ASICS: an R package for a whole analysis workflow of 1D 1H NMR spectra

Abstract

Motivation In metabolomics, the detection of new biomarkers from Nuclear Magnetic Resonance (NMR) spectra is a promising approach. However, this analysis remains difficult due to the lack of a whole workflow that handles spectra pre-processing, automatic identification and quantification of metabolites and statistical analyses, in a reproducible way. Results We present ASICS, an R package that contains a complete workflow to analyse spectra from NMR experiments. It contains an automatic approach to identify and quantify metabolites in a complex mixture spectrum and uses the results of the quantification in untargeted and targeted statistical analyses. ASICS was shown to improve the precision of quantification in comparison to existing methods on two independent datasets. In addition, ASICS successfully recovered most metabolites that were found important to explain a two level condition describing the samples by a manual and expert analysis based on bucketing. It also found new relevant metabolites involved in metabolic pathways related to risk factors associated with the condition. Availability and implementation ASICS is distributed as an R package, available on Bioconductor. Supplementary information Supplementary data are available at Bioinformatics online.

The maturity in fetal pigs using a multi-fluid metabolomic approach

“…Nuclear magnetic resonance. The detailed protocol for sample preparation, spectra acquisition and preprocessing can be found in Lefort et al 60 . Briefly, each sample of plasma and amniotic fluid (200 µ L) was diluted in 500 µ L deuterated water (D 2 O) and centrifuged without the addition of internal standard to improve spectra quality.…”

“…To measure the concentration of metabolites in the three fluids, NMR metabolomic spectra were processed with ASICS, a recently developed R package 60 . Before quantification, spectra were normalized by the area under the curve and aligned with preprocessing functions available in the Bioconductor R package ASICS 60 (version 2.0.0).…”

“…Quantification was performed using the default reference library provided in the package and was processed independently for each fluid but the same maximum chemical shift, set at 0.02, was allowed for each. The library alignment was improved (compared to that described in Lefort et al 60 ) by using a global quality control criterion: the correlation between quantifications and targeted buckets of the spectra was maximized to choose the best alignment between peaks. Finally, metabolites that had at least 25% of quantifications larger than 0 in at least one condition (stages of gestation and genotypes) were retained.…”

“…Spectra quality control. Plasma relative quantifications were previously validated in Lefort et al 60 using biochemical targeted dosages of three metabolites (glucose, fructose and lactate) in a subset of the samples. The results of a an Orthogonal Projections to Latent Structures Discriminant Analysis 14 (OPLS-DA) based on the standard bucket approach were compared with the results of an OPLS-DA based on metabolite quantifications.…”

Identification of Plasmatic Biomarkers of Foie Gras Qualities in Duck by Metabolomics

“…(ii) The recent metabolite method consists in converting the 1 H-NMR spectra into a metabolite relative concentration table with the ASICS R package (R package version 4.0.2 2 ) that contains an automatic approach to identify and quantify metabolites in a complex 1 H-NMR spectrum from their unique peak pattern (fingerprint) (Tardivel et al, 2017;Lefort et al, 2019). It is a relative quantification in function of the whole spectrum and the metabolite relative concentration has no unit.…”