Fig. 8

Benchmarking SCADIE on FL bulk dataset with poor initialization and limited sample size. a Volcano plots for known DEGs from CIBERSORTx, TOAST, SCADIE’s initial Ws and final Ws, for the latter two cases the standard errors were obtained through jackknife. b Number of correctly identified DEG (p-value cutoff 0.01) from CIBERSORTx HiRes, TOAST, and SCADIE. For SCADIE’s initial W and final W cases, their point estimates were different Ws but the same set of standard deviation was used in the hypothesis tests