Extended Data Fig. 1: QC of the scRNA-Seq clustering analysis.

a, Experimental procedure: optic tectum and torus longitudinalis (TL) were dissected from 6-7 dpf WT larvae in eleven experimental batches. The cells were dissociated and single-cell barcoded cDNA was generated using the 10x Chromium system and sequenced. Cells were clustered into t-types according to the similarity of gene expression. b, After batch correction using Harmony⁶³ and removal of immediate early genes (see Methods), we identified a total of 25 clusters. Each dot represents a single cell, colour coded according to the cluster. c, Same UMAP as in b, colour coded according to the batch. Clusters include cells from all batches. d, Same UMAP as in b, colour coded according to the larvae age. Clusters include cells from both 6 and 7 dpf larvae. e, QC metric of the sequenced cells after filtration, according to cluster. Cells were filtered (see Methods) in order to remove outliers that contain either low or high number of genes and UMI (representing either poorly sequenced cells or doublets), as well as cells containing high percentage of mitochondrial genes (representing stressed cells). Different cell types contained different levels of genes and UMIs, matching their biological profile (for example high levels of UMIs in some of the progenitor cells). f, The similarity between clusters was assessed by correlating the specificity scores of the top markers associated with each cluster (see Methods). The vast majority of clusters showed low correlation with other clusters, representing unique t-types. g, The robustness of the clusters was assessed by measuring the overlap of cells grouped together under different clustering parameters in comparison to the parameters used (see Methods). h, Jaccard raincloud plot showing cluster stability. 80% of the data were subsampled 100 times, with pairwise Jaccard indices calculated for each cluster before and after subsampling and reclustering (see Methods). In most instances, we observed that clusters with small numbers of cells were scored low, as a result of infrequent sampling of those cells. Box plots include all data points. The lower and upper hinges represent the first and third quartiles, respectively. Whiskers extend to the largest and smallest values within 1.5 × IQR from each hinge.