The researchers have developed a tool for the complete application of untargeted nuclear magnetic resonance (NMR) based metabolomics for biomarker discovery in different biosamples such as urine, blood or cell extracts.
Untargeted NMR metabolomics comprises the investigation of low-molecular–weight compounds, called metabolites, that are dysregulated under specific conditions such as illness. The use of an untargeted approach does not require previous knowledge on a specific metabolic pathway but provides precious information about fingerprints that these specific conditions reveal.
Untargeted Metabolomics has an important computational content, AlpsNMR covers the full workflow of NMR data from signal preprocessing to machine learning for biomarker discovery. The software is readily available to the metabolomics community.
Santi Marco, Group leader at IBEC and UB Professor
The developed tool, called AlpsNMR (Automated spectraL Processing System for NMR), is an R package that applies a concise workflow in an automated manner, in which users does not need too much knowledge about the procedure. This pipeline comprises all steps of the metabolomics workflow, from 1D spectra until the identification of the putative biomarkers that predict a condition. The pipeline starts with the loading step that imports the directory containing multiple 1H NMR spectra. The package also includes algorithms for the detection of outlier samples, and is able to report and remove those, if applicable. The package is also able to detect, align and normalize of NMR peaks by several algorithms (urinary creatinine, probabilistic quotient normalization, total area…) particularly useful in urine biosamples. The tool makes interactive plots in HTML format that may be edited, zoomed or converted to a PNG figure. After the integration of peaks, the acquired data frame is ready for statistical analysis. Peak (or feature) selection process is carried out based on sophisticated machine learning specially developed for this package tool. This statistical analysis entails the application of Partial Least Square models using Bootstrap. Finally, the identification of relevant features uses the source of the biological sample (plasma/serum, urine and cells) generating ranked metabolite candidates obtained according to the Human Metabolome Database.
Contributing researchers include scientists from Nestlé Research, Lausanne, Switzerland.
Step-by-step workflow for untargeted NMR-based metabolomics analysis using AlpsNMR: An R tool for Automated spectraL Processing System for untargeted NMR-based metabolomics.
Reference article: Francisco Madrid-Gambin, Sergio Oller-Moreno, Luis Fernandez, Simona Bartova, Maria P. Giner, Cristopher Joyce, Francesco Ferraro, Ivan Montoliu, Sofia Moco and Santiago Marco. AlpsNMR: an R package for signal processing of fully untargeted NMR-based metabolomics. Bioinformatics, vol. 36, no. 9 , 2943-2945., 2020.
Software digital object identifier: 10.18129/B9.bioc.AlpsNMR