Fig. 1

Framework of Sensei. a–f show side-by-side the way Sensei (right) models a controlled clinical study (left). a A controlled study involves a control group and a case group for ascertaining the difference in the proportions of T cells between the two groups. b Sensei models the true biological between-group difference and within-group variance using beta distributions. Correlation is also modeled for matched pairs study design. c A biopsy is extracted from each participant and assayed by a single-cell technology. Cell types are identified in silico. d Sensei models technical variations introduced by limited cell number using a binomial distribution (with other technical variations already accounted for in b). e The t-test is performed to identify statistically significant differences. f Sensei infers the distribution of the t-statistics and calculate the false negative (type II error) rates. g A sample input for Sensei. Required are sample sizes, cell numbers, estimated proportions of the cell type and false positive rate (type I error) rate for t-test. h A sample output of Sensei, corresponding to (g). Tabulated are false negative rates for each feasible sample size