GAVIOTA - Generative Analysis based on Variational Inference on multi-Omics datasets for inferring disease Trajectory Architectures
Summary
Current datasets suffer from limited comparability due to technology biases and remain vastly under-exploited. Fast (computationally optimised), interactive frameworks leveraging the generative potential of large datasets are tightly linked to future precision medicine. While recent approaches improved information extraction from rapidly expanding data sources, analyses predominantly focus on identifying differentially expressed genes linked to observed phenotypes. Though generative approaches (e.g. neural networks or deep learning approaches) were attempted, frameworks lack the robustness and scalability to match the pace of data production.
Metabolic dysfunction remains inadequately understood, particularly in the context of obesity and its complications. A key unresolved question is why only certain patients with obesity develop conditions like MASLD (Metabolic Dysfunction-Associated Steatotic Liver Disease), and why only some progress to severe disease. Genetic variants provide only partial explanations for this progression.
Data from academic, healthcare, and industry partnerships, including Astra Zeneca, derived from multivariate studies of biopsies and blood/plasma assays, are vital for developing non-invasive biomarkers that reflect and predict tissue biology. The productive collaboration between wet-lab researchers and bioinformaticians across Dr. Mohorianu's Core Bioinformatics group and AstraZeneca enabled both deep dataset exploration and development of innovative frameworks like Magpie, which optimises the coregistration of spatial transcriptomics and metabolomics datasets and the Muntasir et al 2025 biomarker study.
Project aims
By elucidating the underlying biology and disease drivers of metabolic dysfunction, this work will establish a framework for understanding disease heterogeneity and progression stratification, ultimately creating opportunities for developing targeted treatments that improve patient outcomes and reduce healthcare system burdens.
Contact details
Irina Mohorianu - iim22@cam.ac.uk
Opportunities
This project is open to applicants who want to do a:
- PhD