These are basic tools to interact with UMLS lexica, namely UMLS, SNOMED and RxNorm, using Python 3 scripts. For each of the three databases there are scripts (2 Bash and 1 Python) that facilitate import of the downloaded data into a local SQLite 3 database.

You will need a UMLS license to download UMLS lexica.

For a simple start, run one of the files (umls.py, snomed.py, rxnorm.py) in your Shell and follow the instructions. The scripts will prompt you to download and install the databases and, when completed, print a simple example lookup.

There are also utility scripts that offer help for specific use cases, see below.

An auto-generated documentation (via Sphinx) is available but not very exhaustive at the moment. See below for some quick examples.

There are XYLookup classes in each of the three files which can be used for database lookups (where XY stands for UMLS, SNOMED or RxNorm). The following example code is appended to the end of the respective scripts and will be executed if you run it in the Shell. You might want to insert XY.check_databases() before this code so you will get an exception if the databases haven't been set up.

look_umls = UMLSLookup()
code_umls = 'C0002962'
meaning_umls = look_umls.lookup_code_meaning(code_umls)
print('UMLS code "{0}":     {1}'.format(code_umls, meaning_umls))

look_snomed = SNOMEDLookup()
code_snomed = '215350009'
meaning_snomed = look_snomed.lookup_code_meaning(code_snomed)
print('SNOMED code "{0}":  {1}'.format(code_snomed, meaning_snomed))

look_rxnorm = RxNormLookup()
code_rxnorm = '328406'
meaning_rxnorm = look_rxnorm.lookup_code_meaning(code_rxnorm, preferred=False)
print('RxNorm code "{0}":     {1}'.format(code_rxnorm, meaning_rxnorm))

You would typically use this module as a submodule in your own project. Best add this as a git submodule but that really is up to you. But if you do use this module as a Python module, you can't use the name py-umls because it contains a dash, so you must checkout this code to a correctly named directory. I usually use UMLS.

Normalizes an NDC (National Drug Code) number. It's built after the pseudo-code published by NIH, first identifies the format (e.g. "6-3-2") and then normalizes based on that finding. The pseudocode always normalizes the NDC to 5-4-2, padded with leading zeroes and removing all dashes afterwards. This implementation achieves the same normalization.

It also handles NDC codes that come in the "6-4" format, which usually is a 6-4-2 format with omitted package code. Those codes first get normalized to 6-4-2 by appending "-00", then go through the standard normalization.

This script precomputes ingredients, generics, treatment intents, drug classes and mechanisms of action per RxNorm concept (of specific TTYs). The script will run for a couple of minutes but NOT STORE ANYTHING as-is. You can uncomment a line that prints the generated document or do your own things, or even better: use the script rxnorm_link_run.sh after you've set up your MongoDB credentials and the script will store one JSON document per RXCUI and TTY combination into a NoSQL database. Look into rxnorm_link_run.py (note the file ending) where the server connections are set up and insertion happens; a Couchbase implementation is missing.

Create a graphical representation (PDF) of relationships for a given RXCUI. You need to have dot installed (part of the GraphViz package). Run the script from command line and provide a RXCUI as the first argument. The script will then traverse the relationship graph of that concept to a specific depth (8 by default) and plot the relationships into a dot file and a PDF.