Performance of MCMC estimation using the ensemble model and prior, in simulation of multi-lab experimental design in S3 Table.

The experiment has a total of 194 million to 289 million cells from each of five participants, distributed among four labs, encompassing 474 to 569 wells per participant. Bias and absolute error represent over- or underestimation of effect sizes, as difference in log10 of fold-change. For example, typical batch effects are estimated to be 10−0.202 = 63% of the simulated truth. See Methods (“Validation by simulation: Multiple labs”) for simulation details.