allaway
commented
4 years ago We also talked about including the 3 LVs that are common to all the tumortypes... but we can reconsider those if we think they would add less to our discussion about the differences between tumortypes.
Yeah, after thinking a bit wasn't clear to me that these were particularly worth highlighting, especially if the point of these plots is to explain that a latent variable is really a ranked list of genes, and convey that some are associated with pathways while others are not not.
happy to add if there is a compelling reason that i am not seeing, but there is already a lot here.
I think this is figure 4-ish in the paper. This is a variant of what we discussed in our meeting today.
I plotted expression and loading for the top 3 LVs with an associated pathway for each tumor type. Filtering for unique LVs only (since there are some dupes in the top 40 lists for each) this left me with 6 LVs overall.
Then, we can segue into the not-well-defined LVs - so I did the same thing with these. I ended up with 7 unique LVs.
Maybe we can use the pheatmap @sgosline plotted to indicate which is associated with which tumor type(s).