Strange artifacts when wildcart (*) present in SMILES

[luispintoc]

Hi, I found a few artifacts when dealing with SMILES that contain the wildcard *, specially in rings. In the attached screenshot, you can see that a ring becomes a 'square' after encoding and decoding. This happens when converting from BRICS to SAFE.

[FNTwin]

Hi @luispintoc ,

thanks for opening the issue! The main issue is that while cleaning the attach points to construct the molecule in the encoder method, the dummies are getting considered and removed by our regex.

A simple work-around would be (in converter.py) to change the attach_regexp.sub (line 334) from:

scaffold_str = attach_regexp.sub(val, scaffold_str)

to:

n_matches= len(attach_regexp.findall(scaffold_str))
scaffold_str = attach_regexp.sub(val, scaffold_str) if n_matches > 1 else scaffold_str.replace(attach, "*")

This is counting the number of matches and in case substitute them. An attachment point in the safe string molecule should have at least 2 matches so we should naively find the dummies by only counting the number of matches by the attach regex string, and in case just substitute them back with a dummy *.

Now if we run your example:

Don't forget to add remove_dummies=False in the decode otherwise you will remove the dummy.

We'll fix it soon enough with a PR!

Edit: branch opened https://github.com/datamol-io/safe/tree/wildcard_fix

[luispintoc]

Thanks for the quick response, it's working as expected now!