assays that were developed decades ago by medical chemists. To get more out of each drop of blood, however, will require a single package that integrates protein chemistry, microfluidics, and software. This will also require teams with a much wider range of expertise — teams that include “protein chemists, chip designers, molecular biologists, data scientists, material scientists, biophysicists, optical engineers, microfluidics engineers, bioinformaticists, software engineers, and more."