Figure 5.

Analysis of hESC-derived cell types by scRNA-seq and logistic regression

(A–C) UMAP projection of hESC-derived cells after quality filtering showing cells colored by their Louvain cluster’s membership (A), day of differentiation and analysis (B), and their log-library size normalized expression of STMN2 (C).

(D and E) Wheel plot showing hESC-derived midbrain cell types (dots) generated by the protocol developed in this study (D) or the protocol by Kriks et al. (2011) (E), compared by logistic regression to endogenous standards (wheel): human embryonic ventral midbrain cell types from La Manno et al. (2016) and vascular leptomeningeal cells from Marques et al. (2018).

(F) Heatmap showing genes with highest coefficients from logistic regression for progenitor midline (ProgM), progenitor floor plate (ProgFP), dopaminergic neurons (DA), andoculomotor and trochlear nucleus (OMTNs). Log-library size normalized gene expression of top-similar hESCs to in vivo counterparts derived from current protocol and Kriks et al. (2011) are shown. Permutation test for each reference cell type was performed with H1: The sum of average expression of genes in the heatmap. Permutation was performed 1,000 times on normalized counts, p values are shown.

(G) Track plot showing log-library size normalized gene expression of selected genes of cells at days 21 and 28 of differentiation (top). Heatmap showing similarities between differentiated cells and reference endogenous cell types from La Manno et al. (2016) and Marques et al. (2018), scoring using logistic regression (bottom).

(H) Dot plot showing clusters from days 21 and 28 of differentiation with average similarity scores, as determined by logistic regression, between 0.5 and 0.79 (orange) or between 0.8 and 1.0 (purple). Only clusters with similarity scores >0.5 are shown.

(I) Violin plot showing genes enriched in cluster 24 and 35. Log-library size normalized gene expression is shown.