Fragment properties¶

You can annotate fragments with molecular properties and then use those properties to constrain which fragments are selected during generation. This is a two-part workflow: add property columns to the database, then filter on them at generation time.

Add properties with cremdb_add_prop¶

cremdb_add_prop -i fragments.db -p mw logp rtb tpsa fcsp3 -c 8

Option	Default	Description
-i, --input	— (required)	CReM fragment database
-p, --properties	all five	Properties to compute (see below)
-c, --ncpu	1	Number of CPUs
--fetch-batch	50000	Rows fetched per batch
--write-batch	10000	Rows per write
--imap-chunk	500	Multiprocessing chunk size
-v, --verbose	off	Print progress to stderr

Built-in properties:

Name	Property
mw	molecular weight
logp	Crippen logP
rtb	rotatable bonds
tpsa	topological polar surface area
fcsp3	fraction of sp³ carbons

The command is schema-aware: in a v1 database it adds columns to the frags table; in a v0 database it adds them to each radiusN table. It is also incremental — only rows whose property values are still NULL are computed, so rerunning after adding new fragments fills just the new rows.

Filter on properties at generation time¶

Pass property ranges as extra keyword arguments to any generation function. A value may be a single number (exact match) or a (low, high) tuple (inclusive bounds):

from rdkit import Chem
from crem.crem import mutate_mol

mols = list(mutate_mol(
    Chem.MolFromSmiles("c1ccccc1N"),
    db_name="fragments.db",
    set_names="chembl",
    min_freq=5,
    mw=(50, 180),
    logp=(0.0, 3.5),
))

Add custom properties (Python API)¶

crem.db.add_fragment_props adds the built-in properties and, optionally, arbitrary custom columns computed by your own functions:

from crem.db import add_fragment_props
from rdkit import Chem
from rdkit.Chem import Descriptors

def num_rings(smi):
    return Chem.MolFromSmiles(smi).GetRingInfo().NumRings()

# Built-ins only
add_fragment_props("fragments.db")

# Built-ins plus a custom column
add_fragment_props(
    "fragments.db",
    properties="all",
    custom_props={"nrings": num_rings},
    ncpu=8,
)

# Only a custom column, targeting the H-collapsed table
add_fragment_props(
    "fragments.db",
    custom_props={"heavy": lambda s: Descriptors.HeavyAtomCount(Chem.MolFromSmiles(s))},
    table="frags_h",
)

Picklable functions (named functions, functools.partial) run on ncpu workers; lambdas and closures run serially. Custom properties can target either frags (default, keyed on core_smi) or frags_h (keyed on the H-capped SMILES). See crem.db for the full signature.