2015-03-04 toby

Analysis of the information overlap between OMIM, SemMedDB, and Implicitome.

The main program is databases/venn_diag.py. It depends on some functions that
are laid out into subfolders based on the database that it deals with.

The main part of the algorithm is simple pair matching:
Given the structured information pair (subject1, object1), (subject2, object2),
the information is the same if subject1 == subject2 and object1 == object2.

Since each database uses different identifiers for concepts, most of the code
deals with converting identifiers to a common standard, which is
(Entrez gene ID, UMLS CUI).

For converting OMIM identifiers to UMLS CUIs and Entrez gene IDs, I used the
NCBI's Eutils. The code for doing these conversions can be found in the project
"global_utils" (global utilities, which is common code I use for many other
projects), in the file "convert.py".

For converting Implicitome to Entrez gene IDs and UMLS CUIs, I used the dblink
table included in the MySQL version of Implicitome. I performed a left join
to do the conversion in bulk.

Finally, SemMedDB is natively indexed by UMLS CUI and Entrez gene ID, so no
conversion was necessary.