Accurate, fully-automated NMR spectral profiling for metabolomics

Area of Science:

• Biochemistry
• Analytical Chemistry
• Computational Biology

Background:

• Metabolic profiles in biofluids change with disease, offering diagnostic potential.
• Nuclear Magnetic Resonance (NMR) spectroscopy can extract metabolic information from biofluids.
• Current NMR metabolomics are manual, slow, and error-prone, limiting clinical use.

Purpose of the Study:

• To develop an automated system, BAYESIL, for rapid and accurate metabolic profiling from NMR spectra.
• To overcome the limitations of manual NMR spectral analysis in metabolomics.

Main Methods:

• Developed BAYESIL, a system for autonomous metabolic profiling from 1D 1H NMR spectra of biofluids (serum, CSF).
• Employed spectral processing and matching against a reference compound library using probabilistic graphical models.
• Utilized a Bayesian inference approach for spectral matching.

Main Results:

• BAYESIL achieved ~90% correct identification and ~10% quantification error for over 50 compounds.
• The system processed complex mixtures, including real biological samples, in under 5 minutes on a single CPU.
• Performance met or exceeded that of trained experts in quantitative NMR spectral profiling.

Conclusions:

• BAYESIL is the first fully-automatic, publicly accessible system for quantitative NMR spectral profiling.
• This tool promises to accelerate high-throughput metabolomics and expand NMR applications in clinical settings.