The changing face of Arabian dust storms
01 June 2023
Published online 30 October 2014
A team of scientists have devised a new computer-based statistical model that is able to detect subtypes of cancer cells within a tumour.
Malignant tumours can contain different subtypes of cancer cells with different sets of mutated genes. However, most existing computer-based models, usually assume that the tumour is made of only one type of cancer cells, failing to detect other subtypes that can potentially grow and even escape chemotherapy treatment.
The international research team’s new model, published in Bioinformatics1, can now estimate and classify the fractions of each subtype of cancer cells in an individual tumour by studying the gene expression. “This model is potentially useful for a physician to recommend a combination chemotherapy that can target all the subtypes of cancer cells in a tumour rather than just the majority subtype,” says lead researcher Patrick Flaherty.
Known as the Gaussian-Laplace-Dirichlet (GLAD) model, the researchers from the Worcester Polytechnic Institute, USA and Al Akhawayn University, Morocco, first used it on computer-generated data of tumours with two subtypes of cancer cells expressing 500 genes and the model successfully detected different cancer cell subtypes.
It was also able to classify a mixed population of liver, lung and brain tissues from rats, by detecting the tissue-specific gene expression. The model also succeeded when applied to experimental data from primary brain tumours containing four subtypes (classical, proneural, neural and mesenchymal) of cancer cells.
Saddiki, H. et al. GLAD: A mixed-membership model for heterogeneous tumor subtype classification. Bioinformatics http://dx.doi.org/10.1093/bioinformatics/btu618 (2014)