Methods

To compare gene expression landscapes between species, we followed the same pipeline as in Telley et al., 2019, using published dataset from Nowakowski et al., 2017 (embryonic human cortex at 10, 14, 18 and 22 post-conception weeks) and Pollen et al., 2019 (cerebral human organoids at 3, 5, 8, and 10 weeks in vitro).

Note that while in Telley et al., 2019, mouse cortical neurons directly born from apical progenitors at specific embryonic stages were followed during their differentiation thanks to FlashTag labeling, the other datasets presented here comprise all cortical cell populations at each developmental stage.

For each dataset, a pseudo-age and a pseudo-differentiation scores were calculated as follow. Ordinal regression (bmrm R package) was performed using seed cells to build a pseudo-age / pseudo-differentiation model (see Telley et al., 2019 for details). The pseudo-age and pseudo-differentiation scores were then calculated as the predictions from these models for all the other cells of the datasets.

Embryonic human cortex

Cell types were defined as follow:

Cell types	Subtypes annotated by the authors
aRG	vRG
BP	IPC-new, oRG
iN	IPC/newborn, Newborn
mN	Deep layer, Layer IV, Layer VI Pan-area, Parietal and Temporal, Upper Layer

Models were built as follow:

• pseudo-age: 100 cells / embryonic stage (n = 25 aRG, 25 BP, 25 iN, 25 mN per embryonic stage)
• pseudo-differentiation: 200 cells / type

Model classes	n cells per embryonic stage
aRG	50
BP	25 oRG & 25 IPC
N	25 iN & 25 mN

Prediction scores: (n = 150 random cells per type and embryonic stage)

Cerebral human organoids

Cell types were defined as follow (low quality annotated cell subtype was excluded):

Models were built using n = 50 cells per in vitro stage and per cell type.

Prediction scores: (n = 200 random cells per type and in vitro stage - for IPC: n = 100 at 3, 150 at 5, 250 at 8 and 300 at 10 weeks in vitro)