Towards a large-scale human inflammation atlas at single-cell resolution

[eng] Inflammation is a physiological response essential for maintaining homeostasis, but when the immune system becomes dysregulated, it can drive a range of pathological conditions and become a key contributor to many diseases. Understanding these complex, disease-specific inflammatory mechanisms is crucial for developing effective treatments. Single-cell sequencing technologies offer a powerful approach to study the immune system at high resolution, providing detailed insights into cellular heterogeneity and immune dynamics. To fully exploit their potential in large-scale studies, standardized protocols for sample collection and processing, coupled with advanced computational methods for robust data analysis, are needed. Since immune-driven diseases often target specific tissues and exhibit distinct inflammatory profiles, a comprehensive approach is required. Blood, as a minimally invasive sample, offers a valuable opportunity to monitor inflammation by tracking cytokine levels and cellular activity, making it ideal for elucidating disease mechanisms and therapeutic responses. In this thesis, experimental protocols and computational approaches were developed for large-scale studies focused on analyzing human inflammatory profiles using single-cell RNA sequencing (scRNA-seq). First, we implemented FixNCut, a protocol that preserves transcriptional profiles and tissue composition by fixing samples at the time of collection, enabling subsequent tissue dissociation with minimal stress-induced artifacts. We then generated a comprehensive reference of the inflammatory spectrum in circulating immune cells using single-cell transcriptomic profiles from over 1,000 patients suffering from 20 immune-driven diseases and healthy individuals. By analyzing the immune cell composition and inflammatory molecule expression, we characterized the immune response across diseases, identifying both commonalities and disease-specific differences. Using interpretable machine learning algorithms, we captured disease-specific inflammatory genes at the cell type level that could serve as clinical biomarkers. Furthermore, we proposed a framework for classifying patients based on their blood-derived inflammatory profiles. Finally, we studied the immunomodulatory effects of albumin treatment in patients with acute decompensated cirrhosis, which resulted in immune transcriptional reprogramming that primarily affected B lymphocytes and enhanced neutrophil antimicrobial activity, boosting the patients' immune response against infections. In summary, this thesis integrates scRNA-seq experimental and computational methods to facilitate large-scale, decentralized research on immune-driven diseases with the goal of applying these approaches for disease diagnosis, monitoring progression, and assessing therapy responses in clinical settings.