Machine learning maps changing immune response over time in sepsis

Researchers have mapped how the immune response in people with sepsis changes over time. 

Sepsis is a life-threatening condition that occurs when there is misfiring of the immune response to infection. In sepsis, there is failure of the vital organs, and it can be fatal even when treated quickly. It’s estimated that there are over 160 million cases and about 21 million deaths from sepsis worldwide each year.

Current treatments for sepsis focus on treating the underlying infection with antimicrobials and providing supportive care for failing vital organs. Despite efforts to treat the misfiring immune system, none have been successful at achieving improved outcomes for patients.

In the new study, published today in Immunity, researchers sought a more detailed understanding of the immune response mechanisms that change over time in patients with sepsis.

They analyzed blood samples collected at four different timepoints (between admission to and discharge from critical care) from critically ill patients with sepsis at Guy's and St Thomas' NHS Foundation Trust. To build a detailed picture of the immune response, called an 'immune profile,' the researchers looked at multiple layers of immune response information in the blood samples, including data on immune cells, gene expression and protein expression analyses that are changing. 

Using machine learning approaches, they then combined these layers of information to generate a more comprehensive immune profile for the first time in patients with sepsis. 

The analyses revealed an immune trajectory that had three distinct, temporal immune states (referred to as STImS) between admission and recovery, with each state involving different immune cell activity, and immune response programmes. 

Importantly, these sepsis immune states didn't match to the clinical stage of sepsis. For example, the 'early' immune state (STImS1) was not the same as the early clinical stage of sepsis, which is often the day a clinician diagnoses sepsis. 

The researchers say these findings could have important implications when considering how best to treat sepsis patients – specifically with what treatments and when. 

The research was carried out by Dr Matthew Fish during his PhD at King's College London in the laboratory of Professor Manu Shankar-Hari. His work was supported by a National Institute of Academic Anesthesia BJA/RCoA fellowship. Dr Fish is now a postdoctoral researcher at Massachusetts General Hospital where he continues to study sepsis and the immune system. 

The study also brought together researchers from King's College London, including Dr Chad Swanson, Reader in the School of Immunology & Microbial Sciences, and Professor Mervyn Singer, Professor of Intensive Care Medicine at University College London. 

The main aim of this work was to build a profile of sepsis immune responses over time. When people are admitted to hospital with sepsis, they are usually classed as having 'early' sepsis – but our findings show that this isn't necessarily the case – their immune system may already be at stages of the immune response. Knowing exactly what is happening to a patient's immune system during sepsis could identify which treatments are likely to work best." 

Dr. Matthew Fish, former PhD student at King's College London and lead author of the paper

Professor Manu Shankar-Hari, Professor of Critical Care Medicine at King's College London, Honorary Consultant at Guy's and St Thomas' NHS Foundation Trust and senior author of the paper, said: "This research shows the importance of looking at the changing architecture of the immune system in sepsis over time, rather than just taking a snapshot view. We need to find better ways to treat the misfiring immune system. Only by understanding the intricacies of the immune system in all its component parts - by integrating and dynamically mapping cell and molecular immunobiology to determine why the very system designed to protect us from infections is misfiring in sepsis - can we begin to improve outcomes for patients by treating the misfiring immune system." 

The authors say the next step for this work is to understand what causes changes in the immune system between onset of infection to the development of sepsis. Understanding these changes could help identify new treatment targets and approaches to reduce either the progression to sepsis or the severity of sepsis, with the potential to improve outcomes and make a difference to lives of millions affected by serious infections worldwide.