All We Have Yet To Understand

A tech giant partners with a bioinformatics pioneer to create an entirely new kind of genetic map.

T cells attack a cancer cell.

I believe that all reality is information, and all information creates reality. I am not alone in this belief, but it is nevertheless controversial. Regardless, around this maddening thesis revolves nearly all the intractable problems, paradoxes, and opportunities of modern science, technology, and quite possibly policy and politics. The more we informatize the physical world, the more we can ply its unknown depths.

But there’s so much of it, this information. The recursive joke of an acroamatic god — we understand that information drives everything, but there’s simply too much information to understand. Start with our very minds — comprised of 100 billion neurons connected in no less than 100 trillion paths. Each synaptic firing across one of those hundred-trillion possibilities comprises an informational declaration — and each neuron may fire up to two hundred times a second. Don’t ask me how much potential information that is — I can’t do the damn math.

But in computer science, humanity may have found a path forward. In computer science we have forged a rosetta, a cipher capable of mindlessly absorbing nearly-infinite amounts of information and identifying the patterns which matter to us mere mortals. We spread these patterns as light across an array of pixels — patterns which built novel services like the world wide web, or NetFlix, or Google, or…Pokemon Go.

But information science has so much more to give us, if we could only jump from lighted pixels to the analog, messy, water-based chemistry of our flesh and blood. If we could only turn every process in the human body into information, modeled and monitored in real time, well, we could solve our most immortal of questions, could we not? We would be as gods.


Well if we are as gods, we best get good at it, no?

We had to start somewhere. From the moment Crick and Watson figured out the helical, four-protein structure of the human gene, scientists longed to decode it, to write out god’s original programming. And decode it we did — far faster than we thought we would. This led to an astonishing march of scientific progress — culminating in millions of ordinary citizens voluntarily spitting into small envelopes and mailing them to for-profit companies, all in some kind of devil’s bargain — you might learn that you’re 12.5 percent Ashkenazi Jew, but then again, you might also learn that you’re prone to Crohn’s disease. Or, I don’t know, that you can (or can’t) smell asparagus in your pee.

It all seems so…random and unsatisfying.

Which leads me to set down the bong and tell you the actual subject of this article: A partnership between a total NewCo called Adaptive Biotechnologies, and Microsoft, and the audacious thing they’re trying to pull off: Mapping not the genome (been there, done that), nor engaging robotic CRISPR’d nanoprobes in pursuit of pathogens (that comes later), but rather, the creation of a very special kind of map; a map, if proven accurate, that might just offer the foundation for a very new kind of medicine.

Adaptive and Microsoft call the new territory to be mapped the “immunome,” and the paltry Wikipedia page for the term belies its relative novelty. My first stab at explaining the concept, after an hour on the phone with Adaptive executives, is this: “All genetic material informing the possibility matrix of human disease.” When I bounced that one off Dr. Harlan Robins, the co-founder of Adaptive Biotechnologies, he responded thusly: “The immunome is the set of DNA sequences that code for T cell and B cell receptor genes. These genes are special because they rearrange with the immune cell, and are therefore different between cells (all other DNA is the same in each cell — excluding mutations). These genes code for receptors that recognize pathogens through binding.”

Chad and Harlan Robins, co-founders, Adaptive Biotechnologies (image)

Don’t worry, if you keep reading, I promise to try to make sense of that.

There’s a lot of information in the human genome. But there’s far more information in the immunome — by a factor of two or three. Mapping the genome took ten years. Adaptive and Microsoft hope to map the immunome in the next seven or eight.


Are you still here? Good. Let’s pull back for a second. Why might anyone want to map the immunome? The short answer is to diagnose diseases far earlier and with far more precision than we’ve been able to so far. That alone would revolutionize medicine.

But to really answer that question requires an explanation of how the human immune system works. Here’s the short version: Our immune systems are driven in large part by special cells (the T and B cells Robins mentioned above) which cruise around our bloodstream identifying and attacking bad hombres that have invaded our bodies. These bad hombres have scientific names and properties like “antigens” and “pathogens,” but I’m trying to keep it real here. What really ties this all together is that the T and B cells use informational keycodes to do their work. These keycodes are comprised of snippets of genetic information that match the baddies’ own genetic signatures. When a particular T cell finds a match, it hulks out (a process called clonal expansion), and proceeds to seek and destroy every baddie it can find. Some of those T cells hang around in the blood stream for a long time after, just in case. That’s why vaccines work (sorry, Jenny McCarthy).

Now, there are more than a million baddie keycodes in the known human disease universe. But because of mutations, there are literally trillions of possible baddies, and good old Mother Nature has therefore given our B and T immune cells the potential to make trillions of different kinds of these baddie killers. Each baddie contains roughly eight to ten amino acids — the building blocks of proteins. That’s their keycode. Add up all the informational possibilities and you’re at … well, we’re back to the quandary at the top of this piece. It’s vast. It’s virtually incomprehensible. It’s 10-to-the-fifteenth complicated. There are words for these kinds of numbers. (It’s “quadrillion” and “billiard,” for what it’s worth).

Here’s a better explanation!

If only we could identify those baddies when they’re just starting to blossom in our bloodstreams, before they take root in our cells and start to do their damage….Oh, the anguish we could prevent! Imagine if, as part of a standard annual blood draw, your doctor could discover every latent baddie that has yet to meet its Hulk destroyer? Since we can’t do that, we wait till those baddies dig in and present their damage as “symptoms” — multiple sclerosis, for example, or pancreatic cancer. This often takes years, and by the time the symptoms are expressed, we’re often done for. But if we could find them and target therapies at them before they do their worst damage…

But I get ahead of myself.

Diagnosis is often the foggiest area of medical science. Tests, tests, and more tests is usually the prescription for patients presenting symptoms that might be any number of diseases. We lack precision diagnostics for so many diseases precisely because diseases pose extraordinary informational challenges. Until recently, identifying diseases from their genetic keycodes wandering the bloodstream has been impossible — we didn’t have the map and we didn’t have the key.

This is where Adaptive Biotechnologies comes in. Founded in 2009, the company has pioneered an immunosequencing platform that identifies immune cell receptors — it reads out the baddies’ keycodes. But as the company began to scale its platform, it quickly realized it lacked the computational power, and the machine-learning brawn, to slice through the quadrillions of possibilities that its research had begun to uncover.

Enter Microsoft. As I hit publish on this piece, the two companies have announced a major partnership that combines Adaptive’s sequencing platform with Microsoft’s Healthcare NExT initiative, itself a combination of Microsoft’s formidable research, cloud computing, artificial intelligence, and machine learning capabilities. Microsoft is not only investing in Adaptive (the company has already raised nearly half a billion dollars), it’s committing both full time research staff and substantial compute resources to the effort. The companies won’t quantify the deal, but CEO Chad Robins (Harlan’s brother and co-founder) tells me it will take “hundreds of millions of dollars” to map the human immune system. This deal is a major step toward completion of that goal.

But while mapping the entire human immune system is the goal, the work will likely yield benefits well before such a map is complete. That’s because creating diagnostic tests for just a small part of the map — certain types of cancers, or something as simple as early stage diabetes, for example — could yield extraordinary therapeutic benefits. CEO Robins thinks those tests are only two to three years away. To start, Adaptive plans to pick a few thousand or so antigens, cross tabulate them with several million T cells, and run the machine learning to find where the informational matches light up (it’s far more complicated than that, but I never pretended to be a science writer*). Once the matches are found, a diagnostic test can be created that identifies the early presence of antigens in a patient’s bloodstream in a matter of days, if not hours.

And once those kinds of diagnostic tests are commonplace, we’ll have the informational tools in place to loose those therapeutic CRISPR nanobots in a targeted and life-saving fashion. Not to mention, we’ll have a rich new database of early stage disease progression, which will at the very least inform entirely new approaches to disease treatments.

But it all starts with finding the informational signal in the biological noise that is the human body. Perhaps with intentional irony, Robins compared the initial stages of his project to running a massive Excel lookup table. Bill Gates would be proud.

Adaptive Biotechnologies will be represented at the Shift Forum this coming Feb. 26–28. Join us there to learn more.


  • Any and all mangling of the science involved in this post are mine and mine alone. I promise to correct what I can should readers, and the companies involved, care to alert me to my errors!
One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.