Improve GO semantic similarity visualizations

[shellywanamaker]

I generated GO semantic similarity heatmaps with this code https://github.com/hputnam/Geoduck_Meth/blob/master/RAnalysis/Scripts/GO_cluster.Rmd to look at relationships among significant GO terms. For each pH comparison (Day 10, Day 135, and Day 145) I generated a BP semantic similarity heatmap and an MF one. Plots are here: https://github.com/hputnam/Geoduck_Meth/tree/master/RAnalysis/Output.

For an example plot, here's the BP semantic similarity heatmap for GO terms significantly enriched on Day 10. It's a symmetrical heatmap with terms listed as rows (I removed them from columns because that would be redundant). The heatmap is colored by similarity with 1 being the most similar (red) and 0 being the least (blue). Often times, 0.5 is chosen as a semantic similarity threshold, but this can be optimized https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4521860/.

I also generated a plot with colored clusters to help with visualization but the cluster cutoff was arbitrarily chosen. Since we are just trying to summarize results, I'm not sure how much this matters.

.

For the paper, we may want to just extract a few clusters from each heatmap and show focused heatmaps of clusters with high semantic similarity.

Any other ideas?