hms-dbmi / dseqr

single-cell and bulk RNA-seq analyses from counts → pathways → drug candidates.
https://docs.dseqr.com
Other
20 stars 4 forks source link

Determine if RNA-Seq or microarray gives better drug query results against CMAP02 #127

Closed alexvpickering closed 5 years ago

alexvpickering commented 5 years ago

It would be useful to know for future datasets that we might generate. Here are some studies that have profiled CMAP02 drugs using both RNA-Seq and microarray platforms:

image

alexvpickering commented 5 years ago

No impressive difference between techs (AUC Microarray: 0.759, RNA-Seq: 0.727):

image


One observation that aligns with previous observations (single signatures generated for each compound perform worse than individual signatures for each compound_cell_dose) is that sorting by minimum correlation has a benefit over sorting by average correlation (AUC min: 0.759, avg: 0.672):

image

Same but for RNA-Seq (AUC min: 0.727, avg: 0.664):

image