Extended Data Fig. 6: Glial diversity. | Nature

Extended Data Fig. 6: Glial diversity.

From: Molecular architecture of the developing mouse brain

Extended Data Fig. 6

a, Cells from clusters 172 to 320 were pulled from the complete dataset, and a t-SNE embedding was calculated. The dendrogram was cut into 15 metaclusters. Cells are coloured by their original cluster (left) and metacluster cut from the dendrogram (right). b, Cells are coloured by their labels for differential gene-expression testing: brown, glioblast; coral, neurogenic; grey, not included in testing. c, Select genes are annotated on a volcano plot illustrating differential expression between neurogenic and gliogenic radial glia. The dotted line denotes the significance threshold (q = 10−5). The range of the x axis was chosen to capture statistically significant genes. Cell-cycle genes are coloured red. d, For each metacluster, the size of the bar indicates the percentage of cells from each tissue (top), each embryonic stage (middle) or that are cycling (>1% cell-cycle UMIs) (bottom). e, Expression dot plots are shown for select genes. The size of the dot represents the fraction of cells expressing a gene. The colour represents normalized mean expression, in which the mean expression for each gene is normalized by the maximum mean expression of the gene across the metaclusters. f, Cells selected for region classification are coloured on the t-SNE. Grey cells were not selected. g, For a range of parameter values (max_features), a gradient boosting classifier was fit to each training set and tested on the five remaining training sets, coloured as in f. Average precision, recall and F1 scores are plotted for each parameter value and training set. h, F1 scores for each gradient boosting classifier (max_features = 50) are shown for forebrain, midbrain and hindbrain cells. The colour of each line indicates which cells were used to train the classifier (coloured as in g). i, Feature importance for each gradient boosting classifier was used to rank genes, and the intersection of the 100 most important genes for all 6 classifiers is shown on the x axis. Expression dot plots are shown for each of these genes in radial glia, glioblasts and OPCs from the forebrain, midbrain and hindbrain . j, Cells are coloured by log-transformed gene expression for the indicated genes. Grey indicates no expression. k, Cells that are positive for both Egfr and Dll1 are coloured blue.

Back to article page