Displaying common fingerprint bits as SMARTS in a sleeker way

[PaulC61]

Hi there,

So I've been using this snippet of code to visualize FP bits as smarts for quite a while now. It has been very useful!

https://github.com/greglandrum/rdkit_blog/blob/master/notebooks/Looking%20at%20Morgan%20bit%20counts.ipynb

Given the following smarts strings I can generate the attached plot:
['[OH;D1]-[c;R;D3]',
'C;D3(-[N;D3])=[O;D1]',
'NH;D2-C;D3=[O;D1]',
'CH2;R;D2-[CH2;R;D2]',
'CH2;R;D2-[CH2;R;D2]-[C;R;D4]',
'CH2;D2-N;D3-[CH2;D2]',
'c;R;D3(:[cH;R;D2]:[cH;R;D2]):[cH;R;D2]:[c;R;D3]',
'C@H;R;D3(-[CH2;R;D2])-[C;R;D4]',
'cH;R;D2:c;R;D3:[c;R;D3]',
'CH2;R;D2-[C;R;D4]']

As you can see there is information lost between the SMARTS and the actual MolFromSmarts drawing. I know the original code was designed to visualise the smarts with molecule examples. However, I would like a sleek way of looking at the SMARTS strings themselves like in your recent blog posts (synthons) with attachments and bonding. If you already have some sort of converter, that would be great, if not, some pointers on where to start would be great. I'll get going myself - it seems fine

[greglandrum]

We don't include atomic SMARTS because it can very quickly make structure drawings unreadable.
However, you can always set the atom labels yourself using the atomLabel property on atoms:

It probably makes sense to make the font a bit smaller when you do this:

[PaulC61]

Well actually I don't want to view the atom smarts, I want something close to what you have when you visualise morgan bit fragments like in this blog post of yours:

https://greglandrum.github.io/rdkit-blog/posts/2025-03-29-updated-bit-rendering.html

For example:
'[OH;D1]-[c;R;D3]'

• OH would be shown and c would have two dummy atoms [][] with correct bonding - as for when you visualise morgan bits. From my understanding they're currently drawn with a molecule example, but the atoms-wise SMARTS should have enough info to do the same implicitly no?

[PaulC61]

So I figured it out in a kind of "hacky" way which seems a little buggy but it does work

[greglandrum]

I'm glad you found something that works for you.
In the case of the Morgan bits it's easy to know what to draw (and what a sensible coloring scheme is for the atoms) since we know what atomic features are taken into account in the fingerprint and which "extra" atoms need to be added. This allows the rendering code to look for those specific features.
A general SMARTS pattern is much more complicated to work with.