scMoMaT jointly performs single cell mosaic integration and multi-modal bio-marker detection

Abstract

Single cell data integration methods aim to integrate cells across data batches and modalities, and data integration tasks can be categorized into horizontal, vertical, diagonal, and mosaic integration, where mosaic integration is the most general and challenging case with few methods developed. We propose scMoMaT, a method that is able to integrate single cell multi-omics data under the mosaic integration scenario using matrix tri-factorization. During integration, scMoMaT is also able to uncover the cluster specific bio-markers across modalities. These multi-modal bio-markers are used to interpret and annotate the clusters to cell types. Moreover, scMoMaT can integrate cell batches with unequal cell type compositions. Applying scMoMaT to multiple real and simulated datasets demonstrated these features of scMoMaT and showed that scMoMaT has superior performance compared to existing methods. Specifically, we show that integrated cell embedding combined with learned bio-markers lead to cell type annotations of higher quality or resolution compared to their original annotations.

scGPT: Towards Building a Foundation Model for Single-Cell Multi-omics Using Generative AI

“…The first two sequencing batches contain gene expression and protein abundance data from CITE-seq, and the second two batches have chromatin accessibility and protein abundance measurements from ASAP-seq. In the benchmark experiment with scMoMat [65], scGPT demonstrates superior batch correction performance as shown in Figure S.1).…”

“…scMultiomic integration We benchmarked scGPT in two integration settings, paired and mosaic, against the recent scMultiomic integration methods Seurat v4 [24], scGLUE [9] and scMoMat [65] respectively. In the paired data integration experiment, we benchmarked scGPT with scGLUE [9] and Seurat v4 [24] on the 10X Multiome PBMC [14] dataset.…”

“…The same 1,200 highly variable genes and 4,000 highly variable peaks were used as input to all methods. In the mosaic data integration experiment, we benchmarked scGPT with scMoMat [65] on the ASAP PBMC [38] dataset.…”

“…Batch correction metrics are not applicable to this setting. For the mosaic ASAP PBMC [38] dataset, scGPT was benchmarked with scMoMat [65] for cell type clustering and multiomic integration performance, evaluated on eight biological conservation and batch correction metrics.…”

Multimodal generation of astrocyte by integrating single-cell multi-omics data via deep learning

“…Such an integration is necessary to overcome the current limitations in modality scalability and cost associated with currently accessible scMulti-omics sequencing technologies. Nevertheless, multimodal mosaic integration is quite challenging [56,65]. A key challenge involves addressing the diversity of modalities and handling technical variations across different batches.…”