Purpose of AUCell scoring
AUCell uses within-cell feature rankings to examine whether a selected gene set is enriched near the top of a cell’s ranking. The notes explore whether this approach can help characterize regulatory activity and cell populations in scATAC-seq data. Two motivations are:
- Program activity: examine patterns associated with selected regulatory regions or gene sets.
- Cellular heterogeneity: compare scores across cells to identify groups with different activity patterns or states. The AUCell documentation describes the implemented gene-expression workflow. An accessibility-based adaptation needs an explicit definition of how regions become the features used for scoring.
Preparing gene sets or regulatory regions
First, define the biological sets of interest. For scATAC-seq, the notes consider genomic regions such as enhancers or promoters linked to genes of interest. The region coordinates and the feature identifiers need to be compatible with the accessibility data.
Connecting peaks to genes
Peak annotations can connect regions to nearby genes or to selected regulatory elements. This mapping is a distinct step: its genomic coordinates, annotation reference, and mapping rules affect which features enter each set.
Presence and absence versus ranking values
The notes consider a binary representation that records whether a region or set is present in a cell. That presence/absence representation must be distinguished from the values used to rank features. For scoring, each cell needs a feature ranking derived from an explicitly chosen signal, such as read counts, accessibility values, or a derived activity matrix. A binary matrix has many ties and does not contain the same information as a ranking based on signal intensity.
Calculating a ranking-based score
Within a cell, features are ordered by the chosen signal. The analysis then follows where members of the selected set appear near the top of that ordered list and computes an area-under-the-curve score. In the conceptual plot:
- The horizontal axis follows the ranked features.
- The vertical axis follows the cumulative recovery of set members. The AUC summarizes how concentrated those members are near the top of the ranking. A score alone is not a significance test or an experimentally validated cell-type assignment.
Comparing cells and conditions
Scores can then be examined across cell populations, conditions, or other cellular measurements. Comparability depends on consistent feature preparation and scoring settings. This page records an exploratory adaptation. Before using it as an scATAC-seq pipeline, specify the region-to-gene representation, ranking input, gene-set definition, and interpretation or statistical tests for the chosen dataset.