Feature selection for preserving biological trajectories in single-cell data.

Area of Science:

• Single-cell biology
• Computational biology
• Genomics

Background:

• Single-cell technologies measure thousands of molecular features during dynamic biological processes.
• Identifying key features driving continuous biological processes from noisy data is challenging.
• Confounding biological variation and irrelevant features can obscure cellular trajectories.

Approach:

• Introduces DELVE (dynamic selection of locally covarying features), an unsupervised feature selection method.
• DELVE uses a bottom-up approach to mitigate unwanted variation and model cell states.
• Identifies representative subsets of dynamically expressed molecular features that recapitulate cellular trajectories.

Key Points:

• DELVE mitigates confounding variation by modeling cell states from dynamic feature modules.
• Demonstrated improved characterization of cell populations and cell type transitions.
• Successfully applied to simulations, single-cell RNA sequencing, and immunofluorescence imaging data.

Conclusions:

• DELVE offers an alternative framework for enhancing trajectory inference in single-cell studies.
• Improves the discovery of co-varying molecular features along biological trajectories.
• The DELVE method is available as an open-source Python package.