Determining whether a gene is mutated in a tumour may not be enough. In many cases, it is also important to know how many copies of that mutated gene are present in the cancer cells and how strongly this alteration affects the behaviour of the disease.

This is the starting point of the study led by Giulio Caravagna, Professor of Computer Science at the University of Trieste and head of the Cancer Data Science Laboratory, published in Nature Genetics. The researchers proposed a new way of reading tumour DNA that more closely reflects the true complexity of cancer: mutations are considered not only in terms of whether they are present or absent, but also in relation to the tumour's overall genomic architecture.

In cancer cells, somatic mutations - those that arise during a person's lifetime in cells other than germ cells - can combine with changes in DNA copy number. As a result, some regions of the genome may be amplified while others are lost. The combination of mutations, amplifications and losses of genomic regions gives rise to the concept of gene mutant dosage. The aim is to estimate how many mutated copies of a gene are actually present in a tumour and assess whether this information has implications for prognosis and metastatic spread.

To address this question, the research team developed INCOMMON - INference of COpy number and Mutation Multiplicity in ONcology - an open-source computational method that estimates the number of mutated copies of a gene and mutation multiplicity from targeted sequencing panels routinely used in clinical oncology.

One of its most significant features is the ability to derive new information from data already generated during diagnostic workflows. INCOMMON uses advanced statistical inference and machine-learning models to interpret the results of clinical molecular tests in the light of large collections of tumours sequenced at whole-genome level.

“INCOMMON represents a new approach to obtaining higher-resolution analyses than current standards, without changing the type of data generated. This new opportunity reveals features of oncogenic processes that may prove relevant from a prognostic and predictive perspective,” explains Giulio Caravagna.

In the study, supported in part by the AIRC Foundation for Cancer Research, the researchers analysed more than 60,000 clinical patient samples, covering 39 different cancer types and over 500,000 mutations. By stratifying more than 20,000 patients according to the mutant dosage of different oncogenes and tumour suppressor genes, the team identified 46 cancer-specific biomarkers associated with overall survival.

The most significant finding is that 13 of these biomarkers, across 12 cancer types, would not have been detected using conventional models based solely on whether a gene is mutated or not. Gene mutant dosage can therefore reveal prognostic signals that a simpler reading of tumour DNA may miss.

The findings also provide new insight into the metastatic behaviour of tumours. In particular, the researchers identified 26 biomarkers recurrently associated with metastatic spread in 10 cancer types, as well as 20 biomarkers whose presence predicts a tumour's tendency to spread to specific organs in five cancer types.

“The data collected in this study show that genomic information already contained in molecular analyses routinely performed in clinical practice can be reinterpreted with far greater precision using advanced statistical inference and machine-learning systems,” says Nicola Calonaci, an AIRC-funded postdoctoral researcher at the University of Trieste and co-first author of the paper.

“One particularly interesting aspect is that gene mutant dosage seems to capture biological properties of tumours that do not emerge when mutations are treated simply as present or absent,” adds Eriseld Krasniqi, a medical oncologist at the IRCCS Regina Elena National Cancer Institute in Rome, a University of Trieste PhD candidate and co-first author of the paper. “This is especially relevant to the study of metastasis, where it has historically been very difficult to identify genomic features that distinguish primary from metastatic tumours.”

“This research is an excellent example of how collaboration between highly specialised clinical centres and computational research groups can translate into knowledge that benefits cancer patients,” says Giovanni Blandino, Scientific Director of the IRCCS Regina Elena National Cancer Institute (IRE) in Rome. “IRE's participation in a study of this scale reflects our commitment to working at the frontier of translational research, where clinical expertise meets the most advanced technologies for genomic data analysis. Extracting new information from molecular tests already available in clinical practice is one of the most promising directions for precision oncology, and an area in which we intend to continue investing.”

The work also opens the way for future studies integrating information on the treatments received by patients, to determine whether gene mutant dosage has not only prognostic value but also predictive value for response to specific therapies.

Further validation will be required before any potential clinical implementation. The direction indicated by the research is nevertheless clear: to make better use of information already available in molecular data, in order to understand disease evolution and contribute to the development of new biomarkers for precision oncology.

*********************************
STUDY PUBLISHED IN NATURE GENETICS, 31 JULY 2026

Gene mutant dosage is associated with prognosis and metastatic tropism in 60,000 clinical cancer samples

Nicola Calonaci1,†, Eriseld Krasniqi2,†, Daniel Colic3, Stefano Scalera4, Giorgia Gandolfi1, Salvatore Milite5, Konstantin Bräutigam6,7, Andrea Sottoriva5, Trevor A. Graham6, Leonardo Egidi8, Biagio Ricciuti9, Marcello Maugeri-Saccà10, and Giulio Caravagna1,11

1 Department of Mathematics, Informatics and Geosciences, University of Trieste, Italy
2 Phase IV Clinical Studies Unit, IRCCS Regina Elena National Cancer Institute, 00144 Rome, Italy
3 Institute for Research in Biomedicine, Barcelona, Spain
4 Clinical Trial Center, Biostatistics and Bioinformatics Division, IRCCS Regina Elena National Cancer Institute, Rome, Italy
5 Computational Biology Research Centre, Human Technopole, Milan, Italy
6 Centre for Evolution and Cancer, Institute of Cancer Research, London, United Kingdom
7 Institute of Medical Genetics and Pathology, University Hospital Basel, Basel, Switzerland
8 Department of Economics, Business, Mathematics and Statistics “Bruno de Finetti”, University of Trieste, Italy
9 Lowe Center for Thoracic Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA, USA
10 Division of Medical Oncology 2, IRCCS Regina Elena National Cancer Institute, Rome, Italy
11 Area Science Park, Trieste, Italy
† These authors contributed equally.

DOI: 10.1038/s41588-026-02666-z