Fig. 1

Overview of lute framework for deconvolution in Bioconductor. A Schematic of a deconvolution experiment using lute. Inputs include (far left) matrix/flat tabular format, (second from left) SummarizedExperiment [29], (second from right) SingleCellExperiment [29, 30], and (far right) cellScaleFactors. B The lute framework showing (top) terms (defined in Results 2.1) for deconvolution and pseudobulk operation, including \(Y\) (bulk), \(P\) (proportions), \(S\) (cell type sizes), and \(Z\) (cell type reference), noting k (number of cell types) and G (marker genes) with arrows indicating applicable input classes and scaling factors corresponding to differences in cell sizes across cell types available in the cellScaleFactors (hexagon) R/Bioconductor package [19], (middle) the typemarkers() function to select marker genes from the reference and bulk, and (bottom) the deconvolution() generic for calling the user’s choice of the deconvolution method. C Schematic illustrating supported deconvolution algorithms in lute including a parent class (a.k.a. “deconvoParam”), referencebasedParam, independentbulkParam, nnlsParam, musicParam, epicParam, deconrnaseqParam, bisqueParam, and music2Param