A single tumour, or even a single cancer cell, can present multiple genetic alterations at the same time: point mutations, lose or gain of specific genetic fragments, among others. Understanding how these alterations determine a differential gene expression, called ‘transcriptional identity’, is a key aspect to decipher the characteristics of a tumour, such as the differentiation state of cancer cells, its capacity to develop metastasis or to hide from the immune system. Defining these specific characteristics would show light about the particular vulnerabilities of each cancer and how to treat them efficiently.
A team from the Columbia University in New York, with the participation of Dr. Álvaro Aytés, first author of the paper and group leader at the Bellvitge Biomedical Research Institute (IDIBELL) and the Catalan Institute of Oncology (ICO), have analysed the genomic data of 10.000 tumours from the 20 most frequent cancer types, by means of computational algorithms.
The study, published in the journal Cell, identifies up to 112 cancer subtypes defined by 407 master regulator proteins that channel the information of the differential gene expression. Moreover, the analysis reveal that the 112 subtypes can be classified according to the activation/inactivation protein states of 24 unique master regulator modules. These means that the combination of 24 basic features of a tumour can define the cancer subtype, enabling a more reliable prognosis of the patient.
Dr. Aytés states that ‘these results show a hierarchical control of the properties that define each cancer type’, and continues ‘the activation or inactivation state of each of the 24 regulatory modules here identified, define the transcriptional identity of a tumour, enabling a better understanding and treatment.’
Dr. Andrea Califano, head of the Systems Biology Department of Columbia University, and leader of this work, claims that ‘to date, personalised medicine has focused on which gene, among thousands, is responsible of a disease and hoped for an available drug to fight it.’ However, he also adds that ‘this study suggests that instead of searching for a drug for each gene, we could work with a limited number of drugs to act against the here defined 24 regulatory modules.’
Researchers have validated some of these basic features described in prostate or kidney cancers, among others. Using genetic edition techniques and pharmacological treatments, they confirmed that the algorithms can predict accurately the experimental models.
Noteworthy, all information and mathematical models of this study have been immediately placed at disposal of the scientific community through a web application. The platform allows any researcher to analyse genomic data of cancer patients with these new predictive models.