DELVE

DELVE identifies representative molecular features from single-cell RNA sequencing and iterative immunofluorescence imaging to robustly capture cellular trajectories by modeling dynamic gene and protein modules associated with core regulatory complexes using evolutionary algorithms.

Key Features:

• Unsupervised feature selection: Selects a representative subset of molecular features without supervised labels to capture biological variation.
• Bottom-up strategy: Mitigates confounding sources of variation by building feature sets from dynamic modules rather than individual markers.
• Module-based state modeling: Models cell states using dynamic gene and protein modules associated with core regulatory complexes.
• Evolutionary algorithms: Leverages evolutionary algorithms (as implied by the name Dynamic Expression Landmarking via Evolutionary algorithms) for feature selection.
• Trajectory preservation: Prioritizes features that preserve biological trajectories within single-cell datasets.
• Validation inputs: Uses simulations, single-cell RNA sequencing data, and iterative immunofluorescence imaging data for evaluation and validation.

Scientific Applications:

• Cell cycle analysis: Identifies dynamic features that define and order cell cycle states in single-cell data.
• Cellular differentiation studies: Selects features that capture transitions during cellular differentiation.
• Cell type and transition definition: Improves identification of cell types and transitional states by focusing on dynamic gene/protein modules.

Methodology:

Unsupervised feature selection using evolutionary algorithms with a bottom-up, module-based model of cell states, evaluated using simulations, single-cell RNA sequencing, and iterative immunofluorescence imaging.