To Dr. Stephen Williams, Chief Medical Officer of Standard BioTools, proteins and their network of patterns hold clues to the health of our bodies. Measuring enough of them will, over time, reveal insights into our longevity and physical health. His proteomics platform, including SomaSignal Tests, uncovers details from blood proteins to decode potential medical issues, from cardiovascular concerns to cancer risks.
“The likelihood of developing cancer is very dependent on environmental exposure and immune surveillance,” he says. “We think proteins are really good at following those things.”
While the DOC community spent time with Dr. Williams’ technology at DOC 2025’s Living Room Lab, you can hear more from his presentation during the “Blood (Proteomics), Breath and Microplastics” session in the video or read our lightly edited transcript below.
TRANSCRIPT
Dr. Stephen Williams
Thank you. Good to be here. We are all in the information business. We care about us, ourselves living longer and our patients living longer. So we want to predict future risks. We want to work out, can we predict what interventions would be good for someone? We want to be able to detect whether an intervention actually worked, and we want to be able to discriminate between snake oil and things that really work. We’re all in the information business. The question is where is it? Where does this miracle of information about the whole person and about all these different systems reside? The bet that we’ve taken is that it’s in proteins.
If you measure enough proteins, you can get a map like this. There are 7000 spots on this map, each one of which is a protein, and it’s an information landscape. The spots are distributed mathematically over that over there, over the picture here, ones that are more alike close together and ones that are less alike a further apart. They’re color-coded on the left. You can see those blue spots are the ones that are predominantly influenced by genetics. You can see that Francis Collins was half right when he said that genetics is going to explain everything about every disease known to humans, but only half because the other half of the proteins are mostly influenced by something else in the environment.
But the nice thing is that proteins catch that both of those kinds of information, the environment and genetics. And we’re going to start looking at what use that might be. This is a volcano plot. You may not be familiar with volcano plots, but like real volcanoes, the bits you care about are the bits of lava that fly the highest and move the furthest. Those are the protein spots at the top. And of these, of these graphs, the key point here is if you don’t measure it, you can’t find it, and not all of these proteins are things that you would predict from what we already know from human physiology. Our thesis is if you measure enough proteins, you can find effects that are completely unexpected. They don’t work on their own. In medicine we do use individual proteins. But proteins have evolved in networks to transmit information in patterns. If you measure thousands of proteins, you can find the patterns. But the serious point here is you can develop a technology that’s precise and sensitive and useful, but if nobody trusts it, then it won’t get used and it won’t be useful. We have a lot of publications, many of which have got hundreds of citations.
Now we move on to what is the point of all of this? What we’ve done over the past decade or so is to measure thousands of proteins in about 250,000 people with about a million participant years worth the follow up data. And what we’ve done with that is to try and find the protein network patterns that relate to outcomes and that relate to current health state and the ones that change when you intervene in a good way, and the ones that change when you intervene in a way that might be harmful. Today we’ve developed this group of SomaSignal Tests, each of which is a mathematical model, a subset of proteins within the assay itself. The assay has two versions today. One measures 7000 proteins at once. The other measures 11,000 proteins at once, these models between them. Today, they only use about 600 proteins with very little overlap. I’ll talk a bit more about that in the future.
We hope that people like yourselves will become equivalent to genetic counselors, but we proteomic counselors because even though the SomaSignal Tests are useful, they don’t cover every impact of every intervention and every lifestyle change. We think that as we learn more about human physiology, will actually be able to interpret individual proteins as well. But that’s for the future. These are the tests that have been validated so far. And in those publications that we talked about. When we saw earlier about, about the whole person, it’s not simply, you know, a urologist or a single system. What we’re saying here is that you can measure any one of these or all of them from the same blood test at the same time. You can measure changes over time. So you can see they do cover cardio, metabolic health. That was all that was our focus, and cancer. We may have a chance to talk about that later. The likelihood of developing cancer is very dependent on environmental exposure and immune surveillance. We think proteins are really good at following those things. But those tests are in the process of being validated.
So how does this work in practice? How does this help you work out whether a particular change in lifestyle or pharmacology actually did something useful? Well, this is the results of a of a test of, of actual of a dietary intervention. It was a pretty severe caloric restriction study. Each of the rows is a proteomic test. Each of the spots says, well, what happened to that test during the year in this study? You can see that dietary intervention in people who lost more than ten kilos, the purple spots improved a lot of body composition. The glucose tolerance test, the proteins mimic the result of a glucose tolerance test without fasting and without glucose. You can see that there was a huge improvement in those tests, as well as cardiovascular events and body composition.
As you might have expected, some of these changes weren’t slow. That even though that was a year study, as a component of one of the studies, there was that you could measure these things over time and you can see changes quickly. And the tests have also been used by far more. We’ve heard some this some information this morning about cheap ones. Novo has published on using the SomaScan assay, and they used it in both ways that I talked about at the beginning. They measured the changes in individual proteins. You see one of those volcano plots where they found 495 proteins that change many of those in pathways that impact the some of the benefits and some of the this benefits like loss, muscle mass loss.
On the right they also measured the cardiovascular event prediction test. Now you might say, why on earth do you care? Because they did this big outcome study and the event rate was less. So why do they need to show the protein effect? Well, the key point here is that their cardiovascular events study was 57,000 participant years worth of data. It took them seven years to run the benefits. Those cardiovascular benefits to those patients were all delayed while they ran a seven-year study. We were interested in measuring those things today in individuals, and that’s what they did on the right, the the bars crossing the dashed line of zero change in the placebo groups and the ones that went fell below that other proteomic predictor of cardiovascular event rates falling over time in much smaller studies.
Talking of much smaller studies, this is a randomized control study of providing this information. So this was run by Rosalind Gale, who’s sitting there at the back. People told us that behavioral change. Now maybe I don’t think this audience believes it, but behavioral change is a lost cause that even if you provided people with an individualized, accurate assessment of their likelihood of dying, developing heart failure, or having a stroke, they wouldn’t change anything.
What we did was to take a group of 400 people, and we randomly assigned them to be informed or not informed, and we measured prescribing rates of enhanced cardio-protective drugs. That’s the plot on the left. The blue section is the people who were informed, and these were all people who should have been on an SGLT2 inhibitor or GLP-1 or more intensive statins, but weren’t. You can see the prescribing rates went up to between 30 and 40%. And on the gray side, you can see that the uninformed group there was a clinical trial effect from extra attention, but they only went up to about 10%. So providing people with individualized information changed behavior. If you want to visit the Learning Lab, you can see whether it changes your behavior, because you can get this done yourselves. Thanks.