GlAIcomics: a deep neural network classifier for spectroscopy-augmented mass spectrometric glycans data

Area of Science:

• Biochemistry
• Analytical Chemistry
• Computational Biology

Background:

• Carbohydrate sequencing is crucial in biochemistry but challenging due to numerous isomers.
• Gas phase vibrational laser spectroscopy and mass spectrometry offer promising sequencing methods.
• Analyzing complex vibrational fingerprints requires advanced recognition techniques.

Purpose of the Study:

• To develop an automated method for identifying and classifying monosaccharide vibrational fingerprints.
• To create a robust algorithm for carbohydrate sequencing using spectroscopic data.
• To enable AI-driven glycomics applications.

Main Methods:

• Utilized a Bayesian deep neural network model.
• Trained the algorithm on vibrational fingerprints of various monosaccharides.
• Integrated spectroscopy-augmented mass spectrometry with artificial intelligence.

Main Results:

• Developed a high-performing trained algorithm named GlAIcomics.
• Demonstrated accurate identification and classification of monosaccharide vibrational fingerprints.
• Achieved high confidence in discriminating contamination and identifying molecules.

Conclusions:

• GlAIcomics enables automated and accurate carbohydrate sequencing.
• AI combined with spectroscopy-augmented mass spectrometry is a powerful tool for glycomics.
• This approach significantly advances the field of carbohydrate analysis.