DeepFilter

DeepFilter improves peptide identification from metaproteomics datasets by applying deep learning to MS/MS spectra from liquid chromatography–coupled tandem mass spectrometry to enhance detection of peptide-spectrum-matches (PSMs) and proteins.

Key Features:

• Deep Learning Framework: Uses a deep learning-based approach applied to MS/MS spectra from liquid chromatography–coupled tandem mass spectrometry without ad hoc training or fine-tuning.
• Enhanced Identification Accuracy: Identifies up to 12% more peptide-spectrum-matches (PSMs) and up to 9% more proteins compared to Percolator, Q-ranker, PeptideProphet, and iProphet across marine, soil, and human gut metaproteome samples.
• Taxonomic Analysis: Increases species detection by up to 7% in marine, 10% in soil, and 14% in human gut samples in taxonomic analyses.
• Generalization Capability: Generalizes to previously unseen peptide-spectrum-matches, enabling application across diverse metaproteomics datasets.

Scientific Applications:

• Microbiome functional profiling: Improves recovery of peptides and proteins for functional characterization of microbial communities from metaproteomics data.
• Taxonomic profiling: Enhances species-level detection in taxonomic analyses of metaproteomes from marine, soil, and human gut ecosystems.

Methodology:

Applies deep learning to MS/MS spectra from liquid chromatography–coupled tandem mass spectrometry to improve peptide identifications without ad hoc training or fine-tuning.