Comparing the transcriptome of developing native and iPS-derived mouse retinae by single cell RNA sequencing

Seurat_velocyto_monocle directory

Demultiplexing, alignment, filtering, barcode counting, UMI counting, and aggregation of multiple runs were conducted using Cell Ranger v2.1.1 (10X Genomics).  Further filtering, k-means clustering, and UMAP projection were conducted using Seurat software suite (https://satijalab.org/seurat/) 28. Differential expression analyses to identify genes that are upregulated in specific cell types when compared to all other ones (Cell type > Others) or that are differentially expressed between NaR and RO in a given cell type (NaR > RO and RO > NaR) were performed with the Findmarkers function in Seurat (https://satijalab.org/seurat/) 28. Velocity analysis was performed using the Velocyto R 15 and scVelo 14 packages. Single-cell trajectory inference and pseudotime analyses were conducted using Monocle2 (http://cole-trapnell-lab.github.io/monocle-release/) (Trapnell et al., 2014).

Dev_path_width directory

To test whether the developmental trajectories were more tightly regulated in NaR than in RO, we computed the average distance (computed as the Euclidian distance in 2D-UMAP space, i.e.√((x_1-x_2 )^2+(y_1-y_2 )^2 ) ) between 500 randomly selected NaR and 500 randomly selected RO cells and their n nearest neighbors (with n ranging from 1 to 50).  The number of cells per developmental stage was adjusted between NaR and RO by down sampling to the number of the least populated source. The corresponding calculations were performed five times. The curves shown in Figure 2D correspond to the averages across the five replicates.  The grey confidence zone in Figure 2D is bounded by the maxima and minima across the five replicates. The corresponding script was written in Perl (Dev_path_width.pl) and the graph generated in R (Path_width.R).  

Dev_stage_clique_enthropy directory

To compare cell type diversity within developmental stage between NaR and RO, we first equalized the number of cells within developmental stages between NaR and RO by randomly dropping cells from the most populated source. We then sampled two cells within cell source (NaR and RO) and developmental stage and checked whether they were from the same cell type or not. This was repeated 1,000 times yielding a measure of cell type diversity akin to (1-Entropy). Down-sampling of cells was repeated 100 times. Each data point in Figure 2E corresponds to (1-Entropy) for one such random sample. The corresponding script was written in Perl (entropy.pl) and the graph generated in R (Entropy.R).

Reactome analysis directory

Pathway enrichment analyses were conducted using the on-line Reactome analysis tools 37,38. Mouse gene identifiers were converted to human counterparts. Pathway analysis results were downloaded as flat files. A total of 392 pathways with enrichment p-value ≤ 0.01 in at least one analysis were kept and manually sorted according to Reactome hierarchy (Man_processed_reactome_output.txt). A pathway is enriched in a list of genes if it contains more components of the pathway than expected by chance (given the number of genes in the list). The overlapping genes (“found entities”) hence define the enrichment. The same pathway can be enriched in two gene lists due to the same, distinct or partially overlapping sets of “found entities”. We quantified the degree of overlap between sets of “found entities” for the 1,313 pathway enrichments using principal component (PC) analysis in a space defined by the presence/absence of 1,335 genes. The distance between sets of “found entities” in a space consisting of the 20 first PCs was projected in 3D space using t-distributed stochastic neighbor embedding (tSNE) implemented with the Rtsne R function 42. 3D tSNE coordinates were converted to hexadecimal RGB codes and used to color the sets of “found entities” (corresponding to the enrichment of a pathway in a specific gene list) when generating 2D tSNE graphs (Figure S4), or when generating a tile showing the pathways enriched in specific analyses (Cell type > Others, NaR > RO or RO > NaR) and cell types within the analysis (NE, RPE, ERPC, LRPC, NRPC, RGC, HC, AC, PRP, C, R, BP or MC) (Figure 3B). The corresponding scripts were written in Perl (Reactome_analysis.pl) and R (Reactome_analysis.R).

Sridhar_human_data directory

We reanalyzed the Sridhar et al. (Cell Reports 30: 1644-1659, 2020) using Seurat and the Sridhar_Human_Data_Seurat_Rscript script.