Aim

Macrophages in atherosclerotic plaques constantly shift between functional states, but how these transitions unfold in human disease is still poorly understood. We built a single-cell atlas of human plaque macrophages from seven independent cohorts and developed an optimal transport (OT) based method to characterise directed cell-state transitions. 

Integration of scRNA-seq data from seven cohorts; Sinkhorn optimal transport combined with RNA velocity (UniTVelo) to infer directed transitions; transcription factor activity (decoupleR), driver gene analysis (CellRank) and leave-one-cohort-out validation.

We identified six macrophage meta-clusters connected by a network of directed transitions along three axes: monocyte diversification, inflammatory reactivation and routes to fibrosis. Tissue-resident macrophages acquire inflammatory programs while preserving their original identity (transcriptional layering), whereas monocyte-derived cells undergo selective reprogramming.

Building on the atlas, we are developing deep learning sequence-to-function models conditioned on continuous OT-derived cell-state coordinates. The goal is to predict how genes behave along macrophage transitions and identify the regulatory sequence features driving them. Next, we plan to apply the framework to tumour-associated macrophages.