HyCell Copyright (C) 2026 Tencent. Licensed under the Apache License, Version 2.0; see the LICENSE file at the root of this repository. This folder contains or was derived from the following third-party data, each of which remains subject to its original license terms. 1. CZ CELLxGENE Census (release of 8 November 2025) The HyCell weights were trained on the human primary cells of this release, and the gene vocabulary in src/hycell/assets/genes.tsv.gz (Ensembl gene ids and gene symbols) is taken from its gene metadata. Data in the Census are available for re-use under the Creative Commons Attribution 4.0 International license (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). CZI Cell Science Program et al., "CZ CELLxGENE Discover: a single-cell data platform for scalable exploration, analysis and modeling of aggregated data", Nucleic Acids Research 53(D1), D886-D900 (2025), https://doi.org/10.1093/nar/gkae1142. Source: https://cellxgene.cziscience.com/ 2. scIB integration-task datasets, human lung atlas tests/data/lung_512.h5ad holds 512 cells of this atlas (128 randomly chosen cells from each of four Drop-seq batches), with their raw counts, cell-type labels and batch labels; genes are given by Ensembl id and symbol. The cells and counts are otherwise unchanged. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). M. D. Luecken et al., "Benchmarking atlas-level data integration in single-cell genomics - integration task datasets", figshare (2022), https://doi.org/10.6084/m9.figshare.12420968. M. D. Luecken et al., "Benchmarking atlas-level data integration in single-cell genomics", Nature Methods 19, 41-50 (2022), https://doi.org/10.1038/s41592-021-01336-8.