ArchVelo: archetypal velocity modeling for single-cell multi-omic trajectories

Area of Science:

• Genomics and Computational Biology
• Single-cell Multi-omics Analysis
• Systems Biology

Background:

• Inferring dynamic cellular processes from static single-cell genomics data is a significant challenge.
• Understanding gene regulation and cell fate decisions requires integrating different molecular layers.
• Existing computational methods often struggle with accuracy in trajectory inference and latent time alignment.

Purpose of the Study:

• To introduce ArchVelo, a novel computational framework for modeling gene regulation and inferring cellular trajectories.
• To leverage paired single-cell chromatin accessibility (scATAC-seq) and transcriptomic (scRNA-seq) data for dynamic modeling.
• To identify regulatory programs (archetypes) and their influence on transcription and cell differentiation.

Main Methods:

• Developed ArchVelo, a framework representing chromatin accessibility as archetypes (shared regulatory programs).
• Modeled the dynamic influence of archetypes on gene transcription.
• Benchmarked ArchVelo on mouse brain and human hematopoiesis datasets, comparing trajectory inference accuracy and latent time alignment.

Main Results:

• ArchVelo demonstrated superior performance in trajectory inference accuracy and gene-level latent time alignment compared to existing methods.
• The framework enabled trajectory decomposition into archetypal components and identification of underlying transcription factors.
• Applied to CD8 T cells in viral infection, ArchVelo revealed distinct differentiation and proliferation trajectories, including progenitor exhausted CD8 T cells.

Conclusions:

• ArchVelo provides a principled and accurate framework for modeling dynamic gene regulation from multi-omic single-cell data.
• The method successfully infers cellular trajectories and regulatory mechanisms, particularly in immune cell differentiation.
• ArchVelo offers valuable insights into sustained immunity and immunotherapy response by characterizing progenitor exhausted CD8 T cell differentiation.