Chemical dataset curation
First, you must initialize the molecule database by running the following command on one or more SDF files. For testing, I've been using the ChEMBL 33 database.
cura -d DATABASE store MOLECULE_FILENext, you can search for molecules in the existing DATABASE with the query
subcommand:
cura -d DATABASE query -f FORCEFIELD -s PARAM_FILEFORCEFIELD should be the name of an OpenFF-compatible force field (either
built-in like openff-2.1.0.offxml or a local file), and PARAM_FILE should be
a file containing a sequence of parameter IDs, one per line. cura defaults to
searching for ProperTorsions, but this can be changed to another type of
ParameterHandlerwith the --parameter-type/-p flag. You can also apply
filters at this stage, using the --filters/-x flag. Currently supported
filters are:
| Flag | Description |
|---|---|
inchi:FILENAME |
Filter out molecules matching the InchiKeys in FILENAME |
natoms:N |
Filter out molecules with more than N atoms |
For example, to filter out InchiKeys in inchi.dat and molecules with more than
100 atoms:
cura -d DATABASE query -f FORCEFIELD -s PARAM_FILE -x inchi:inchi.dat -x natoms:100The query results are also stored back in the database, with separate results
tables for each unique force field name. If you want to re-run a query for a
given force field, overwriting the previous results, you can use the
--reset/-r flag.