Particle.news

Open-Source AI Maps Centrosome Defects Across Breast Tumours

Distinguishing two distinct centrosome states, the tool will be paired with genomic and molecular data to test whether spatial centrosome signatures can become clinical biomarkers.

Overview

  • The University of Southampton released CenSegNet, an open-source AI that maps centrosomes at single-cell, spatial resolution and was used to analyse more than 330,000 centrosomes across 911 tumour specimens from 127 breast cancer patients.
  • CenSegNet separates centrosome abnormalities into two independent defect types: cells with excess centrosomes and cells with enlarged centrosomes, which tend to occupy different regions within tumours.
  • Tumours with high levels of enlarged centrosomes were associated with more aggressive features, including higher grade, lymph node involvement, and specific genetic changes.
  • Researchers have applied the tool to other tissues such as kidney and colon, but CenSegNet remains a research tool and the team plans to combine its spatial outputs with genomic, transcriptomic, and proteomic data to develop and validate centrosome‑based biomarkers.
  • Centrosome abnormalities have long been seen as a cancer hallmark, yet direct study in patient tissue was hard; this AI approach makes large-scale, single‑cell mapping possible but requires independent validation before it can change patient care.