Sepsis, the overreaction of the immune system in response to infection, causes about 20% of deaths worldwide and up to 20-50% of deaths in US hospitals each year. Despite its prevalence and severity, the disease is difficult to diagnose and treat effectively.
The disease can lead to decreased blood flow to vital organs, inflammation throughout the body, and abnormal blood clotting. Therefore, if sepsis is not recognized and treated promptly, it can lead to shock, organ failure, and death. But it can be difficult to identify which pathogen is causing sepsis, or whether an infection is in the bloodstream or elsewhere in the body. And in many patients with symptoms that resemble sepsis, it can be difficult to determine if they really have an infection.
Now, researchers from 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 to advanced genomic data from the microbe and host – to identify and predict cases of sepsis. As reported on October 20, 2022 in Natural microbiologythe approach is surprisingly accurate and has the potential to far exceed current diagnostic capabilities.
Sepsis is one of the top 10 public health problems facing mankind. One of the main challenges of sepsis is diagnosis. Existing diagnostic tests are unable to capture the two-sided nature of the disease – the infection itself and the host’s immune response to the infection.”
Chaz Langelier, MD, Ph.D., senior author, associate professor of medicine, UCSF division of infectious diseases, and CZ Biohub researcher
Current diagnostics for sepsis focus on detecting bacteria by growing them in culture, a process that is “essential for appropriate antibiotic therapy, which is essential for survival from sepsis,” according to the researchers behind the new method. But culturing these pathogens takes time and does not always correctly identify the bacteria causing the infection. Similarly to viruses, PCR tests can detect that viruses are infecting a patient but do not always identify the particular virus causing sepsis.
“As a result, clinicians are unable to identify the cause of sepsis in approximately 30 to 50 percent of cases,” Langelier said. “It also leads to a mismatch between the antibiotic treatment and the pathogen causing the problem.”
In the absence of a definitive diagnosis, doctors often prescribe a cocktail of antibiotics in an attempt to stop the infection, but the overuse of antibiotics has led to an increase in antibiotic resistance around the world. “As physicians, we never want to miss a case of infection,” said Carolyn Calfee, MD, MAS, professor of medicine and anesthesia at UCSF and co-lead author of the new study. “But if we had a test that could help us determine precisely who doesn’t having an infection, it could help us limit the use of antibiotics in these cases, which would be really good for all of us.”
The researchers analyzed whole blood and plasma samples from more than 350 critically ill patients who had been admitted to UCSF Medical Center or Zuckerberg San Francisco General Hospital between 2010 and 2018.
But rather than relying on cultures to identify pathogens in these samples, a team led by CZ Biohub scientists Norma Neff, Ph.D., and Angela Pisco, Ph.D., instead used the next-generation metagenomic sequencing (mNGS). This method identifies all nucleic acids or genetic data present in a sample and then compares this data to reference genomes to identify the microbial organisms present. This technique allows scientists to identify genetic material from entirely different kingdoms of organisms – be they bacteria, viruses or fungi – present in the same sample.
However, detection and identification of the presence of a pathogen alone is not enough for an accurate diagnosis of sepsis. Biohub researchers therefore also performed transcriptional profiling – which quantifies gene expression – to capture the patient’s response to infection.
Next, they applied machine learning to mNGS and transcriptional data to distinguish between sepsis and other serious illnesses and thus confirm the diagnosis. Katrina Kalantar, Ph.D., senior computational biologist at CZI and co-first author of the study, created an integrated host-microbe model trained on data from patients in whom sepsis or non-infectious systemic inflammatory diseases had been established, which allowed a diagnosis of sepsis with very high precision.
“We developed the model by looking at a metagenomic dataset alongside results from traditional clinical trials,” Kalantar explained. To begin, the researchers identified changes in gene expression between patients with confirmed sepsis and non-infectious systemic inflammatory conditions that appear clinically similar, then used machine learning to identify which genes might best predict these changes.
The researchers found that when traditional bacterial culture identified a pathogen causing sepsis, there was usually an overabundance of genetic material from that pathogen in the corresponding plasma sample analyzed by mNGS. With this in mind, Kalantar programmed the model to identify organisms present in disproportionately high abundance relative to other microbes in the sample, and then compare them to a benchmark of well-known sepsis-causing microbes.
“On top of that, we also noted down any viruses that were detected, even if they were at lower levels, because they really shouldn’t be there,” Kalantar explained. “With this relatively simple set of rules, we were able to do quite well.”
“Nearly perfect” performance
The researchers found that the mNGS method and their corresponding model worked better than expected: they were able to identify 99% of confirmed cases of bacterial sepsis, 92% of confirmed cases of viral sepsis, and were able to predict sepsis in 74% of cases clinically. suspects. who had not been definitively diagnosed.
“We expected good performance, even great performance, but it was almost perfect,” said Lucile Neyton, Ph.D., postdoctoral researcher at Calfee Lab and co-first author of the study. “Using this approach, we get a pretty good idea of what’s causing the disease, and we know with relatively high confidence whether a patient has sepsis or not.”
The team was also thrilled to discover that they could use this combined method of host response and microbial detection to diagnose sepsis using plasma samples, which are routinely taken from most patients. as part of standard clinical care. “The fact that you can actually identify patients with sepsis from this type of widely available and easy-to-collect sample has big implications for practical utility,” Langelier said.
The idea for the work stems from previous research by Langelier, Kalantar, Calfee, UCSF researcher and president of CZ Biohub, Joe DeRisi, Ph.D., and their colleagues, in which they used mNGS to effectively diagnose lower respiratory tract infections in critically ill patients. . Because the method worked so well, “we wanted to see if the same kind of approach could work in the sepsis setting,” Kalantar said.
The team hopes to build on this successful diagnostic technique by developing a model that can also predict antibiotic resistance in pathogens detected with this method. “We’ve had success doing this for respiratory infections, but no one has found a good approach for sepsis,” Langelier said.
Additionally, the researchers hope to eventually be able to predict sepsis patient outcomes, “such as mortality or length of hospital stay, which would provide key insights that would allow clinicians to better care for their patients and tailor patient resources. who need it the most,” Langelier said.
“There’s a lot of potential for new sequencing approaches like this to help us more accurately identify the causes of a patient’s severe disease,” Calfee added. “If we can do that, it’s the first step towards precision medicine and understanding what’s going on at an individual patient level.”
Kalantar, KL, et al. (2022) Integrated host-microbe plasma metagenomics for the diagnosis of sepsis in a prospective cohort of critically ill adults. Natural microbiology. doi.org/10.1038/s41564-022-01237-2.