Hundreds of cardiac MRI traits derived using 3D diffusion autoencoders share a common genetic architecture.
Emanuel
Soda
Computational scientist building ML and Bayesian models for multimodal biomedical data — spatial transcriptomics, single-cell genomics, and digital pathology.
PhD Researcher at Human Technopole & Politecnico di Milano
Visiting at Wellcome Sanger Institute (Bayraktar group)

About
Decoding biology
through computation
I'm a computational scientist with 5+ years of experience developing machine learning, Bayesian statistical models, and reproducible data pipelines for multimodal biomedical data. My PhD focuses on characterising the spatial tissue microenvironment in Inflammatory Bowel Disease using Visium spatial transcriptomics, computational pathology, and clinical outcome modelling.
Currently a visiting PhD student at the Wellcome Sanger Institute (Bayraktar group, Hinxton, UK), where I'm designing multimodal deep learning frameworks integrating GeoMx RNA profiles with H&E histology features to characterise crypt-level heterogeneity in IBD.
I believe in open, reproducible science — most of my work ships as open-source tools and public pipelines.
Research Interests
Selected Work
Research highlights
IBD tissue microenvironment
Characterising spatially resolved microenvironments across 180 IBD and HC samples using probabilistic deconvolution and NMF-based niche discovery.
→Deep LearningGeoMx + H&E multimodal framework
Deep learning frameworks integrating GeoMx RNA profiles and histology features to characterise crypt-level heterogeneity in IBD at the Wellcome Sanger Institute.
→ToolScreenR — R/Bioconductor package
Automated hit identification in shRNA loss-of-function screens, integrating edgeR-based statistical testing within a Tidyverse-native pipeline.
→Publications
Recent papers
Automated workflow for the cell cycle analysis of (non-)adherent cells using a machine learning approach.