DeepLC

DeepLC predicts peptide retention times for liquid chromatography-mass spectrometry (LC-MS) analyses using deep learning to support peptide identification, including peptides with post-translational or other modifications.

Key Features:

• Atomic composition encoding: Employs a peptide encoding based on atomic composition to represent amino acids and modifications, enabling prediction for modified peptides not present in the training data.
• Performance and accuracy: Achieves performance comparable to state-of-the-art methods for unmodified peptides and demonstrates improved accuracy for peptides with novel modifications.
• Application in open modification searches: Provides retention time estimates that can be used to flag potentially incorrect identifications during open modification searches.
• Use with complex biological datasets: Has been applied to datasets such as CD8-positive T-cell proteome data to assist identification of modified peptides.

Scientific Applications:

• LC-MS peptide identification: Improve peptide identification in LC-MS workflows by supplying predicted retention times as orthogonal evidence.
• Open modification searches: Support open modification searches by enabling retention time prediction for peptides carrying known and novel modifications, reducing identification ambiguity.
• Post-translational modification analysis: Aid detailed analysis of post-translational modifications by providing retention time-based validation for modified peptide assignments.

Methodology:

DeepLC uses a deep learning model with an atomic-composition-based peptide encoding and is trained to predict liquid chromatography retention times for peptides, including modified species.