Closed bschilder closed 7 months ago
Because the vast majority of diseases only have 1 gene annotated to them in the HPO (7671/8631 diseases = 88.9%), neither EWCE nor phenomix would be very useful for finding enrichment. Instead, I've taken the following strategy:
(please forgive the screenshot, copy-and-paste doesn't bring over the latex equations)
This is symptom-cell type linking procedure is now automated as follows:
results <- MSTExplorer::load_example_results()
results <- MSTExplorer::add_symptom_results(results = results)
I've summarised this in a figure, which is currently designated as Supplementary Fig S1:
Previously did this using a series of many millions of simple gene enrichment tests. #18 But @NathanSkene noted that EWCE's lower gene limit (min of 4 genes in each gene list) is arbitrary and could be lifted. Given the large number of tests, I will need to reduce the number of bootstrap iterations 100k to ~1k (will do some time estimations to make sure that can be run in a reasonable amount of time).