Published December 6, 2024 | Version v3
Model Open

Dual mode MS2Deepscore model

  • 1. ROR icon Wageningen University & Research
  • 1. Wageningen University & Research
  • 2. Hochschule Düsseldorf University of Applied Sciences

Description

An MS2Deepscore model trained on both positive and negative ionisation mode spectra. It is suitable to predict chemical similarity between two mass spectra, both between the same ionmode, but also between positive and negative ion mode spectra. See https://github.com/matchms/ms2deepscore for more details. 

This new version is trained with a new training algorithm resulting in improved model performance. 

In addition a embedding evaluator model has been added. This model is not necessary for running MS2Deepscore, but can be used to estimate the prediction error for the embedding of a spectrum, this can help to filter out spectra with low information content, which will likely get unreliable MS2Deepscore predictions.

Files

embedding_model_settings.json

Files (416.9 MB)

Name Size Download all
md5:3f8fcc2c926c50df6d1ecf92ad2b61ae
791.9 kB Download
md5:713e3fb8330954870e39f373236b400c
273 Bytes Preview Download
md5:9fba8e3ac989963579e885a442ad7db6
416.1 MB Download
md5:b1f295bf1f6b3d08744227c2b41c235c
2.2 kB Preview Download