Research

An imaging biomarker for inflammation in Type 2 Diabetes

In metabolic disorders, inflammatory signalling from adipose tissue shifts PBMC chromatin organization: nuclear volume, heterochromatin content and activation increase while chromatin organization heterogeneity and lamin decrease, relative to healthy controls.

Type 2 Diabetes is accompanied by chronic low-grade inflammation. Adipose tissue shifts toward pro-inflammatory signalling and circulating immune cells respond, but measuring that response normally means a panel of clinical assays. We asked whether it is already written into the chromatin of a patient's peripheral blood mononuclear cells, and visible under a microscope. A representation-learning framework reads chromatin organization from PBMC images, including nuclear volume, heterochromatin content, lamin and the heterogeneity of chromatin packing, and recovers the immune alterations that accompany the disease at 83% accuracy, from a sample that is already routinely drawn.

Paper

A 3D in-vitro model of the tumor microenvironment

Under compressive force and TNF-alpha stimulation, cells separate into weak and strong responders, distinguished by the degree of NF-kB nuclear translocation and chromatin condensation.

Cells inside a tumor sit in a soft three-dimensional matrix, under mechanical compression and constant inflammatory signalling. A flat culture dish reproduces none of that. I developed a 3D model that embeds cells in a collagen gel and applies both, compressive force together with TNF-α, and used it to ask how an apparently uniform population responds. It splits in two. Nuclear and cell morphology features, combined with machine learning, separate weak from strong responders, distinguished by how far NF-κB (p65) translocates into the nucleus and how much the chromatin condenses. The difference is legible in cell shape alone, before any molecular readout.

Paper

Correlative light and soft X-ray tomography of heterochromatin

Correlative light and soft X-ray tomography resolves native heterochromatin organization in the mammalian nucleus, showing a phase boundary between heterochromatin domains and surrounding euchromatin at length scales of roughly 80 to 120 nanometres.

Heterochromatin organization at the mesoscale, between individual nucleosomes and whole chromosome territories, is hard to observe, because the two techniques that could see it are rarely applied to the same cell. I established a framework that registers confocal fluorescence microscopy to synchrotron soft X-ray tomography on intact, vitrified cells, with no sectioning or staining artifacts, along with the computational pipeline to segment and quantify chromatin density from the reconstructed volumes.

Paper