idea: summary chart of source ontologies

[allaway]

IT would be interesting to auto-generate a plot for the readme that summarizes the source ontologies that we are building on.

[allaway]

Here's a pretty rough script to start:

library(tidyverse)

linkmls <- glue::glue("{getwd()}/{list.files('./modules/', recursive = T, full.names = T)}")



extract_meanings <- function(dat) {
  dat %>%
    map_depth(3, pluck, c("meaning"), .ragged = T) %>%        # Extract 'meaning' from within each 'permissible_values'
    discard(is.null) %>% 
    unlist()                                  # Flatten the result into a vector
}

extract_sources <- function(dat) {
  dat %>%
    map_depth(3, pluck, c("source"), .ragged = T) %>%        # Extract 'meaning' from within each 'permissible_values'
    discard(is.null) %>% 
    unlist()                                 # Flatten the result into a vector

}

sources <- sapply(linkmls, function(x){
  res1 <- yaml::read_yaml(x) %>% flatten()
  c(extract_meanings(res1), extract_sources(res1)) 
}) %>% unlist 
  
library(dplyr)
library(stringr)

source_analyzed <- tibble(sources = sources) %>%
  mutate(simple_source = 
           case_when(
             # Grouping academic publications under a single label
             str_detect(sources, "pubmed.ncbi.nlm.nih.gov") ~ "Publication",
             str_detect(sources, "doi.org") ~ "Publication",
             str_detect(sources, "ncbi.nlm.nih.gov/pubmed/") ~ "Publication",
             str_detect(sources, "ncbi.nlm.nih.gov/pmc/articles") ~ "Publication",
             str_detect(sources, "ncbi.nlm.nih.gov/books") ~ "Publication",
             str_detect(sources, "journals.plos.org") ~ "Publication",
             str_detect(sources, "nature.com") ~ "Publication",
             # Existing and other cases
             str_detect(sources, "ONCOTREE") ~ "ONCOTREE",
             str_detect(URLdecode(sources), "obolibrary") ~ str_extract(URLdecode(sources), 'obo/[:alpha:]+'),
             str_detect(sources, "www.ncbi.nlm.nih.gov/geo") ~ "GEO",
             str_detect(sources, "github.com/HumanCellAtlas") ~ "Human Cell Atlas",
             str_detect(sources, "Sage Bionetworks") ~ "Sage Bionetworks",
             str_detect(sources, "edamontology.org") ~ "EDAM",
             str_detect(sources, "thermofisher.com") ~ "Thermo Fisher Scientific",
             str_detect(sources, "promega.com") ~ "Promega",
             str_detect(URLdecode(sources), "ebi.ac.uk/efo") ~ "ebi/EFO",
             str_detect(URLdecode(sources), "ebi.ac.uk/cmpo") ~ "ebi/CMPO",
             str_detect(sources, "bioassayontology.org") ~ "BioAssay Ontology",
             str_detect(sources, "cognitiveatlas.org") ~ "Cognitive Atlas",
             str_detect(sources, "bioconductor.org") ~ "Bioconductor",
             str_detect(sources, "gatk.broadinstitute.org") ~ "GATK",
             str_detect(sources, "software.broadinstitute.org/gatk") ~ "GATK",
             str_detect(sources, "en.wikipedia.org") ~ "Wikipedia",
             str_detect(sources, "patents.google.com") ~ "Google Patents",
             str_detect(sources, "docs.gdc.cancer.gov") ~ "GDC",
             str_detect(sources, "creativecommons.org") ~ "Creative Commons",
             str_detect(sources, "cellosaurus") ~ "Cellosaurus",
             str_detect(sources, "mesh") ~ "MeSH",
             str_detect(sources, "www.ncbi.nlm.nih.gov/Taxonomy") ~ "NIH Taxonomy",
             str_detect(sources, "obo\\:NCIT") ~ "obo/NCIT",
             str_detect(sources, "jax.org") ~ "JAX",
             str_detect(sources, "illumina") ~ "Illumina",
             TRUE ~ "Other" 
           ))

ggplot(source_analyzed) +
  geom_bar(aes(x = fct_infreq(simple_source)))  +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 50, hjust = 1)) +
  labs(x = "source", y = "count")

[allaway]