OUTCAST-AI: Unraveling Omics-Data for the Identification of Long-Post-COVID Therapeutic Targets: An Artificial Intelligence-Based Approach
Funded by: Federal Ministry of Research, Technology and Space
Duration: 01/2025 - 12/2026
Aim of the project
The COVID-19 pandemic has had a deep impact on global health. While much attention has been focused on the acute phase of the disease, a significant proportion of patients continues to experience debilitating symptoms long after their initial recovery, a condition now referred to as Long /Post-COVID (LC/PC) syndrome. Despite its high prevalence and heterogenous clinical mani festation the pathomechanisms driving LC/PC and its various subtypes along with the development of targeted and individualized treatments still have to be elucidated.
This project aims to develop innovative artificial intelligence (AI) based methods by a multidisciplinary approach to promote best practice therapeutic interventions for patients suffering from the LC/PC syndrome. By integration of expertise in biochemistry, clinical medicine, biometry, medical informatics, bioinformatics and AI molecular, clinical as well as routine data sets of large Long-/Post COVID patient cohorts will be analyzed to identify and understand the heterogenous and complex pathomechanisms driving LC/PC and its subtypes.
These AI-based approaches for identification of LC-/PC specific molecular alterations and clinical determinants will allow for the differentiation and classification of distinct LC/PC subtypes. By establishment of an AI pipeline for the identification of potential therapeutic targets tailored to these LC-/PC- subtypes novel treatment strategies for LC/PC syndrome will be developed, thus paving the way for individualized treatment approaches. Project results will have a great impact on the whole process of healthcare management for patients suffering from the LC/PC syndrome: The identification of the underlying LC-PC specific patho mechanisms will allow for early diagnosis of the disease and refinement of therapeutic strategies as well as health service utilization. These achievements will lead to a substantial improvement of patient`s outcome and a reduction of the socioeconomic burden of the LC/PC syndrome.
Identifying new therapeutic targets is traditionally laborious and time-consuming.
Machine Learning (ML) and Artificial Intelligence (AI) increasingly accelerate this process while improving precision, paving the way towards more individualized and efficient therapies.
Our research team aims to identify novel therapeutic targets for Long COVID / Post COVID (LC/PC) using ML/AI.
Project objectives:
- Integrate and analyze omics and clinical data to create a consistent dataset.
- Investigate pathways altered upon SARS-CoV-2 infection and identification of enduringly active ones in LC/PC syndrome
- Differentiate and classify LC/PC subgroups based on molecular and clinical signatures.
- Establish an AI pipeline to identify potential therapy targets and develop new treatment strategies for LC/PC.
Contribution of the Department of Medical Bioinformatics
Subproject No. 2
AI-based network analysis of COVID-19 blood protomics and transcriptomics: hunting for new therapeutic targets
COVID-19 shows a highly variable disease course – ranging from asymptomatic cases to severe respiratory failure – and many patients continue to suffer from long-term consequences (LC/PC). The underlying mechanisms are still poorly understood.
In preliminary work, PDI was identified as a marker of severe disease, but biomarkers alone are insufficient for developing new therapies, as changes can be both cause and consequence of the disease.
This project therefore aims to integrate molecular data (gene and protein expression) with clinical data and analyze them using advanced probabilistic models. These models can distinguish true biological relationships from spurious or indirect effects.
A central focus is, that the screening for differentially expressed genes from bulk transcriptomics/proteomics can be confounded by cellular compositions, meaning that differential gene expression can be the consequence not just of altered gene regulation but also of different cellular compositions.
Therefore, we will build on our expertise in high-dimensional statistical modeling and cell-type deconvolution to holistically analyze blood transcriptomic and proteomic data and their relationships to clinical variables in COVID-19 patients.
Summary:
• We will establish estimates of cellular compositions from blood transcriptomics data in the NAPKON study using digital tissue deconvolution (DTD) and extensions thereof (6–8). These estimates will serve for improved downstream analysis, revealing disease associated cell populations and gene regulatory mechanisms.
• We will explore cell- and sample specific gene-expression estimates using the TissueResolver. This will resolve gene-expression differences between patient groups (e.g., LC/PC yes vs. no) on the level of cell types from bulk measurements.
• We will use probabilistic graphical models to infer the complex inter-dependencies among transcriptomic and proteomic variables, and their relation to clinical parameters.
• We will combine probabilistic graphical models with cell composition estimates, to better distinguish genuine from erroneous associations in the NAPKON study.
Coordination and Project Partners
Project coordinator:
Prof. Dr. Hassan Dihazi,
University Medical Center Göttingen,
Department of Nephrology and Rheumatology
Prof. Dr. Sabine Blaschke-Steinbrecher,
University Medical Center Göttingen
Emergency Department
Additional project partners:
Prof. Dr. Michael Altenbuchinger
University Medical Center Göttingen
Department of Medical Bioinformatics •
Prof. Dr. Julian Kunkel
Georg-August-University Göttingen Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen (GWDG)
Department of Computer Science
Prof. Dr. Eva Grill
Ludwig-Maximilians-University Munich
Institute for Medical Information Processing,
Biometry and Epidemiology
Prof. Dr. Dagmar Krefting
University Medical Center Göttingen
Department of Medical Informatics