Integration of RNA from RNA-seq + Multiome

[mspivakov]

Is it to be expected that single-modality and multiome RNA-seq data are very difficult to batch-correct?

The leftmost plot shows no correction (the left blob are RNA data from two multi-experiments, the right blob are RNA data from two single-modality experiments), the next plot to the right plot shows the same with Seurat integration. Liger (third from left) performs similarly, Harmony (rightmost plot) slightly better, but the populations still don't really merge.