InstaNovo & InstaNovo+

InstaNovo and InstaNovo+ are state-of-the-art AI models for de novo peptide sequencing, developed by InstaDeep in collaboration with the Department of Biotechnology and Biomedicine at the Technical University of Denmark (DTU) and featured in Nature Machine Intelligence. Available on the DeepChain platform, these tools are designed for proteomics researchers, drug discovery scientists, and biomedical researchers who need to identify novel, previously uncharacterised peptides directly from mass spectrometry data—without relying on reference protein databases.

Traditional peptide sequencing methods are constrained by their dependence on existing protein databases, limiting their ability to detect unknown proteins, novel modifications, or organisms not represented in those databases. InstaNovo and InstaNovo+ overcome these limitations by combining transformer-based sequence prediction with diffusion-based iterative refinement, enabling high-confidence discovery of novel and modified peptides across a wide range of biological samples.

Core Models and Architecture

• InstaNovo is a transformer-based model that interprets mass spectrometry data by mapping fragment ion peaks to peptide sequences, analogous to how speech recognition converts audio into text. It predicts amino acids in a peptide sequence directly from mass spectra without requiring a reference database.
• InstaNovo+ is a diffusion-based iterative refinement model that enhances sequence accuracy by processing entire peptide sequences holistically, rather than predicting one amino acid at a time as autoregressive models do. It begins with an initial sequence—either from InstaNovo or generated randomly—and refines it step by step, mimicking the manual refinement process used by researchers.
• Together, the two models balance precision and exploration: InstaNovo provides initial high-quality predictions, while InstaNovo+ significantly reduces false discovery rates (FDR) and expands the range of potential peptide sequences identified.
• InstaNovo incorporates Knapsack Beam Search decoding, which prioritises peptide sequences consistent with precursor mass constraints, reducing errors and increasing confidence in identifications.

How the Sequencing Workflow Operates

1. A biological sample is ionised and fragmented into smaller peptides via mass spectrometry, generating a mass spectrum based on the mass-to-charge ratio of the fragments.
2. InstaNovo interprets the resulting mass spectrum, mapping fragment ion peaks to peptide sequences without consulting a protein database.
3. InstaNovo+ then refines these initial predictions iteratively, improving alignment with real-world proteomic data and reducing false positives.
4. The combined pipeline delivers scalable, AI-driven de novo peptide sequencing with enhanced reliability and broader proteomic coverage.

Key Capabilities and Demonstrated Applications

• Novel protein discovery: In testing on HeLa cells—a well-studied sample—InstaNovo identified 1,338 previously undetected protein fragments, expanding the known proteomic database.
• Nanobody sequencing: InstaNovo enables high-precision sequencing of nanobodies (small antibody fragments), accurately characterising their sequences to optimise binding to disease targets, with potential applications in infectious diseases, autoimmune disorders, and cancer.
• Complex biological mixtures: In snake venom samples, InstaNovo increased peptide spectrum matches (PSM) by 20% and detected venoms from species outside the original experiment scope, demonstrating its ability to identify organisms not referenced in the study design.
• Infection diagnostics: InstaNovo has demonstrated success in identifying undetected bacteria in wound fluid exudates, supporting more accurate infection diagnosis and antibiotic selection.
• Immunopeptidome analysis: InstaNovo+ enhances detection of MHC-bound peptides, identifying 12,965 novel PSMs and significantly expanding the known immunopeptidome. This supports cancer immunotherapy research, vaccine design, and real-time cellular-level disease insights.

Future Development Roadmap

• Expansion of training datasets to improve peptide prediction accuracy across diverse species and environments.
• Broader support for post-translational modifications (PTMs), enabling detection of both natural and induced protein modifications.
• Development of more user-friendly platforms to make InstaNovo accessible to a wider range of researchers and accelerate adoption across protein science disciplines.

InstaNovo and InstaNovo+ are accessible through the DeepChain platform, which is designed for life sciences R&D teams seeking AI-powered tools with minimal setup and customisable workflows. All performance claims are substantiated by the research paper InstaNovo enables diffusion-powered de novo peptide sequencing in large scale proteomics experiments.