Count-based fingerprints in opensource nodes?

Hi guys,

Is there a way to obtain count-based fingerprints (e.g. ECFC, FCFC) using opensource nodes? That is that I want to represent molecule with a count-vector of fragment frequencies instead of fragment presence/absence vector.


Dear Gio,

Nice coincidence. We are working on a couple nodes that allow to manipulate (expand, collapse, count) byte vector "fingerprints". The type definition is already there but we are lacking a couple preprocessing nodes. As soon as that is done I'm sure some of the vendors and community contributions will add the corresponding producer nodes (like e.g. RDKit).


Indeed. We should be able to either modify the existing node to optionally provide count-based fingerprints or create a new count-based fingerprint node as soon as the preprocessing code is there (more or less).




Dear Bernd and Greg,

this is a great news! Excellent!

In my opinion, in certain situations, count-based fingerprints can have better performances respect presence/absence-based fingerprints and hence it would be great having them available in some knime node.

I'll stay tuned. Thanks a lot.