yuanyueli / spectralentropy Goto Github PK

Thank you for the extensive investigation of 43 similarity scores leading to the conclusion of Spectral entropy outperforms MS/MS dot product similarity. I noticed the one bond difference is defined to classify true or false identification. I am wondering which tools are used for calculating the bond difference between molecules for Extended Data Fig. 7.

Add paper reference, add WIKI and enable discussions
T.

Thank you for the extensive investigation of 43 similarity scores leading to the conclusion of Spectral entropy outperforms MS/MS dot product similarity. I noticed the one bond difference is defined to classify true or false identification. I am wondering which tools are used for calculating the bond difference between molecules for Extended Data Fig. 7.

Hello! I would like to reproduce the evaluation reported in your paper and extend it with some new methods. However, I am puzzled by your definition of the binary classification metrics, and I would greatly appreciate your clarification.

I understand that for each spectrum, you define a set of candidate spectra for retrieval based on the first block of the corresponding InChI keys. According to the paper, if we denote a subset of retrieved InChI keys (for a certain similarity threshold) as S and a query InChI key as q:

• A true-positive prediction means that q belongs to S.
• A false-positive prediction means that q does not belong to S and S is not empty.

Following the same logic, I assume that a true-negative indicates that S is empty. However, I am struggling to understand what exactly a false-negative implies. Could you please explain your concept of a negative class and the definitions of true-negative and false-negative predictions?

Thank you in advance!

I found that when calculating cosine similarity, the score for two identical spectrograms is not 1. After simple debugging, it was found that there seems to be an indentation error in line 173 of the tools.py file. Some code that seems to be in the while loop is not in the loop