Sepsis is one of the top 10 public health issues in the world, causing an estimated 20% of deaths globally. Despite its prevalence, the cause of sepsis goes unknown in an estimated 30 to 50% of cases.
Sepsis is the overreaction of the immune system in response to an infection—and its this infection that causes diagnosis uncertainties. It is difficult for clinicians to identify which pathogen is causing sepsis, or whether an infection is in the bloodstream or elsewhere in the body. Even in patients with symptoms that resemble sepsis, it can be challenging to determine whether they truly have an infection at all.
Now, researchers at the Chan Zuckerberg Biohub (CZ Biohub), the Chan Zuckerberg Initiative (CZI), and UC San Francisco (UCSF) have developed a new diagnostic method that applies machine learning and metagenomic next-generation sequencing (mNGS) to both microbe and host to identify and predict sepsis cases with 99% accuracy.
Current sepsis diagnostics focus on detecting bacteria by growing them in culture, but that is a time-consuming process—one that doesn’t even correctly identify the infection-causing bacterium sometimes. In the absence of a definitive diagnosis, doctors often prescribe a cocktail of antibiotics in an effort to stop the infection, which just leads to another global health problem—increased antibiotic resistance.
For the study, published in Nature Microbiology, the researchers analyzed whole blood and plasma samples from more than 350 critically ill patients who had been admitted to UCSF Medical Center or the Zuckerberg San Francisco General Hospital between 2010 and 2018.
Rather than cultures, the team instead used mNGS. This method identifies all the genetic data present in a sample, and compares the data to reference genomes to identify the microbial organisms present. The technique allows scientists to identify genetic material from entirely different kingdoms of organisms that are present within the same sample. The researchers also performed transcriptional profiling, which quantifies gene expression, to capture the patients’ responses to infection.
Based on the results, the researchers discovered that when traditional bacterial culture identified a sepsis-causing pathogen, there was usually an overabundance of genetic material from that pathogen in the corresponding plasma sample analyzed by mNGS.
With that in mind, they turned to artificial intelligence—specifically, a machine learning model created by Katrina Kalantar, lead computational biologist at CZI and co-first author of the study.
Kalantar trained the integrated host–microbe model using data from patients in whom either sepsis or non-infectious systemic inflammatory illnesses had been established, which enabled sepsis diagnosis with very high accuracy. She also programmed the model to identify organisms present in disproportionately high abundance compared with other microbes in the sample, and to then compare those to a reference index of well-known sepsis-causing microbes.
The combination of mNGS and machine learning worked even better than the scientists expected. According to the study, they were able to identify 99% of confirmed bacterial sepsis cases, 92% of confirmed viral sepsis cases, accurately predicted sepsis in 74% of clinically suspected cases that hadn’t been definitively diagnosed previously.
“We were expecting good performance, or even great performance, but this was almost perfect,” said Lucile Neyton, a postdoctoral researcher at UCSF and co-first author of the study. “By using this approach, we get a pretty good idea of what is causing the disease, and we know with relatively high confidence if a patient has sepsis or not.”
Importantly, the researchers noted that the detection method works using plasma samples, which are collected from most patients as part of standard clinical care.
“The fact that you can actually identify sepsis patients from this widely available, easy-to-collect sample type has big implications in terms of practical utility,” said senior study author Chaz Langelier, M.D., Ph.D., an associate professor of medicine at UCSF and a CZ Biohub Investigator.
Langelier said one of the goals of the research is to eventually be able to predict outcomes of patients with sepsis, including mortality and length of hospital stay.
“[These] would provide key information that would allow clinicians to better care for their patients and match resources to the patients who need them the most,” he said.
The team also hopes to help address another public health crisis—rising antibiotic resistance. They are looking to develop a model that can predict antibiotic resistance from pathogens detected with the new mNGS/machine learning combo technique.
“There’s a lot of potential for novel sequencing approaches such as this to help us more precisely identify the causes of a patient’s critical illness,” concluded co-senior author Carolyn Calfee, M.D., professor of medicine and anesthesia at UCSF. “If we can do that, it’s the first step toward precision medicine and understanding what’s going on at an individual patient level.”