ur-whitelab/exmol

CODEX Example

Closed this issue · 6 comments

While messing around with CODEX, I noticed it wants to compute ECFP4 fingerprints using a different method and this gives slightly different similarities. @geemi725 could you double-check the ECFP4 implementation we have is correct, or is the CODEX one correct?

image

@whitead this is the reason. Not sure how to pick "what's the best" though.

Screen Shot 2021-11-23 at 10 11 04 AM

@whitead also codex is taking 1-similarity with the bit vectors but the output must be similarity, not distance?

So which is more accepted?

My understanding is that Morgan FPs by default uses the counts. If we use bit vectors we will be losing info pertaining to the counts. But I have seen bitvectors used more commonly (in the past few weeks).

For future references: Visualizing which atoms contribute to similarity between two molecules.
Screen Shot 2021-11-23 at 10 55 43 AM

Interesting, maybe we could use that for @navneeth3005's project