AutoGeneS

AutoGeneS identifies informative genes to deconvolve bulk RNA-seq and estimate cell-type proportions, thereby uncovering cellular heterogeneity without requiring prior marker genes.

Key Features:

• Automatic informative gene selection: Extracts dataset-specific informative genes rather than relying on pre-selected marker gene sets.
• Multi-criterion optimization: Simultaneously minimizes pairwise gene-expression correlation and maximizes inter–cell-type distance to improve separation of cell types.
• Robustness to noisy references: Improves deconvolution accuracy when reference profiles are noisy or when cell types are closely correlated.
• Compatibility with reference profiles: Operates with reference profiles derived from single-cell experiments and sorted cell populations.
• Implementation: Provided as a Python package implementation.

Scientific Applications:

• Cell-type deconvolution: Estimation of cell-type proportions from bulk RNA-seq by selecting genes that distinguish cell types.
• Oncology, immunology, and developmental biology: Dissecting tissue cellular composition in studies of cancer, immune responses, and development.
• Benchmarking and validation: Benchmarking deconvolution performance using empirical data, including validation on human peripheral blood with flow cytometry ground truth.

Methodology:

Automatically extract informative genes by optimizing multiple criteria—minimizing correlation and maximizing distance between cell types—using reference profiles from single-cell experiments or sorted cell populations to improve bulk RNA-seq deconvolution.