Abstract
Inferring cellular dynamics from static single-cell data remains a central challenge in genomics. We introduce ArchVelo, a computational framework for modeling gene regulation and inferring trajectories from paired single-cell chromatin accessibility (scATAC-seq) and transcriptomic (scRNA-seq) data. ArchVelo represents chromatin accessibility as archetypes—shared regulatory programs—to model their dynamic influence on transcription. It outperforms existing methods in trajectory inference accuracy and gene-level latent time alignment, enables trajectory decomposition into archetypal components, and identifies the underlying transcription factors. After benchmarking on mouse brain and human hematopoiesis datasets, we apply ArchVelo to CD8 T cells in viral infection and reveal distinct trajectories of differentiation and proliferation. Focusing on progenitor exhausted CD8 T cells, critical for sustained immunity and immunotherapy response, we identify differentiation from Ccr6− to Ccr6+ progenitors, shared between acute and chronic infections. ArchVelo provides a principled framework for modeling dynamic gene regulation and trajectory inference in multi-omic single-cell data across biological systems.
