Network biology approaches towards the identification of predictive cancer biomarkers
Schulc Klára
Molecular Medicine Division
Dr. Várnai Péter
SE, Belgyógyászati és Hematológiai Klinika
2026-10-08 13:00:00
Pathobiochemistry
Dr. Ligeti Erzsébet
Dr. Csermely P. és Dr. Veres D
Dr. Fekete Tibor János
Dr. Czakó Bence
Dr. Wiener Zoltán
Dr. Bihary Dóra
Dr. Szaniszló Tamás
Cancer is a disease where malignantly transformed cells use the available resources for self-replication instead of their original functions, evading immune response and therapeutic efforts. Therapy resistance is one of the largest challenges in oncology today, leading to unnecessary treatments and larger mortality. Signalling changes marking therapy resistance in cancer can be studied in multiple ways. In this work, we used in silico approaches to model cancer signalling in carcinogenesis and therapy resistance, and both in vitro and in silico methods to identify predictive biomarkers.
Network topology analysis was used to model colon cancer development, using open-source RNA microarray data to create normal, adenoma and carcinoma network models from a pre-published network skeleton. The results show a larger network diameter indicating loosened signalling structure in colon adenoma, and a smaller network diameter pointing to strictly organised signalling in colon carcinoma. While the VEGFR pathway becomes more expressed with carcinogenesis, the EGFR and MAPK pathway models have higher weighted degree and median abundance in normal network models than in the adenoma network models.
Network motifs are small, frequently present network structures indicating strong co-regulation of the participating nodes. Here, we showed that intrinsically disordered proteins, difficult to target, but central players in signalling networks, are enriched in the triangle motifs containing onco-therapeutic targets. Using network topology features such as motif information and protein annotations including IDP databases, a powerful biomarker prediction tool called MarkerPredict was established on target-neighbour pairs. Validation showed high performance, and 2084 potential biomarkers were predicted using the established Biomarker Probability Score. These predictions are shown to be highly relevant, waiting for further validation before clinical use.
LncRNAs are promising biomarker candidates, as they can not only be detected in tumours, but in bodily fluids like peripheral blood, promising minimal-invasive therapeutic decision-support. A pan-cancer MAPK-pathway-specific lncRNA signature was established with RNA sequencing on lung and liver cancer cell lines with effective but not lethal MAPK pathway inhibition, ready for validation on patient samples.
