Fig. 1

Overview of scDC workflow. This illustrates the main components of the scDC procedure. The key functions have been included on top of each sub-figure, where relevant. a Single cell data is collected and analysed. Publicly available experimental data is obtained. Simulated data is generated using R package PowSimR. b This corresponds to step 1 and 2 of the 4-step procedure. Resampled data is generated using stratified bootstrap and then clustered using clustering algorithm. Each cluster is matched to cell-type using reference cell labels in the original dataset. This step is repeated n times. c In step 3 and 4, cell count output from each bootstrap is fitted using GLM. The coefficient estimates from each individual GLM model are pooled using Rubin’s rules and tested for significance. d User can extract the overall estimates of statistics. e Each bootstrap re-sampling gives an estimated distribution of cell-types composition for each patient. f the result can be visualised graphically