Advance metabolite identification from tandem mass spectra using deep generative models

Tandem mass spectrometry (MS/MS) is a modern technique for measuring metabolites. MS/MS spectra represent molecules by the fragment patterns of compounds that contain structural features of the precursor molecules. The database-searching strategy is the most popular for metabolite identification among its peers. It matches the query MS/MS spectrum to a database of molecule candidates, identifying the metabolite that best matches the query spectrum. This study uses the database-searching strategy and focuses on developing a novel machine learning identification tool. This tool applies autoencoders to map metabolite structures and MS/MS spectra to latent spaces separately. Then, we train a classifier to identify real metabolite-spectrum matches based on the latent space features of metabolites and spectra. Further, we build a generative adversarial network (GAN) to optimize the classifier as the discriminator. A large number of experiments are conducted. The experimental results verify the effectiveness of our tool.

Grantor dc:publisher

University of Tennessee at Chattanooga

Year dc:date.available

2027

Author dc:creator

• Tsai, Meng Hsiu

Contributors dc:contributor

• Wang, Yingfeng
• Liang, Yu; Jain, Hemant (Hemant K.); Qin, Hong; Wu, Dalei
• College of Engineering and Computer Science

• Deep learning (Machine learning)
• Generative adversarial networks (Computer networks)
• Metabolites--Identification
• Tandem mass spectrometry

Licence

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language

English, eng

Repository record dc:identifier

https://scholar.utc.edu/theses/1030

OAI identifier oai:identifier

oai:scholar.utc.edu:theses-2212