Figure 1 Discussion

[gwaybio]

Let's use this issue to discuss modifications to Figure 1. Code added in #19.

Figure 1. Cell Painting and L1000 data provide complimentary measurements of compound perturbations across doses. (a) Percent replicating scores calculated via median pairwise correlations between replicates of compound perturbations across dose. We generated the non-replicate distribution from matched randomly sampled non-replicate groups. See methods for full details about sampling and data processing. The dotted blue line indicates the 95th percentile of the non-replicate distribution. (b) The L1000 and Cell Painting assays reproducibly measure a complementary set of compound perturbations. The three numbers represent (from top to bottom) the number of reproducible compounds unique to L1000, the number of overlapping reproducible compounds between the two assays, and the number of reproducible compounds unique to Cell Painting. (c) Percent matching scores calculated via median pairwise correlation of level 5 consensus signatures of compounds mapping to the same mechanism of action (MOA) across dose. The dotted blue line indicates compounds significantly matched to their MOA based on a non-parametric p value (see methods). (d) The L1000 and Cell Painting assays each capture a complementary set of compound MOAs. The three numbers represent counts reflecting the ordering as described in panel b. (e) Consensus signatures of select MOAs demonstrate high heterogeneity across assays and doses.

[gwaybio]

• [x] Some heatmaps are missing dividing lines between compounds for some reason. I just need to slightly tinker with plot size
• [x] panels a thru d can be smaller and panel e can be larger

[gwaybio]

from @shntnu

• [x] Add the "overlapping" category to the legend to reflect this and drop the legend title ("Assay")
• [ ] I think it will be a lot more compelling if these plots show the percentiles (i.e. i,j is the rank of i in the ranked listed of similarity to j) rather than the Pearson correlations. Imagine everything was similarly toxic at a high dose; we'd expect Pearson within an MOA to be high. But the percentile to be low. This is not critical because we know that MOAs do not look the same at high-doses, as seen in 1d

[gwaybio]

• [x] Confirm that the percent replicating numbers and null distribution are being derived from the overlapping compound set

[gwaybio]

• [x] Add supplementary figure comparing Cell Painting compound median replicate correlations with the same replicate counts as L1000