Automated Gas Chromatography Peak Alignment: A Deep Learning Approach using Greedy Optimization and Simulation.

Area of Science:

• Analytical Chemistry
• Computational Biology
• Medical Diagnostics

Background:

• Volatile organic compounds (VOCs) in breath are clinically significant for disease detection.
• Gas chromatography (GC) is used to measure these VOCs.
• Chromatographic peak alignment is a critical and challenging step in VOC analysis.

Purpose of the Study:

• To develop an automated method for chromatographic peak alignment.
• To overcome limitations of traditional semi-automated alignment algorithms (slow, expensive, inconsistent).
• To improve the analysis of breath VOCs for disease diagnosis.

Main Methods:

• A pipeline was developed to train a deep-learning model using simulated artificial chromatograms.
• A small, annotated dataset was used for model training.
• A postprocessing step involving greedy optimization was employed for signal alignment.

Main Results:

• The proposed pipeline enables automated chromatographic peak alignment.
• The deep-learning model, trained on simulated data, effectively aligns signals.
• This automation addresses the challenges of manual intervention in traditional methods.

Conclusions:

• Automated chromatogram analysis using deep learning can significantly improve VOC-based disease diagnosis.
• This approach has the potential to enhance the management of respiratory diseases like asthma, ARDS, and COVID-19.
• The developed method offers a faster, more consistent alternative to manual alignment techniques.