| Summary: | Single-cell sequencing has catalyzed a significant shift in biological modeling and hypothesis generation. In this study, we present advanced algorithms that utilize multiple modalities for peak-gene linking via correlation-based mechanisms. This approach provides a robust framework for bias correction and noise reduction at single-cell resolution.
Our research introduces a novel algorithm that employs peak-gene linking to integrate epigenomic data, thereby translating it into transcriptomic data. This integration facilitates comprehensive secondary analyses on meta-cells and sub-cell types. Crucially, our methodology enables swift computation of modules from any single-cell assay, promoting exploration of intricate biological and disease mechanisms that might remain unmodeled with a pseudo-bulk approach.
By combining snRNA-seq and snATAC-seq data, our method substantially outperforms equivalent tasks of gene expression estimation, showing promising results even with unpaired real data. Furthermore, the modeling of genomic peak modules using our algorithms uncovers additional signal potentially overlooked when examining single peaks.
We envision correlation-based linking as a key aspect of future single-cell multiomic technology, as it allows for correlation between assays at the single-UMI level. As such, improved modeling of data at the single-cell level will enhance our understanding of complex gene regulatory networks and disease mechanisms.
|
|---|