Using machine learning and multi-omic analysis to understand the hyperinflammatory response that leads to haemophagocytic lymphohistiocytosis (HLH).

Background: Normal immune responses are critical to survival, but hyperinflammatory responses are harmful, and can lead to cytokine storm, multi-organ failure and death. Examples of hyperinflammation, or cytokine storm syndromes, include the hyperinflammatory pneumonia caused by COVID-19, the cytokine release syndrome related to CAR-T cell therapy, and the multisystem syndrome haemophagocytic lymphohistiocytosis (HLH). Secondary HLH is a prototypic hyperinflammatory syndrome caused by malignancies, rheumatological conditions, and infections including SARS-CoV-2 [1, 2]. It has a mortality of ~50%, increasing to 80% in people with lymphoma. Management of HLH involves treatment of the cytokine storm alongside identification and treatment of the pathogenic driver.
This syndrome is poorly understood, and a perceived rarity has hampered research. We do not understand why some people mount an exaggerated and harmful immune response to these stimuli and others do not.
Objective: To improve patient outcomes, we need to develop a strong evidence base. This has been lacking to date because of a lack of concentration of expertise. The UCLH HLH Service was set up to address this and we have seen 20 patients in the first 6 months of 2022, compared to approximately 2-4 patients per year prior to the establishment of this service. Patient samples are recruited to a biobank and a matched clinical database (recruitment of 5-10 patients per month ongoing via UCLH biobank). We plan to harness this unique resource to develop a mechanistic model to improve our understanding of hyperinflammatory responses. This project will use immunological and multiple 'omics (multi-omics) analysis approaches to characterise hyperinflammation phenotypes and heterogeneity for improved understanding of this aberrant immune response.
Aims:
1: Curate literature to identify relevant pathogenic pathways and design multi-omic analysis approach.
2: Interrogate multi-omic [immune phenotype (spectral cytometry), targeted transcriptomics/proteomics and metabolomics] and rich clinical data to establish signatures associated with HLH disease using already established analysis pipelines[3-10],
models developed during the rotation project and knowledge gained from attending a 'Datathon' focused on using clinical/registry data for patient stratification.
3: Establish and identify HLH disease endotypes and heterogeneity according to multi-omic, disease and patient outcome signatures.
Outcome: First study to define HLH heterogeneity, identify potential HLH disease endotypes and improve understanding of HLH disease mechanisms.