A decade ago, as a surgical resident at Stanford with a focus on trauma and critical care, Tim Sweeney, MD, PhD, found himself caring for a young man who had survived an ATV rollover accident. The patient’s organs were failing, and despite every advanced intervention the ICU could offer, his condition kept deteriorating. “I had him on all the fancy organ support tools in the ICU, but he was maxed out on therapy,” Sweeney told Inside Precision Medicine. “I went to my attending team and asked, ‘What else can we do?’ And the answer was: nothing. At that point, all there is is prayer.”
The moment crystallized a reality for Sweeney that many clinicians know: in sepsis, once supportive measures are exhausted, there is no specific medicine to offer. Patients’ lives hinge on the hope that their immune system, battered and disordered, will turn the corner on its own. That frustration propelled Sweeney in his residency to join the laboratory of Purvesh Khatri, PhD, now a Stanford professor of medicine and biomedical informatics, who was developing methods to mine large “omics” datasets for hidden patterns. Khatri had trained with the late data pioneer Atul Butte, MD, PhD, who specialized in stitching together datasets that most biologists considered too disparate to yield insights. When Sweeney saw the approach, he suggested applying it to sepsis.
What began as an academic curiosity about 11 years ago soon led to a series of papers showing that gene expression profiles in whole blood could be used to tell whether a hospitalized patient had an infection, to distinguish bacterial from viral causes, and to predict 30-day mortality. The findings suggested that the immune system’s transcriptional response carried enough information to diagnose and prognosticate sepsis more precisely than any conventional marker. Based on those results, Sweeney and Khatri co-founded a company called Inflammatix eight years ago, aiming to turn host-response signatures into clinically usable diagnostics.
Now, more than a decade after the work began, Sweeney and Khatri are celebrating the publication of three papers in Nature Medicine that together mark a transition from possibility to practice. One paper details work supporting the FDA clearance of TriVerity, a 30-minute blood test that uses a compact device to deliver infection diagnosis and prognosis directly at the point of care.
The other two Nature Medicine articles, co-authored with collaborators across dozens of institutions, establish consensus frameworks for using transcriptomics to define subtypes of sepsis and other critical illnesses—life-threatening health conditions that require intensive medical treatment and can lead to vital organ failure. Together, the articles represent both the clinical validation of a new diagnostic platform and the conceptual foundation for precision medicine in the ICU. It’s a feat that could immediately save lives.
“I tend to be cautious about overselling things, because translation always takes time,” Khatri told Inside Precision Medicine. “But this particular set of manuscripts really feels like the beginning of precision medicine in critical illnesses.”
True point-of-care
The focal point of this trio of papers is the validation of TriVerity, a rapid test that distills more than a decade of work into a cartridge-based system that was approved by the FDA in January 2025. With Triverity, clinicians can collect a patient’s blood sample, load it into a disposable cartridge, insert it into a benchtop instrument from Inflammatix called Myrna, and walk away. Within 30 minutes, the system delivers a readout of whether the patient has an infection, the source of that infection (bacterial or viral), and how likely the patient is to progress to critical illness requiring intensive care. Khatri said, “True point of care, with no hands-on except for taking a blood sample, was unprecedented.”
When the Inflammatix co-founders began this work, Khatri hypothesized that host response could be used as a diagnostic, but he never expected it to be as promising as it now appears to be. The expectation had been that host response signatures would fail in immunocompromised patients, such as those with cancer or transplants, who represent a significant minority of ICU cases. Yet the data showed otherwise.
For Sweeney, TriVerity represents not just a technological milestone but also proof that transcriptomics can finally escape the research lab. “This is the culmination of years of work,” he said. “It’s moving a rapid transcriptomic profile from an idea to a clinically validated tool out in the world.”
The broader goal is not just to advance a single diagnostic but to retire an outdated way of thinking about sepsis. “One of the interesting things about this trio of papers is that nowhere do we really get specific about whether it is ‘sepsis’ or ‘not sepsis,’” Sweeney said. “That binary is not meaningful to clinicians. If I tell you the patient in bed six has ‘not sepsis,’ what does that mean? Is it bacterial? Viral? Something non-infectious but severe? That’s not helpful.”
Sweeney draws a comparison to oncology. “It’s like saying someone ‘has cancer.’ What does that mean? Breast, lung, colon? What subtype, what mutation? That’s how cancer treatment advanced, and that’s where critical illness is heading.” Khatri frames it in similar terms. “Critical illnesses today are defined symptomatically. We don’t know the underlying biology. These papers show we can define them by biology, and that’s what enables precision medicine.”
From diagnosis to treatment
For Sweeney, this has been a long-awaited step toward rational therapy. Sweeney said, “If someone is going to go to the ICU, how do we eventually go from zero drugs for sepsis to a framework that actually allows us to find druggable subtypes? That’s what these consensus papers provide.”
The two companion Nature Medicine papers address a deeper problem. Suppose clinicians can now diagnose infection and stratify severity rapidly, as is the case with TriVerity. What comes next? How do those insights lead to therapies that change outcomes? Over the past decade, dozens of research groups have proposed frameworks to classify sepsis patients into subgroups based on protein or gene expression patterns. These “endotypes” or “subtypes” often appeared promising, but Khatri said that there were too many competing standards and no agreement on which to use. The result was paralysis: the insights could not be translated into clinical trials.
The new work in Nature Medicine changes that. Drawing on collaborations across the United States and Europe, the teams harmonized existing schemas and identified a consensus blood transcriptomic framework for sepsis and a consensus immune dysregulation framework for sepsis and critical illnesses. The analyses showed that instead of being discrete, endotypes exist along a continuum of immune dysregulation, most prominently along two axes: myeloid dysfunction and lymphoid dysfunction. Patients may show one, both, or neither. “What we discovered was that there are four endotype schemas in total,” Khatri explained. “But there is no absolute cutoff; they’re on a continuum.”
This continuum not only maps cleanly onto severity—patients with greater dysregulation fare worse—but also ties into drug responsiveness. Using single-cell RNA sequencing, the researchers linked the transcriptional profiles to specific cell types, grounding the framework in biological mechanisms. For example, they were able to clear up some confusion around the use of steroids in sepsis. “We show that steroids, which target lymphoid immune responses, actually benefit patients with lymphoid dysregulation but not those without it,” Khatri said. “That essentially clarifies the controversy surrounding steroids in sepsis. Trial after trial has been contradictory because patients weren’t stratified. Now we can separate them.”
Beyond sepsis
The field of critical illness research has long been mired in a catch-22. Precision medicine requires the ability to identify subgroups of critical illnesses like sepsis at the point-of-care, but without that ability, there is no way to run precision trials. Lacking trials, the field had no path to therapies. “Up until now, we were stuck,” Khatri said. “We couldn’t measure at the point-of-care, so we didn’t know which schema to bring into the clinic. Because we couldn’t bring the schema into the clinic, we couldn’t design trials. Because we couldn’t design trials, we couldn’t measure. That cycle has finally been broken.”
With TriVerity now FDA-cleared for diagnosis and prognosis, and with consensus frameworks using transcriptomics and defining biologically grounded subtypes, the building blocks of precision medicine in sepsis and critical illness are finally in place. “This is why I feel confident for the first time in ten years to say that we’re at the beginning of precision medicine for sepsis,” Khatri said. “We have all the tools, and we’re ready.”
Both Sweeney and Khatri emphasize that the implications extend well beyond sepsis. A 2022 Nature Medicine perspective argued that all of critical illness should be redefined in molecular terms, much as oncology has moved from organ-based labels to molecular subtypes. “Right now, you’re in the ICU because you had a car accident or a major surgery or sepsis,” Sweeney said. “But you’re not still there because of those events. You’re still there because of runaway inflammation and immune dysfunction.”
The frameworks developed in these studies show that dysregulation signatures transcend the initial cause of ICU admission. Patients with severe physical trauma, acute respiratory distress syndrome (ARDS), or burns can be classified along the same myeloid-lymphoid axes as those with infection. “Two patients may both arrive in the ICU—one from sepsis, another from trauma—but if both have lymphoid dysregulation, they may both benefit from corticosteroids,” Sweeney said. “Another group may benefit from anakinra, or from combination therapies. That’s where precision medicine for critical illness begins.”
Khatri underscores the breadth of the vision. “Our framework basically captures the entire critical illness space,” he said. “It’s not just about sepsis. It’s everything. For the first time, we can imagine assigning the right treatment to the right patient at the right time.”
Bedside to bench and back again
The core diagnostic signature behind TriVerity was derived not from proprietary datasets but from mining publicly available gene-expression data. “Public data is like public land,” Khatri said. “People think it’s garbage, but if you keep digging, you’ll find oil or gold.”
That conviction carried through even when mentors warned him against tackling sepsis. Atul Butte, who had moved from Stanford to UCSF, counseled Khatri against the startup. “He told me, ‘Sepsis is the graveyard of startups. It’s where companies go to die,’” Khatri recalled. But he pressed ahead, arguing that if the gene signature was validated across global populations, it stood a higher chance of succeeding clinically.
When FDA clearance came through nine months ago, Khatri wrote to Butte, who was then in the ICU. “I told him, it just happened. We got FDA clearance using publicly available data,” he said. “I’m so glad we achieved it while he was still alive.” Butte passed away in June 2025, a few months after the approval.
The narrative of Inflammatix’s decade-long effort is unusual in that it moves fluidly between clinical practice and computational discovery. Sweeney’s experiences in the ICU set the questions. Khatri’s data mining provided answers. Together, they pushed those answers back into the clinic through a startup that built the hardware and workflows to make transcriptomic diagnostics usable in real time.
“It’s a bedside-to-bench and back-to-bedside story,” Sweeney said. “First, you identify the need at the bedside. Then you do the computational work at the bench. Then you build a device to bring it back to the bedside.” That iterative loop, reinforced by the publication of three Nature Medicine papers, may help turn sepsis research from a “graveyard of startups” into a proving ground for precision medicine.
Neither Sweeney nor Khatri claims that sepsis has been solved. The road to new therapies remains long, and translation to clinical practice is rarely smooth. However, both perceive the accomplishments in these three papers as a significant turning point.
“It’s not that these are the first papers to suggest transcriptomic diagnostics or subtypes,” Sweeney said. “What makes them special is that they demonstrate an FDA-cleared platform for a blood transcriptional profile, and they demonstrate consensus across dozens of institutions. That’s what takes us from an idea to clinical practice.”
For Khatri, the excitement lies not just in validation but in possibility. “After ten years of work, I finally feel confident enough to say this: we are likely at the beginning of precision medicine for sepsis and critical illness,” he said. “We have the tools. And we are ready.”
And for patients like the young man who haunted Sweeney’s early training, the hope is that the next generation of physicians won’t have to choose between supportive care and prayer. They will have precision diagnostics, consensus frameworks, and—if the vision holds—targeted therapies that finally change the trajectory of critical illness.
