CSi-Studio / G-Aligner

G-Aligner: a graph-based feature alignment method for untargeted LC-MS-based metabolomics
17 stars 1 forks source link

G-Aligner and MZmine #1

Open froz9 opened 9 months ago

froz9 commented 9 months ago

Hello,

I am interested in testing your alignment algorithm with the intention of incorporating it into my pipeline. However, I find your documentation a bit confusing.

I regularly use MZmine for preprocessing and my alignment method of choice is RANSAC. Is there a way to import the aligned feature list directly back into MZmine?

Additionally, while I may be a novice in this area, I was wondering if you could provide a simpler example for using G-aligner? I have downloaded MetaPro, but G-aligner does not seem to be available there yet.

Thank you in advance for your assistance.

CSi-Ti commented 9 months ago

Conclusion first. We didn't provide a way to import the aligned feature list back to MZmine.

G-Aligner requires feature extraction results from untargeted software as input, aligns these feature lists into a feature map, and outputs the aligned map to a table file.

The contribution of G-Aligner is to propose a more accurate feature alignment algorithm. However, G-Aligner only focus on the alignment technique, no feature indexing or MS2 transfering functions were implemented, and do not support feature identification in the current version. If you are good at coding, you can match the features in the aligned map back to feature lists by m/z, RT and intensity coordinates, achieve correlation of MS2 spectra, and perform identification in software such as GNPS.

As for MetaPro, MetaPro is a semi-targeted batch analysis platform, which uses different analysis method for feature extraction and alignment. MetaPro uses semi-targeted analysis method, using targeted algorithms (need a list of targeted compound ions) on untargeted data (DDA). Based on the semi-targeted method, MetaPro naturally has higher speed and accuracy than untargeted methods, and supports simultaneous result curation on hundreds of samples. For each target ion, MetaPro performs batch EIC extraction, peak finding, peak alignment, and identification. In the peak alignment procedure, MetaPro first performs internal standard-based global RT alignment, and then performs EIC similarity-based methods for local peak alignment. Comparing with G-Aligner, MetaPro aligns peaks (RT, intensity) rather than features (m/z, RT, intensity), and requires more information (internal standard, EIC, peak coordinate) than G-Aligner (peak coordinate only). As a result, G-Aligner will not be implemented into MetaPro, but may be implemented in our developing untargted platform in the future.