proFIA: A data preprocessing workflow for Flow Injection Analysis coupled to High-Resolution Mass Spectrometry

TL;DR

proFIA is a new R package that efficiently preprocesses FIA-HRMS metabolomics data, enabling high-throughput, robust peak detection, quantification, and missing data imputation, thus facilitating large-scale phenotyping studies.

Contribution

It introduces a novel, fast preprocessing workflow specifically designed for FIA-HRMS data, overcoming limitations of existing tools that require chromatography separation.

Findings

Achieves 96% precision and 98% recall in peak detection

Preprocessing time is less than 15 seconds per file

Demonstrates robustness and efficiency on real metabolomics datasets

Abstract

Motivation: Flow Injection Analysis coupled to High-Resolution Mass Spectrometry (FIA-HRMS) is a promising approach for high-throughput metabolomics. FIA-HRMS data, however, cannot be preprocessed with current software tools which rely on liquid chromatography separation, or handle low resolution data only. Results: We thus developed the proFIA package, which implements a suite of innovative algorithms to preprocess FIA-HRMS raw files, and generates the table of peak intensities. The workflow consists of 3 steps: i) noise estimation, peak detection and quantification, ii) peak grouping across samples, and iii) missing value imputation. In addition, we have implemented a new indicator to quantify the potential alteration of the feature peak shape due to matrix effect. The preprocessing is fast (less than 15 s per file), and the value of the main parameters (ppm and dmz) can be easily inferred from the mass resolution of the instrument. Application to two metabolomics datasets (including spiked serum samples) showed high precision (96%) and recall (98%) compared with manual integration. These results demonstrate that proFIA achieves very efficient and robust detection and quantification of FIA-HRMS data, and opens new opportunities for high-throughput phenotyping. Availability: The proFIA software (as well as the plasFIA data set) is available as an R package on the Bioconductor repository (http://bioconductor.org/packages/proFIA), and as a Galaxy module on the Main Toolshed (https://toolshed.g2.bx.psu.edu/) and on the Workflow4Metabolomics online infrastructure (http://workflow4metabolomics.org). Contacts: alexis.delabriere@cea.fr and etienne.thevenot@cea.fr.