so I am trying to combine raw counts from TCGA and estimated counts from Geo datasets (Kalisto pseudoalignment) but there are issues with scaling as multiple factors need to be taken into consideration. is there a possibility that i can combine them and mitigate batch effects and scaling issues.

It is possible to combine raw counts from TCGA and estimated counts from GEO datasets (Kalisto pseudoalignment) while mitigating batch effects and scaling issues.

I found this article that have integration methods within dimensionality reduction techniques: Such as principal component analysis (PCA) or t-distributed stochastic neighbor embedding1 (t-SNE).

https://saturncloud.io/blog/scipy-hierarchical-clustering-understanding-strange-results-for-over- drive mad -100-datapoints/