Automated Peak Detection Without Conventional Picking

Automated Peak Detection Without Conventional Picking

License: restricted — no clear open-source license detected for the underlying tool; verify licensing before commercial use or redistribution.

Summary

This skill applies image processing and machine learning to chromatography–mass spectrometry data to automatically identify multivariate marker features directly from two-dimensional MS maps (m/z vs retention time), bypassing conventional peak picking which suffers from false detections. It is particularly suited for untargeted metabolomics of complex samples like human breath at parts per billion sensitivity levels.

When to use

Apply this skill when you have a two-dimensional GC–MS or LC–MS dataset (m/z vs retention time) and need to identify discriminative analyte features without relying on conventional peak picking algorithms. This is especially valuable when analyzing complex, low-abundance samples (e.g., human breath, aroma compounds) where conventional peak detection generates false positives, or when untargeted marker discovery is the goal rather than targeted quantitation.

When NOT to use

• When input has already been processed through conventional peak picking and stored as a feature matrix or peak list—the method is designed to work directly on raw or minimally processed 2D MS maps.
• When high-throughput targeted quantitation with predefined compound libraries is the primary goal; this skill prioritizes marker discovery and multivariate discrimination over absolute quantitation.
• When retention time shifts are severe and unaligned across samples; the method assumes retention time consistency or requires pre-alignment of the 2D MS maps.

Inputs

• Two-dimensional MS map (m/z vs retention time) from GC–MS or LC–MS instrument
• Chromatography–mass spectrometry raw data or processed intensity matrix
• Sample metadata (control vs. case grouping for multivariate feature extraction)

Outputs

• Structured feature table with m/z values, retention times, and feature scores
• Visualization of identified marker features overlaid on m/z–retention time space
• Classification results discriminating analyte signals from background

How to apply

Load the two-dimensional MS map into NPFimg as a gridded intensity image. Apply the image processing pipeline to enhance signal-to-noise contrast and normalize the spatial representation of m/z and retention time axes. Deploy the machine learning classifier to discriminate true analyte signals from noise and baseline artifacts across the entire map without pre-specifying peak boundaries. The algorithm automatically outputs marker features with associated m/z values, retention times, and feature scores. Validate results by comparing with conventional peak-picking tools (e.g., XCMS) and examining signal acquisition error rates and marker identification reliability.

Related tools

• NPFimg (Primary tool implementing image processing and machine learning pipeline for automated marker feature detection from 2D MS maps without conventional peak picking) — https://github.com/poomcj/NPFimg
• XCMS (Benchmark comparison tool for conventional peak picking; used to validate NPFimg's lower error rates in signal acquisition and marker identification)

Evaluation signals

• Error rates in signal acquisition and marker identification are lower or comparable to XCMS when applied to the same dataset, as documented in the article's comparison study.
• Identified marker features exhibit m/z and retention time values consistent with expected analyte properties and avoid false detections known to plague conventional peak picking.
• Visualization of marker features on the m/z–retention time space shows spatial coherence and discrimination between control and case groups; features cluster in distinct regions rather than scattered randomly.
• Feature scores output by the algorithm are reproducible across replicate analyses and correlate with analyte abundance or biomarker relevance in downstream validation (e.g., ROC analysis).
• The method successfully identifies multivariate chemo-/biomarker features at parts per billion sensitivity levels in complex matrices (e.g., human breath) without manual peak boundary definition.

Limitations

• Method requires that the two-dimensional MS map be properly registered and aligned across samples; severe retention time drifts or m/z calibration errors will degrade performance.
• Limited documentation and update history noted in the repository; external users may encounter undocumented edge cases or parameter sensitivities not fully described in the primary publication.
• Applicability is demonstrated primarily for GC–MS and LC–MS at parts per billion levels; performance on other chromatographic platforms or at different dynamic ranges has not been systematically evaluated.
• The machine learning classifier is trained and validated on aroma odor and human breath datasets; generalization to substantially different metabolite classes or sample matrices requires independent validation.

Evidence

• [intro] NPFimg automatically identifies multivariate chemo-/biomarker features of analytes in chromatography–mass spectrometry (MS) data by combining image processing and machine learning: "We present a method named NPFimg, which automatically identifies multivariate chemo-/biomarker features of analytes in chromatography–mass spectrometry (MS) data by combining image processing and"
• [intro] Bypasses conventional peak picking which suffers from false peak detections: "Our approach allows us to comprehensively characterize the signals in MS data without the conventional peak picking process, which suffers from false peak detections."
• [intro] Process two-dimensional MS map to identify and visualize marker features: "NPFimg processes a two-dimensional MS map (m/z vs retention time) to discriminate analytes and identify and visualize the marker features."
• [intro] Lower error rates compared to XCMS in signal acquisition and marker identification: "Comparison with the widely used XCMS shows the excellent reliability of NPFimg, in that it has lower error rates of signal acquisition and marker identification."
• [intro] Successfully demonstrated on human breath at parts per billion level: "The feasibility of marker identification is successfully demonstrated in case studies of aroma odor and human breath on gas chromatography–mass spectrometry (GC–MS) even at the parts per billion"
• [intro] Potential applicability to untargeted metabolomics of human breath: "In addition, we show the potential applicability of NPFimg to the untargeted metabolomics of human breath."
• [intro] Broadly applicable to diverse metabolomics/chemometrics using GC–MS and LC–MS: "NPFimg is potentially applicable to data processing in diverse metabolomics/chemometrics using GC–MS and liquid chromatography–MS"
• [readme] README confirms automated identification without conventional peak picking: "Our approach allows us to comprehensively characterize the signals in MS data without the conventional peak picking process, which suffers from false peak detections."