CHAMPOLLION: Robust Multi-Omics Integration via Inverse Optimal Transport Using Paired Cells
Samaran, J.; Peyre, G.; Cantini, L.
Show abstract
Fully capturing cellular identity requires integrating multiple molecular layers. Bridge integration, i.e. aligning unimodal datasets using a paired multi-omic reference, has emerged as a practical solution, yet existing methods offer limited interpretability and use paired information without regularization, making them sensitive to limited size and coverage. We introduce CHAMPOLLION, which uses regularized optimal transport to learn an interpretable cross-modal metric that drives the alignment of unpaired cells while capturing relationships between molecular features. Benchmarks on RNA-protein and RNA-ATAC datasets show that CHAMPOLLION outperforms existing approaches, remaining accurate with few paired cells and even generalizing to unseen cell types. Beyond alignment, CHAMPOLLION reveals biologically meaningful cross-modal relationships, highlighting in scRNA-protein data a potential role for CD18 across multiple cancers, and, in a human tonsil atlas combining scRNA-seq and scATAC-seq, suggesting that MEF2C may regulate inflammatory responses beyond the brain, notably in plasmacytoid dendritic cells.
Matching journals
The top 3 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Joint probabilistic modeling of paired transcriptome and proteome measurements in single cells 98%
- scGPT: Towards Building a Foundation Model for Single-Cell Multi-omics Using Generative AI 98%
- Towards Universal Cell Embeddings: Integrating Single-cell RNA-seq Datasets across Species with SATURN 98%
Similar papers in this journal
Similar papers in this journal
Similar papers in this journal
Similar papers in this journal
- scCausalVI disentangles single-cell perturbation responses with causality-aware generative model 97%
- Automated assignment of cell identity from single-cell multiplexed imaging and proteomic data 97%
- Learning multi-cellular representations of single-cell transcriptomics data enables characterization of patient-level disease states 96%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.