Privacy-preserving federated tensor decomposition of single-cell immune data: recovering multicellular programs across institutions

2026-06-22

AI Summary

This study addresses the challenge of limited data sharing across institutions and ethnic groups in single-cell immunology due to privacy constraints, which hinders collaborative discovery of multicellular programs. The authors propose a privacy-preserving federated tensor decomposition method wherein each participant computes program subspaces locally, and a coordinator aggregates these via stacked singular value decomposition (SVD) combined with federated global mean centering—yielding results equivalent to centralized decomposition without sharing raw cell-level data. This approach enables, for the first time, cross-institutional and cross-ethnic recovery of multicellular programs, accommodates missing cell types at some sites, and is robust to site-label perturbations. Validations on systemic lupus erythematosus and COVID-19 datasets accurately recapitulate interferon programs (AUC=0.998) with high cross-site subspace correlation (0.989); in interstitial lung disease, it outperforms single-cell-type models (AUC 0.96 vs. 0.91). Secure aggregation also substantially mitigates membership inference attack risk (AUC reduced from 0.91 to 0.61).

Abstract

Tensor decomposition of donor $\times$ cell-type $\times$ gene single-cell data recovers \emph{multicellular programs}: coordinated axes of inter-individual transcriptional variation that span cell types and stratify disease. Yet immune single-cell atlases are increasingly multi-institution, multi-ancestry, and governed, so patient cells often cannot be pooled. We present a federated estimator: each site computes a local program subspace, and a coordinator merges these by stacked SVD under federated global-mean centering, provably equivalent (up to truncation) to the centralised decomposition. This centering makes the merge robust to site-label confounding (program AUC $0.957$ vs.\ $0.861$ for naive per-site centering). Only program subspaces leave a site, and aggregation is compatible with secure aggregation. On a 261-donor systemic lupus erythematosus atlas it recovers the canonical interferon program (ISG enrichment AUC $0.998$; case--control separation $0.958$; bootstrap $Δ\text{AUC}=-0.000$, 95\% CI $[-0.004,+0.012]$ vs.\ centralised), across institution-scale and multi-ancestry partitions, and across three \emph{real} COVID-19 sites (subspace correlation $0.989$). It recovers the program when \emph{no site observes all cell types} (correlation $1.000$, exact by construction), which fixed-feature federated PCA cannot. On an interstitial-lung-disease atlas the recovered program predicts disease better than the best single cell type (AUC $0.96$ vs.\ $0.91$; gap 95\% CI excludes zero) and the advantage survives federation; a liver cohort is consistent ($p=0.005$). Membership-inference shows secure aggregation cuts attack AUC from $0.91$ to $0.61$. The method enables cross-institution, cross-ancestry recovery of multicellular immune programs without sharing cells.

Problem

• federated learning
• tensor decomposition
• single-cell data
• privacy preservation
• multicellular programs

Innovation

• federated tensor decomposition
• multicellular programs
• privacy-preserving learning
• secure aggregation
• single-cell immune data