Abstract / Summary
Abstract Background Diffuse large B-cell lymphoma (DLBCL) exhibits marked prognostic heterogeneity, partly driven by interactions between regulated cell death and the tumour immune microenvironment. However, the interplay between immune status and cell death in DLBCL remains unclear. Methods This study integrates bulk and single-cell transcriptomic data to characterise immune–cell death interactions in DLBCL. Immune–cell death–related genes are identified by intersecting differentially expressed genes with immune and cell death gene sets. A cell-death–immune score (CDIS) is constructed via 101 machine-learning algorithm combinations and validated in independent cohorts. We assess immune infiltration, cell death pathway activity, cell–cell communication, ligand–target interactions, pseudotime trajectories, drug connectivity, molecular docking, and virtual gene perturbation. Results We identified 159 immune–cell death–related genes and derived a 16-gene CDIS from 25 candidates. The CoxBoost plus random survival forest (RSF) ensemble model achieved optimal performance. High CDIS was associated with suppressed immunogenic cell death programs and an immune-cold microenvironment, characterised by reduced effector immune infiltration and increased M2-like macrophages. Single-cell analysis revealed that model genes were expressed predominantly in macrophages and identified M2-like macrophage-centred pro-tumour survival signalling involving the BAFF/APRIL axis. Conclusions CDIS offers a robust prognostic framework for DLBCL and uncovers immune escape mechanisms linked to macrophage-driven signalling and dysregulated cell death.