I develop computational methods and open-source software for analysing large-scale multimodal biomedical datasets. My focus is on spatial transcriptomics, computational pathology, and clinical outcome modelling in IBD.
Spatial OmicsWellcome Sanger Institute (Bayraktar group)
Crypt-level heterogeneity in IBD — multimodal deep learning
Designing multimodal deep learning frameworks integrating GeoMx Digital Spatial Profiling RNA profiles and H&E histology features to characterise crypt-level transcriptional and morphological heterogeneity across Inflammatory Bowel Disease and healthy control patients.
PythonPyTorchGeoMxOpenCVH&E histology
Spatial OmicsHuman Technopole & Politecnico di Milano
IBD tissue microenvironment — spatial transcriptomics
Characterising spatially resolved tissue microenvironments across 180 IBD and healthy control samples using probabilistic cell type deconvolution and NMF-based niche discovery. Building Bayesian classification and regression models linking spatial cellular composition to patient-level clinical outcomes.
PythonScanpySquidpySpatialDataPyroVisium
Computational PathologyHuman Technopole
Whole-slide image analysis in IBD
Computational pathology workflows integrating instance segmentation, supervised tissue classification, and self-supervised representation learning (SSL) to quantify crypt-level morphological heterogeneity across IBD whole slide images.
PythonPyTorchLazySlidescikit-imageOpenCV
ML · Live-cell imagingIstituto Europeo di Oncologia (IEO)
Automated cell cycle phase classification
Developed and validated an end-to-end automated ML pipeline for cell cycle phase classification from live cell imaging data, including image processing, feature extraction, supervised classification, quality control, and reproducible model evaluation. Reduced manual processing time by 80%. Published in eLife.
PythonScikit-learnPyTorchImageJ/FijiNumPy
R PackagePolitecnico di Milano
ScreenR — shRNA screen analysis
An R/Bioconductor package for automated hit identification in shRNA loss-of-function screens, integrating edgeR-based statistical testing within a Tidyverse-native pipeline to process barcode count data at scale.
RBioconductoredgeRTidyverse
GenomicsIstituto Europeo di Oncologia (IEO)
AML epigenome-metabolism linkage
Analysed bulk and single-cell Acute Myeloid Leukaemia data to link metabolic alterations to epigenetic changes using integrative omics approaches.
RPythonSeuratDESeq2scRNA-seqATAC-seq