scglue-unpaired-multiomics-integration

Unpaired scRNA-seq and scATAC-seq data integration uses a guidance graph to align modalities and generate joint cell embeddings. This approach maps features across modalities to reveal shared cellular structure and regulatory relationships without requiring matched barcodes.

A guidance graph for single-cell multi-omics integration maps features across modalities using genomic coordinates and optional Hi-C or eQTL data. It directs the graph-based integration model to align unpaired scRNA and scATAC datasets for cis-regulatory inference.

Yes, you can infer transcription factor target networks from unpaired scATAC and scRNA data. The integration model produces feature embeddings that enable cross-modality gene regulatory network construction and cis-regulatory inference from the aligned datasets.

Yes, you need precomputed HVG graphs, ATAC LSI embeddings, and genomic coordinates to perform scGLUE integration. These prerequisites provide the foundational feature structures and genomic context required to align unpaired single-cell modalities.

To incorporate Hi-C or eQTL data into single-cell regulatory inference, you extend the guidance graph using 150kb windows or external interaction data. This flexible graph extension enhances regulatory discovery and cis-regulatory inference during the multi-omics alignment process.

You should not use a guidance graph approach for unpaired multi-omics data if you lack precomputed HVG graphs, ATAC LSI embeddings, or genomic coordinates. The integration model requires these specific inputs to successfully align modalities and produce joint embeddings.