CSIRO’s new statistical methods enable analysis of very complicated microarray data.

Stats outsmarting brain tumours

Reference: 07/85

CSIRO and the Kolling Institute of Medical Research at Sydney’s Royal North Shore Hospital have discovered two new biological markers that can identify different types of aggressive brain tumours.

16 May 2007

The discovery is announced this month in the Journal of Neuropathology and Experimental Neurology. It will help neuropathologists more accurately diagnose brain tumours and better predict patient survival. Diagnostic accuracy is vital in selecting appropriate treatments for individual patients, leading to better treatment outcomes.

“CSIRO’s statistical techniques have advanced the Institute’s brain tumour research by up to five years,” says Dr Kerrie McDonald of the Kolling Institute.

Using samples of brain tumours and normal brain tissue obtained from the Australasian Brain Tumour Bank at Royal North Shore Hospital, Dr McDonald’s team measured the expression of thousands of different genes.

CSIRO statistician Maree O’Sullivan says: “Typically, researchers will observe hundreds of different gene expression changes in tumour samples.

“The genes identified in the study will also become potential targets for new treatments and could help predict patient survival times.”

“Our statistical techniques are able to sort through the data and pinpoint very small sets of genes that reliably differentiate between tumour types.”

Dr McDonald says her team identified a signature gene expression pattern of 18 genes that is unique to aggressive brain tumours and absent in normal tissue.

As well as vastly speeding up the research, the discovery makes developing cheap and fast diagnostics kits a possibility. Currently, brain tumours are categorised by examining sections of the tumour under a microscope. A gene-based diagnostic test could be a valuable support to this.

The genes identified in the study will also become potential targets for new treatments and could help predict patient survival times.

“Two of the genes have been validated at the protein level and we have found that expression levels are associated with survival outcome,” Dr McDonald says.

CSIRO is also using its proprietary statistical techniques to develop diagnostic and prognostic tools for other types of cancer, including predicting survival time for breast cancer patients and diagnosing subtypes of paediatric acute lymphoblastic leukaemia to enable individualised treatment.