WNetAlign: fast and accurate spectra alignment using truncated Wasserstein distance and network simplex

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Computational Biology

Background:

• Liquid chromatography-mass spectrometry (LC-MS) and nuclear magnetic resonance (NMR) spectroscopy are vital for analyzing complex biological samples.
• Overlapping peaks and spectral shifts in LC-MS and NMR data complicate feature matching and data interpretation.
• Accurate alignment of spectral data across multiple runs is essential for reliable compound identification, quantification, and structural elucidation.

Purpose of the Study:

• To develop a robust and scalable algorithm for aligning spectra from LC-MS and NMR experiments.
• To address challenges posed by retention time shifts in LC-MS and frequency shifts in NMR data.
• To enhance the accuracy and efficiency of spectral data analysis in omics and structural biology.

Main Methods:

• A truncated Wasserstein distance-based algorithm was developed for spectral alignment.
• Alignment was formulated as a minimum-cost flow problem solved using the Network Simplex algorithm.
• The algorithm incorporates constraints on maximum transport distance to improve computational speed and accuracy.

Main Results:

• The algorithm achieved high precision (0.97) and recall (0.96) on benchmark LC-MS datasets with a rapid 0.6-s runtime.
• It outperformed existing tools like OpenMS and DeepRTAlign in LC-MS data alignment.
• The method demonstrated effectiveness across various dimensions of NMR data, including 2D, 4D, and 7D analyses.

Conclusions:

• The proposed Wasserstein distance-based algorithm provides a powerful and efficient solution for aligning LC-MS and NMR spectra.
• This method significantly improves the reliability of compound identification and structural analysis in omics studies.
• The open-source implementation (wnetalign) facilitates its adoption in proteomics, metabolomics, and structural biology research.