Systematic errors, including doublet, mapping bias or batch effect, confound downstream analysis of single-cell transcriptome and thus should be removed prior to analysis (Zheng et al., 2017). Doublet frequency in our scRNA-seq were evaluated by counting cells expressing TBR1 or GFAP, which were exclusively expressed in cortical excitatory neuron (referred to as cortical neuron, CN) and astrocyte, respectively (Figure S3C). Mapping ratio and total UMIs were calculated in individual cells, and their biases among clusters were also compared (Figure S4A). Batch effect after Seurat normalization was evaluated between two biological replicates. We calculated Euclidean distances of data point pairs within one replicate or between two replicates. 1,000 pairs were randomly selected from the within group and the between group, and their distances were compared by Wilcoxson rank-sum test. Batch effect after Seurat normalization was evaluated between two biological replicates. We calculated Euclidean distances of data point pairs within one replicate or between two replicates. 1,000 pairs were randomly selected from the within group and the between group, and their distances were compared by Wilcoxson rank-sum test (Figure S3B).
