figshare
Browse
Schematic representation of the parameters for the power simulation using powsim.pdf (264.95 kB)

Schematic representation of the parameters for the power simulation using powsim

Download (264.95 kB)
Version 3 2018-06-07, 06:59
Version 2 2018-06-07, 06:55
Version 1 2018-06-07, 06:47
journal contribution
posted on 2018-06-07, 06:59 authored by RIKEN BiTRIKEN BiT
We think that the power simulation should reflect the measured data of each single-cell RNA-seq method; therefore, the user should enter each appropriate parameter for powsim. We empirically used a number of detected genes based on measured data (including Quartz-Seq2 data and reported single-cell RNA-seq data as the “ngenes parameter” in each powsimR simulation. We already showed that a number of detected genes greatly differed by method. For example, the detected gene count differed greatly between Quartz-Seq2 and Drop-seq (Quartz-Seq2: 6,322 ± 144 genes; Drop-seq: 2,574 genes). Moreover, the number of differentially expressed genes strongly depended on the number of detected genes. In particular, the DESetup object for the simulateDE function had an ngenes parameter in PowsimR. For each method, we indicated a mean number of detected genes for the ngenes parameter of the DESetup object. We had already mentioned this in Quartz-Seq2 paper. In Ziegenhain et al. (2017), the ngenes parameter might be fixed at the same number (i.e., 13,361 genes) for all methods. The parameter presumes that the gene-detection capability of each method is the same; however, in reality, the number of detected genes differs by method. The fixed number of ngenes parameter is very different from the measured single-cell RNA-sequencing data; therefore, we used the number of detected genes based on the measured data as the “ngenes parameter” and calculated the number of true positive differentially expressed genes. The parameter (based on measured data) makes power simulation more meaningful.

History