Fig. 1

Schematic overview of scAbsolute approach. Initially, raw read counts are binned and segmented. The marginal distribution of the segmented read counts is fitted with a constrained Dirichlet process Gaussian mixture model (DPGMM) using stochastic variational inference in order to identify a limited set of plausible solutions. We do this by identifying a constant \(\rho\) that converts the scale from a read count to an absolute copy number scale. There exist a set of values of \(\rho\) that lead to equally possible ploidy solutions (in this case hypotriploid or hypohexaploid solutions). We select the solution with minimum ploidy value subject to per-cell ploidy constraints. Post hoc analysis of ploidy allows to distinguish previously indistinguishable cell populations based on read density, overcoming the limitation of the above approach. The example shows two (exemplary) copy number profiles for cells in different ploidy states, that have previously been indistinguishable based on copy number profiles alone. We demonstrate that cell ploidy can be determined in the absence of differentiating copy number states and other experimental information, based on per-cell read density alone. The iterative approach makes it possible to improve individual predictions by using the information from related cells to restrict the ploidy window for a given set of cells