InstaNovo

InstaNovo translates fragment ion peaks into peptide sequences using a transformer neural network for de novo peptide sequencing in mass spectrometry-based proteomics.

Key Features:

• Transformer-based de novo sequencing: Translates fragment ion peaks directly into amino acid sequences without requiring protein databases.
• InstaNovo+ diffusion model: Iteratively refines predicted peptide sequences to improve sequencing accuracy.
• Large-scale proteomics compatibility: Processes datasets from diverse proteomics experiments with state-of-the-art performance.

Scientific Applications:

• Therapeutic sequencing: Achieves improved sequencing coverage of therapeutic peptides.
• Novel peptide discovery: Identifies previously uncharacterized peptides in proteomics datasets.
• Unreported organism detection: Detects organisms not present in reference databases in metaproteomics or environmental samples.
• Immunopeptidomics: Enables analysis of peptides presented by major histocompatibility complexes without database constraints.
• Dark proteome exploration: Facilitates investigation of protein regions not covered by existing annotation or database resources.
• Direct protein sequencing: Supports emerging approaches for protein-level analysis without enzymatic digestion requirements.

Methodology:

InstaNovo employs a transformer neural network architecture that translates fragment ion peaks into peptide sequences. InstaNovo+ extends this with a diffusion model for iterative refinement of predicted sequences.