So I have tried several of the cannonical correlation analysis packages in R and none of them seem to be particularly efficient in high dimensional cases. Wondering if there is a better one I can use?
I have tried candisc, CCA and yacca. I have a 10K plus dimensional data that I am trying to integrate together so I am using CCA to get a coordinate system shared between them I can train a random forest on. It's single cell rna-seq data, gene expression at the single cell level for those not familiar. In the end I used Seurat but this is kinda annoying as you then have to pull your transformed data out and it enforces some of seurat normalization methods. The other issue is Seurat does not calculate things like coefficients or correlation of the components.
Seurat uses sparse matrices so I suspect that's why it's so efficient. Wondering if anyone knows of other R packages that can handle the size of the data I am supplying?