Research

Thesis title: "Improving Accuracy and Reproducibility in de novo (Meta-)Proteomics Through AI and
Database-Independent FDR Estimation"

Thesis outline: This project advances de novo and metaproteomics by improving accuracy, scalability, and
reproducibility through artificial intelligence and database-independent FDR estimation. It combines open-source bioinformatics with deep learning to overcome the limitations of current MS-based workflows, which depend on well-characterized proteomes and often fail for complex or novel samples.

Aims:

  1. Automated open-source analysis pipeline: Development of a unified workflow integrating existing bioinformatic tools for quality control, visualization, and reproducible analysis.
  2. AI-based de novo peptide prediction: Design of a transformer/graph-based model incorporating ion mobility data from 4D MS (timsTOF) to enhance peptide identification.
  3. Database-independent FDR framework: Creation of a new statistical approach for unbiased confidence estimation in de novo sequencing.

Methodology:
The first phase builds the pipeline using modular workflow systems (e.g., Nextflow, Singularity) to ensure reproducibility and accessibility. The second phase develops and validates the deep learning model using large-scale proteomic repositories (UniProt, PRIDE, Massive-KB) and in-house datasets from pain and microbiome research. Incorporating ion mobility aims to resolve ambiguities in peptide spectra. Evaluation will focus on benchmark accuracy, scalability, and real-world applicability.

Expected Outcomes:
The project will deliver a reproducible, open-source workflow and an advanced AI model for accurate de novo peptide sequencing. The database-independent FDR approach will provide a universal benchmark for algorithmic performance. Together, these tools will broaden access to high-confidence proteomic analysis and enable discovery of novel, low-abundance peptides relevant to disease mechanisms such as inflammatory bowel disease.

Supervisor: Manuela Schmidt, Advisor: David Gomez-Varela

Back