integrated Human Breast Cell Atlas (iHBCA) v1

Highlights

• Harmonizes 2,128,505 cells from 287 donors across seven independently generated studies into a single integrated latent space (scVI).
• Annotated at the level of 11 major cell lineages, consistent across all seven contributing studies.
• Extensively harmonized donor metadata (age, parity, menopausal status, BRCA1/2 genotype, cancer history, BMI) plus technical covariates.
• Provided as a reusable cross-study reference enabling analysis on a common embedding.

Overview

iHBCA v1 is an optimized, integrated Human Breast Cell Atlas: a cross-study single-cell reference of the adult human breast that harmonizes 2,128,505 cells from 287 donors across seven independently generated studies into a single latent space. Using a deep generative model (scVI), cross-study batch effects are reduced while preserving biological variation, placing all contributing studies in one analyzable embedding annotated at the level of 11 major cell lineages.

The atlas carries extensively harmonized donor metadata spanning biological covariates (age, parity, menopausal status, BRCA1/2 germline status, cancer history, BMI) and technical covariates (study, sample, enrichment, dissociation protocol). iHBCA v1 is provided as a reusable community reference enabling cross-study analysis on a common embedding.

Impact

• Places seven independently generated breast studies in one analyzable latent space for direct cross-study comparison.
• Harmonized biological and technical metadata support covariate-aware analysis across the pooled cohort.
• A reusable community reference foundation for the adult human breast, extensible with finer annotation.

Integrated Objects

iHBCA v1.0 — The integrated Human Breast Cell Atlas (iHBCA) v1.0 is a cross-study reference of the non-tumor adult human breast, harmonizing 2,128,505 cells from 287 donors across seven published single-cell RNA-seq studies. It builds on the atlas first introduced in Reed et al. 2024, with a re-optimized integration (scVI), harmonized donor metadata, and cell-type annotation mapped to consensus nomenclature across three broad lineages and 11 major cell-type groups. New datasets can be mapped onto it to transfer these labels.