Batch effects and the effective design of single-cell gene expression studies

Single cell RNA sequencing (scRNA-seq) can be used to characterize variation in gene expression levels at high resolution. However, the sources of experimental noise in scRNA-seq are not yet well understood. We investigated the technical variation associated with sample processing using the single cell Fluidigm C1 platform. To do so, we processed three C1 replicates from three human induced pluripotent stem cell (iPSC) lines. We added unique molecular identifiers (UMIs) to all samples, to account for amplification bias. We found that the major source of variation in the gene expression data was driven by genotype, but we also observed substantial variation between the technical replicates. We observed that the conversion of reads to molecules using the UMIs was impacted by both biological and technical variation, indicating that UMI counts are not an unbiased estimator of gene expression levels. Based on our results, we suggest a framework for effective scRNA-seq studies. We collected single cell RNA-seq (scRNA-seq) data from three YRI iPSC lines using the Fluidigm C1 microfluidic system followed by sequencing. We added ERCC spike-in controls to each sample, and used 5-bp random sequence UMIs to allow for the direct quantification of mRNA molecule numbers. For each of the YRI lines, we performed three independent C1 collections; each replicate was accompanied by processing of a matching bulk sample using the same reagents. This study design allows us to estimate error and variability associated with the technical processing of the samples, independently from the biological variation across single cells of different individuals. We were also able to estimate how well scRNA-seq data can recapitulate the RNA-seq results from population bulk samples. We combined the 96 single cell samples from each C1 chip into their own master mix and sequenced across three lanes of a HiSeq 2500 (3 individuals x 3 replicates x 96 wells x 3 lanes = 2592 files). We prepared two separate library preparations for each bulk sample, combined them all into one master mix, and sequenced across four lanes (3 individuals x 3 replicates x 2 library preparations x 4 lanes = 72 files).