artemsokolov / amp-ad Goto Github PK

The plot produced by

Isolate the code that produces this plot, clean it up and move it to the figs branch.

The MSBB gene expression file has been updated. (We suspect the change has to do with ID anonymization). After verifying that the expression values didn't change, we need to update wrangle/AMP-AD/msbb.R to work with the updated file.

annotables contains precomputed mappings between ENSEMBL gene IDs and HUGO symbols. Preliminary exploration with this package shows that joining Steve's DGE data against this mapping results in a larger number or retained genes.

Action item: Update DGE wrangle script to use annotables instead of EnsDb.Hsapiens.v86.

Current mapping of Dataset and Region vocabulary is in loading functions:

and

Move them upstream to index*() functions to create consistency. Currently, the user has to filter the index differently from the results matrix:

indexBackground() %>% filter( Region == "TempCortex" )

vs.

indexBackground() %>% bkFromIndex() %>% filter( Region == "TCX" )

The following modifications must be incorporated into the metadata table for DGE experiment, to accurately reflect drug identities:

1. ruxolitinib+dsRNA should be changed to ruxolitinib (thus increasing the number of replicates for the latter from 3 to 6).
2. plx-4720 and entinostat annotations should be swapped (i.e., all wells annotated with plx-4720 should be annotated with entinostat, and vice versa).

When plotting -log10(p) values, the log-transformed data gets annotated with the log-transformed values on the corresponding axes. However, these values are harder to interpret than the linear-scale counterparts. The proposal is to relabel the axes with linear-scale values, preserving the log transformation of the data. (Exactly as the third example in Figure 3.4 of https://serialmentor.com/dataviz/coordinate-systems-axes.html)