Target prioritisation pipeline figure

[bschilder]

Two possible new final figures:

1. Which cell types are associated with severe phenotypes?
   • Two row figure covering a) non-congenital phenotypes which always/often cause death, b) non-congenital phenotypes which always/often cause intellectual disability
   • Which cell types are most enriched (relative to baseline association with phenotypes) for these phenotypes?
   • Which cell types are most enriched (relative to baseline association with phenotypes) for these phenotypes?
   • Maybe one of Brian's network plots showing a top phenotypes/cell-type/gene network for one of each
2. We could show the equivalent of @KittyMurphy 's score graph, but only using the other scores to rank within each of the categories shown below (with these used in place of high/medium/low score).
   1. Always death
   2. Often death
   3. Always ID
   4. Often ID
   5. Always blindness
      The plot would also show the most significantly associated cell type, and best gene target
      How many phenotypes/genes are associated with the following, before/after filtering for congenital onset?

[bschilder]

Here a first draft showing

• left: the severity metrics on the left (after removing phenotypes that often/always cause physical malformations). only showing top 10 phenotypes within the Profound or Severe classes. stratified by congenital vs. non-congenital.
• right: top disease, cell type, a gene target for each phenotype. all labels are colored by -log(p-value) of the phenotype-cell type enrichment test. may want to adjust this.

To do:

• Would like to add a column showing the animal model similarity score for a given phenotype as well.
• Consider swapping the cell type text label with another heatmap (similar to leftside) where x-axis is the cell type and fill is the -log(p-value). Helps to show that some phenotypes enriched in several cell types.
• Merge with network plots for several examples (e.g. Tay-Sachs, neuronal loss).

[bschilder]

version 2

Issues

• Some disease names are too long. Wrapping would screw up the alignments between the left/right subplots. Rather than the entire plot super wide for just those diseases, truncated long names with "..."
• We had planned to add the animal model info (phenotypic similarity, association homologous gene-phenotype overlap). Unfortunately, it just happens that none of the examples in the plot above have animal models within the Monarch resources we're searching. What I can at least do is include whether the homologous gene exists in each species.

[bschilder]

version 3

Now contains human gene target + orthologous non-human animal model genes in black boxes