Pseudo-bulk knock-out profiles from the in vivo whole-brain Perturb-seq atlas of Shi et al. (2026), used in RFind-sc
Year: 2026
Type: dataset
Abstract: For each of five neuronal cell types, the UMI counts of the nuclei of each column (a knock-out, a safe-targeting control or the non-targeting control) of the in vivo whole-brain Perturb-seq atlas of Shi et al. (2026; Hugging Face perturbai/wholebrain_crispr_atlas, commit a7c65dc0a64da4bd47cf6ef5f4dec6c7ef745e87, CC BY 4.0) were summed gene by gene. These profiles are the knock-out references of Fig. 4, Supplementary Fig. S3 and Supplementary Tables 1 and 2 of "RFind-sc maps single cells onto biological axes defined by expression signatures and genetic perturbations" (Murano and Miyakawa). Files: pb_g017_DG.npz (DG Glut), pb_CA1.npz (CA1 Glut), pb_g017_CA3.npz (CA3 Glut), pb_GABA3.npz (GABAergic neurons: Pvalb, Sst and Lamp5 summed) and pb_g151.npz (151 TH Prkcd Grin2c Glut). Each NumPy .npz file (numpy.load(path, allow_pickle=False)) holds counts (19,070 genes x 2,046-2,048 columns, float32), genes (gene symbols), targets (the targeted gene or control of each column) and n_cells (the number of nuclei in each column). The code that makes and uses these files is at https://github.com/tmurano/RFind-sc.
Authors: Murano, Tomoyuki, Miyakawa, Tsuyoshi
Institution: Fujita Health University
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7222583415