prism-oncology/ALHENA-references

ALHENA references

The prebuilt single-cell references of ALHENA, an H&E encoder distilled from transcriptomic foundation models for multi-scale slide analysis. ALHENA annotates the cells of an H&E section by retrieval against one of these references.

Package code on GitHub | Paper code on GitHub | Documentation | Model weights

What a reference holds

Each file under shards/ holds one 512-dimensional projection per atlas cell, in ALHENA's cell space, and the cell type the ALHENA mapping assigns to that cell, on the annotation axes lineage_v1 and compartment_v1. Apart from metadata, such as its identity and the number of cells of each class, it holds nothing else: no expression value, gene name, cell barcode or source annotation. registry.json lists the files with their SHA-256 digests, cell counts, class support, tissues and sources.

The references

Fourteen references serve fifteen tissues: tonsil serves both LymphNode and Tonsil.

Reference	Tissues	Cells	Size	Source
bone_marrow	BoneMarrow	421,775	433 MB	DISCO
breast	Breast	174,638	179 MB	DISCO
cervix	Cervix	77,518	80 MB	Gene Expression Omnibus, GSE208653
heart	Heart	624,358	641 MB	DISCO
intestine	Colon	414,406	425 MB	DISCO
kidney	Kidney	103,547	106 MB	DISCO
liver_cell	Liver	367,596	377 MB	DISCO
lung	Lung	221,012	227 MB	DISCO
ovary	Ovary	124,128	127 MB	DISCO
pancreas_cell	Pancreas	188,369	193 MB	DISCO
prostate	Prostate	21,030	22 MB	Tabula Sapiens, through CELLxGENE
skin	Skin	394,594	405 MB	DISCO
tonsil	LymphNode, Tonsil	273,755	281 MB	DISCO
uterus	Uterus	22,029	23 MB	Tabula Sapiens, through CELLxGENE

How ALHENA uses them

ALHENA downloads a reference when it is first needed, together with registry.json, and checks it against the digest the registry declares. Only the reference of the tissue named is downloaded.

pip install alhena
alhena download --tissue Breast

import alhena

typer = alhena.load_celltyper(tissue="Breast")      # a cell typer on the breast reference
reference = alhena.load_reference(tissue="Breast")  # the reference alone

See References in the documentation.

Sources

References	Source	Terms of the source	Credit
bone_marrow, breast, heart, intestine, kidney, liver_cell, lung, ovary, pancreas_cell, skin, tonsil	DISCO	"All data and scripts in DISCO are free for academic research."	Li, M. et al. Nucleic Acids Res. 50, D596–D602 (2022), and the original studies
prostate, uterus	Tabula Sapiens, through CELLxGENE	CC BY 4.0	Tabula Sapiens Consortium et al. Science 376, eabl4896 (2022)
cervix	Gene Expression Omnibus, GSE208653	No restriction stated by GEO	Guo, C. et al. Clin. Transl. Med. 13, e1219 (2023)

License

The precomputed single-cell reference representations are released under the CC BY-NC 4.0 license. They remain subject to the terms of their source atlases, and those derived from DISCO are provided for academic research only, in accordance with the DISCO terms of use. No expression values, gene names or cell barcodes from the source atlases are redistributed.

The full text of the CC BY-NC 4.0 license is in LICENSE.

Citation

[TODO: add link to preprint]

Attribution under CC BY-NC 4.0 is given by citing the ALHENA preprint:

@article{alhena,
  title   = {{ALHENA}, an {H\&E} encoder distilled from transcriptomic foundation models
             for multi-scale slide analysis},
  author  = {Giraud-Sauveur, F{\'e}licie and Blampey, Quentin and Marza, Pierre and
             Benkirane, Hakim and Vakalopoulou, Maria and Courn{\`e}de, Paul-Henry and
             Christodoulidis, Stergios},
  journal = {TODO: preprint server},
  year    = {TODO},
  doi     = {TODO}
}

Please also cite the source of each reference you use, listed under Sources, and UNI2-h, as its authors ask:

@article{chen2024uni,
  title   = {Towards a general-purpose foundation model for computational pathology},
  author  = {Chen, Richard J. and Ding, Tong and Lu, Ming Y. and Williamson, Drew F. K. and
             Jaume, Guillaume and Song, Andrew H. and Chen, Bowen and Zhang, Andrew and
             Shao, Daniel and Shaban, Muhammad and others},
  journal = {Nature Medicine},
  volume  = {30},
  number  = {3},
  pages   = {850--862},
  year    = {2024},
  doi     = {10.1038/s41591-024-02857-3}
}