Breast cancer is a heterogeneous disease, falling into ten major subtypes according to the latest genomic and transcriptomic analyses. Continued research efforts in profiling the molecular characteristics of breast cancer are likely to result in even more stratified classifications. This wealth of biological data may help determine which cancer profiles are best suited to which treatment strategies. One of the major hurdles to be overcome in reaching this goal, is deciphering which elements of these datasets will best inform decisions on particular treatments. Sophisticated computer analyses may help resolve this problem. In a study published in Genome Biology, including Aneleen Darmen, Obi Griffith and colleagues from the laboratory of Joe Gray at Oregon Health and Science University, USA, a computational approach is presented for calculating the type of biological data needed to predict how breast cancer tumours might react to chemotherapy.
The use of chemotherapy in treating breast cancer is sometimes administered before surgery to shrink tumours or after surgery to prevent recurrence of tumours. The sensitivity of a particular breast cancer to a specific chemotherapeutic agent is in part dependent on its molecular profile. Personalised medicine aims to use genetic and molecular data from individual patients to predict which treatment will work best for an individual. How clinicians choose which types of data to pay attention to when making those predictions is however a problem that needs to be solved before personalised medicine can be adopted.
Gray and colleagues used profiles of 70 breast cancer cell lines to develop their method and then tested their approach on tumours from The Cancer Genome Atlas. They looked at the different types of data that could indicate the response of cancerous tissue to 90 different chemotherapy agents, both experimental and approved for use. Data types analysed included measurements of copy number aberrations, mutations, gene and isoform expression, promoter methylation and protein expression.
Transcriptional subtype was shown to strongly impact predictions of response to 25 percent of chemotherapy agents tested. When other biological data was incorporated, response predictions were improved for 65 percent of agents tested. Although no single type of data predicted the response of all tumours to all treatments, a number of genetic and biological properties had to be considered. By analysing the data, the authors determined which of the genetic and other biological data should be considered a ‘signature’ for use in indicating which chemotherapy treatment the tumour type would respond to.
The findings suggest that the application of these techniques in clinical practice may help doctors prescribe the most beneficial chemotherapy treatment to target specific types of breast cancer. The authors also suggest that implementing such a systems biology approach to clinical trials may improve the identification of patient cohorts most likely to respond to a new therapy.