Exhaustive fragmenter returns oversaturated fragment containers #1119

ToLeWeiss · 2024-10-20T17:19:01Z

Dear CDK-Developers,

i encountered oversaturated fragments in the exhaustive fragmenter, when extracting the fragments as atom containers via getFragmentsAsContainers. In contrast, when extracting the fragments via the getFragments method, it returns the expected fragments with the right saturation. So this occurs only when extracting fragments as containers.

Code to replicate

IFragmenter fragmenter = new ExhaustiveFragmenter();
SmilesParser tmpSmiPar = new SmilesParser(SilentChemObjectBuilder.getInstance());
SmilesGenerator tmpSmiGen = new SmilesGenerator((SmiFlavor.UseAromaticSymbols));
IAtomContainer tmpOriginalMolecule = tmpSmiPar.parseSmiles(
    // DL-Phenylalanin CID	994
    "C1=CC=C(C=C1)CC(C(=O)O)N");
fragmenter.generateFragments(tmpOriginalMolecule);
String[] smilesFragments = fragmenter.getFragments();
IAtomContainer[] atomContainerFragments = fragmenter.getFragmentsAsContainers();
System.out.println("Fragments as SMILES:");
for (String fragment : smilesFragments) {
    System.out.println(fragment);
}
System.out.println("Fragments as containers:");
for (IAtomContainer container : atomContainerFragments) {
    System.out.println(tmpSmiGen.create(container));
}

the above, results in the following output:

Fragments as SMILES:
O=C(O)C(N)C
c1ccccc1
NCCc1ccccc1
c1ccc(cc1)C
Fragments as containers:
C[CH2]([CH](=O)O)N
c1ccccc1
C([CH3][cH]1ccccc1)N
C[cH]1ccccc1

The text was updated successfully, but these errors were encountered:

johnmay · 2024-10-20T18:07:13Z

Thanks will take a look, probably just the container one is reusing atoms. You can probably fix it by calling atom typing/add hydrogens but should clean it up as we go.

I actually think it makes more sense to leave the atoms undervalent (radicals) but some might not expect that

johnmay · 2024-10-20T18:55:01Z

Yes it is as I expected, basically internally it does this before generating each fragment:

AtomContainerManipulator.clearAtomConfigurations(frag);
for (IAtom atom : frag.atoms())
    atom.setImplicitHydrogenCount(null);
AtomContainerManipulator.percieveAtomTypesAndConfigureAtoms(frag);
CDKHydrogenAdder.getInstance(frag.getBuilder()).addImplicitHydrogens(frag);
Aromaticity.cdkLegacy().apply(frag);

I'll look a bit more if there is an easy fix, like I said you might not like what I think the correct answer is:

C1=CC=C(C=C1)CC(C(=O)O)N:
 [c]1ccccc1
 [CH2]C(C(=O)O)N
 [CH](Cc1ccccc1)N
 [CH2]c1ccccc1

or with attachment points:

 *c1ccccc1 |$_AP1$|
 *CC(C(=O)O)N  |$_AP1$|
 *C(Cc1ccccc1)N  |$_AP1$|
 *Cc1ccccc1  |$_AP1$|

ToLeWeiss · 2024-10-21T01:45:50Z

Thank you for your quick response

I actually think it makes more sense to leave the atoms undervalent (radicals) but some might not expect that

I also prefer this variant because it lets the user of the CDK decide whether to saturate or not.
I would be interested in implementing these changes if it would be helpful.

johnmay · 2024-10-21T21:22:18Z

Sure, I think actually you just remove the code I pasted above... perhaps add an option for it though.

ToLeWeiss mentioned this issue Feb 6, 2025

Exhaustive fragmenter saturation fix and optimization ToLeWeiss/cdk#1

Draft

3 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Exhaustive fragmenter returns oversaturated fragment containers #1119

Exhaustive fragmenter returns oversaturated fragment containers #1119

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Exhaustive fragmenter returns oversaturated fragment containers #1119

Exhaustive fragmenter returns oversaturated fragment containers #1119

Comments

Code to replicate

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!