Hi @xengie.doan,
For the results of your analysis in syn18345337, what precisely is the heatmap depicting? Is it representative genes? The most variable? Those that drive the PCs? Also, was any normalization across samples done (e.g. z-score, rank)?
thanks,
sara
Created by Sara Gosline sgosline Yes, it seems that 2-009 xenograft and cell lines are outliers and may need to be removed. I tried removing batch effects using a limma function and visualized it by PCA to see how effective it was in ${entitylist?desc=false&list=syn18361348}
It looks like consortium has a helps correct batch effects, but some cell line/xenograft samples are very different from the other samples and batch correcting for cell line or xenograft doesn't help. I think the transformation is fine as long as it's the same, but can you try z-normalizing each sample to try to account for batch effects? That might help re-center the values as well...
thanks,
sara I used a regularized log transformation and then chose the top 100 varying genes. After a deeper dive, the rlog transformation may shrink low counts because they are closer to the observed count, so this might be why the top 100 varying genes are all highly positive. If you have any suggestions for different normalizations, filtering, etc I'd be happy to incorporate them!