Single-cell RNA sequencing technologies have evolved rapidly over the past few years, but the newest protocols and platforms are still limited by bias and noise. These technical issues can present serious challenges to downstream data analyses when samples are collected from multiple human donors recruited from multiple sites. My dissertation outlines methods for assessing bias and quantifying reproducibility in single-cell RNA sequencing studies. These tools are applicable to a new class of single-cell studies of human disease that move beyond tissue-level case–control comparisons. I have implemented these methods in two software packages, scone and scRAD, both developed to improve the quality of biological insights derived from single-cell RNA sequencing data.