Join GitHub today
GitHub is home to over 31 million developers working together to host and review code, manage projects, and build software together.
Sign upMinor curiosity: why are some scorings (e.g MoNA) not normalized? #24
Comments
This comment has been minimized.
This comment has been minimized.
schymane
commented
Feb 12, 2019
All scores should be normalized when the total score is calculated (between 0 and 1, then adjusted with the weighing given), but in the results file the raw scores are exported, as these are also useful (and it’s easy to repeat the normalization). MoNA Exact Similarity should be between 0 and 1, whereas the MoNA MetFusion-style score can be greater than one – this is a function of the way it is calculated (see e.g. the MetFusion paper). If the MetFusion score is far above one it’s a sign that there are a lot of similar spectrum/structure combinations adding weight to the identification, if they are well below one there’s not much similar on record … does that help?
|
This comment has been minimized.
This comment has been minimized.
Hi Emma, Thanks for your explanation! This is quite interesting, I will need to have a better look at the MetFusion paper though. For now if I understand correctly what is desired here is to keep the original scaling in order to judge what is going on. I do wonder know though if it is 'fair' that >1 MetFusion scores should be able to influence the final score with a greater weight since they are not normalized? Perhaps it would make more sense to report the non-normalized scores but normalize them before calculating the total score? |
This comment has been minimized.
This comment has been minimized.
schymane
commented
Feb 14, 2019
The MetFusion scores are normalized along with everything else, so they only count a maximum of 1 to the total score in the end. The values that are greater than one that you see are the non-normalised “raw” values. (c.f. publications, also normalized to 1 in the final score, but sometimes in the 100s or 1000s in the “raw” values)
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
rickhelmus commentedFeb 12, 2019
When trying to reconstruct relative scores from the absolute values returned by MetFrag CL I noticed that some scorings (e.g. MoNA) are not normalized. I am just curious why this is.
Thanks,
Rick