New Machine-Learning Methods for High-Dimensional, Population-Scale Proteomics
- Post by: Alper Idrisoglu
- 13 November, 2025
- Comments off
ELLIIT Funds New Research Project in Applied Health Technology: Machine Learning for High-Dimensional Proteomics (Project F8)
We are pleased to announce that the project “New Machine-Learning Methods for High-Dimensional, Population-Scale Proteomics” (Project F8) has been selected for funding within ELLIIT Call F. The initiative is led by:
- Principal Investigator: Prof. Andreas Jakobsson, Lund University
- Co-Principal Investigator: Prof. Peter Anderberg, Applied Health Technology, Blekinge Institute of Technology (BTH)
The project brings together methodological expertise in machine learning and information theory with clinical and biomedical research, aiming to advance personalised medicine for critical illnesses such as sepsis and acute respiratory distress syndrome (ARDS).
Scientific focus
The team will analyse high-dimensional proteomic data from two major Swedish biobanks:
- SWECRIT (SomaScan 11,000-protein panel)
- SNAC-Blekinge (Olink proteomics)
Through both supervised and unsupervised machine learning approaches, the project seeks to:
- Identify non-linear and higher-order protein interactions that traditional statistics cannot detect
- Discover proteome-defined disease endotypes linked to clinical outcomes
- Develop interpretable models for clinical decision support, including SHAP, counterfactuals, and rule-based summaries
Role of BTH
The Health and Technology Research Lab (HTRL) at BTH will lead:
- Computational optimisation
- Efficient training methods for large-scale proteomic models
- Scalable and transparent ML pipelines tailored to medical data-sharing constraints
- Development of clinically deployable interpretability layers
Project organisation
The work is structured into two interconnected Ph.D. projects:
- Lund University: ML–MI theory development and endotype discovery
- BTH: Optimised computation, efficient pipelines, and explainable decision-support tools
Project number: F8
Duration: 2026–2030
We look forward to sharing results, publications, and project milestones.