NeatMS

NeatMS performs signal labeling and filtering of untargeted liquid chromatography-mass spectrometry (LC-MS) data to improve peak detection and reduce false positives in metabolomics analyses.

Key Features:

• Machine Learning-Based Classification: NeatMS employs a convolutional neural network (CNN) for peak classification to distinguish true chemical signals from noise.
• Pre-Trained Model: The package includes a pre-trained model that encapsulates expert knowledge to aid detection of true peaks.
• Customization and Transfer Learning: Users can train new models or refine existing ones using transfer learning for dataset-specific adaptation.
• Integration with LC-MS Workflows: Designed for integration into untargeted LC-MS analysis pipelines to enable scalable signal curation.

Scientific Applications:

• Metabolomics peak curation: Improves peak curation by filtering false positives to increase the reliability of metabolomics datasets.
• Biomarker discovery and metabolic profiling: Provides more reliable LC-MS signal sets for biomarker discovery and metabolic profiling in large-scale experiments.
• Systems biology and large-scale studies: Supports systems biology analyses by reducing noise in LC-MS datasets for robust downstream interpretation.

Methodology:

NeatMS applies a CNN-based classification approach, provides pre-trained models and support for transfer learning, and validates performance by comparison with existing peak-detection methods to quantify reduction of false peaks.