Interpretable Differential Abundance Signature (iDAS).

Single-cell technologies have revolutionized the understanding of cellular dynamics by allowing researchers to investigate individual cell responses under various conditions, such as comparing diseased versus healthy states. Many differential abundance methods have been developed in this field, however, the understanding of the gene signatures obtained from those methods is often incomplete, requiring the integration of cell type information and other biological factors to yield interpretable and meaningful results. To better interpret the gene signatures generated in the differential abundance analysis, iDAS is developed to classify the gene signatures into multiple categories. When applied to melanoma single-cell data with multiple cell states and treatment phenotypes, iDAS identified cell state- and treatment phenotype-specific gene signatures, as well as interaction effect-related gene signatures with meaningful biological interpretations. The iDAS model is further applied to a longitudinal study and spatially resolved omics data to demonstrate its versatility in different analytical contexts. These results demonstrate that the iDAS framework can effectively identify robust, cell-state specific gene signatures and is versatile enough to accommodate various study designs, including multi-factor longitudinal and spatially resolved data.