DeepMRM

DeepMRM applies deep learning object-detection algorithms to detect and quantify targeted peptides from MRM, PRM, and DIA mass spectrometry data for improved targeted proteomics quantification using stable isotope-labeled (heavy) peptide standards.

Key Features:

• Input data support: Accepts raw mass spectrometry data from Multiple Reaction Monitoring (MRM), Parallel Reaction Monitoring (PRM), and Data-Independent Acquisition (DIA) together with a target list.
• Peptide detection and quantification: Detects and quantifies endogenous (light) peptides using stable isotope-labeled (heavy) peptide standards.
• Output metrics: Produces peak intensities, peptide abundances, and associated quality scores for targeted peptides.
• Algorithmic approach: Implements deep learning algorithms developed for object detection to identify peptide peaks.
• Automation: Reduces the need for manual intervention in targeted proteomics data analysis.
• Validation: Evaluated on internal datasets and publicly available datasets with reported superior accuracy compared to Skyline.
• Integration: The algorithm has been incorporated into Skyline as an external tool.

Scientific Applications:

• Targeted proteomics data interpretation: Automated detection and quantification of targeted peptides from MRM, PRM, and DIA experiments.
• Clinical proteomics quantification: Quantitative measurement of endogenous peptides using stable isotope-labeled (heavy) peptide standards for clinical proteomics workflows.
• Benchmarking and validation: Evaluation and benchmarking of targeted proteomics workflows against community standards such as Skyline.

Methodology:

Accepts MRM, PRM, or DIA raw data plus a target list; applies deep learning object-detection algorithms to detect and quantify endogenous (light) peptides using stable isotope-labeled (heavy) peptide standards and outputs peak intensities, abundances, and quality scores; performance was evaluated on internal and publicly available datasets.