cc.search_compound confidence metric?

[Maggie Cook]

Hello,

When using cc.search_compound, is there a way to also return some sort of score which quantifies the confidence of the returned identifiers?

Thanks!

[Elad Noor]

Hello,

Yes, search_compound is simply a wrapper for the CompoundCache function "search": https://gitlab.com/equilibrator/equilibrator-cache/-/blob/develop/src/equilibrator_cache/compound_cache.py?ref_type=heads#L262

Here is a code example:

from equilibrator_api import ComponentContribution
cc = ComponentContribution()
search_results = cc.ccache.search("glucose", page=1, page_size=100)

The result will be a list of tuples: (Compound, score).

The score is based on the "ratio" function from the Levenshtein package: https://rapidfuzz.github.io/Levenshtein/levenshtein.html#ratio

So the highest score will be 1 (perfect match) and will go down as the edit distance increases.

Best,

Elad