HeapMS

HeapMS automates peak picking and quality assessment of multiple reaction monitoring (MRM) chromatograms for targeted proteomics quantification.

Key Features:

• AI-Driven Peak Picking: Uses convolutional neural networks (CNNs) trained on transformed chromatogram data to automate peak detection and provide confidence assessments.
• Heatmap Transformation: Converts two-dimensional histograms of light and heavy peptides into encoded heatmaps for CNN input.
• Rule-Based Filtering: Applies rule-based filters to remove chromatograms with low interference and high-confidence peak boundaries identified by Skyline prior to AI analysis.
• Categorized Peak Picking: Classifies results into Uncertain (manual inspection), Deletion (flagged for removal or re-examination), and Automatic (high-confidence) categories.
• Integration with Skyline: Imports chromatograms and peak-picking boundaries from Skyline and exports processed results back to Skyline.

Scientific Applications:

• Quality Control of MRM Data: Detects low-quality peaks and flags chromatograms for manual review to improve data curation.
• Detection of Low-Abundance/High-Interference Signals: Enhances identification of peaks with low signal or high interference in targeted proteomics.
• Support for Quantitative Proteomics Workflows: Provides consistent peak boundaries and Skyline-compatible outputs to aid protein abundance measurement.

Methodology:

HeapMS first applies rule-based filters to pre-select chromatograms, then encodes two-dimensional peptide histograms as heatmaps and applies CNNs to perform peak picking into three confidence categories.