Researchers at the Szegedi Biológiai Kutatóközpont (SZBK), part of the Hungarian Research Network (HUN‑REN), have developed an artificial intelligence (AI)‑based workflow that links microscopic visual features with molecular profiling to study individual cancer cells. The work was carried out by the Lendület Microscopic Image Analysis and Machine Learning Research Group led by Horváth Péter, in international collaboration.
The project addresses tumour heterogeneity: a tumour consists of multiple cell populations that can behave differently. Some cell groups present at diagnosis may later seed metastases or resist therapy. Detecting these critical subpopulations early is challenging because conventional molecular assays often analyse bulk tissue and average out localized differences.
Technology and method
The SZBK team used an automated single‑cell research facility they established. This system can isolate AI‑identified cells around the clock without human intervention and perform downstream molecular analyses. The approach is based on Deep Visual Proteomics: AI first analyses histological images, recognizes relevant cell types or clusters, and those cells are then precisely excised from the sample with a thin laser beam for proteomic profiling.
What the Szeged group added is parallel single‑cell genetic analysis: for the image‑selected tumour cell groups they measured not only protein composition but also which genes are active in the same cells. This dual readout — morphology linked to both gene activity and protein expression — enables a more complete understanding of how individual cells function.
Application to renal cancer
In a report published in EMBO Molecular Medicine, the researchers applied the method to samples from clear cell renal cell carcinoma, the most common and deadliest kidney cancer type. They say this is the first time that both genetic and proteomic composition of tumour clones were measured at single‑cell resolution in this cancer. The combined analysis reveals how morphologically distinct regions of a tumour differ biologically.
Why this matters
Tumour cell aggressiveness arises from multiple interacting factors: gene activity, protein function, metabolism and immune interactions. Morphology alone does not fully explain a cell population’s behaviour. By assessing the same cell groups for both gene expression and protein-level function, the method allows more precise identification of processes that could drive metastasis or therapy resistance.
The SZBK’s automated infrastructure together with AI‑guided Deep Visual Proteomics thus enables detailed single‑cell mapping of intratumour heterogeneity, which may support development of more targeted diagnostics and treatment strategies over time.
