New Computer Model May Aid Personalized Cancer Care

Dana-Farber Cancer Institute scientists have developed a mathematical model to predict how a patient’s tumour is likely to behave and which of several possible treatments is most likely to be effective.

Reporting in the journal Cell Reports, researchers combined several types of data from pre- and post-treatment biopsies of breast tumours to obtain a molecular picture of how the cancer evolved as a result of chemotherapy.

“Better understanding of tumour evolution is key to improving the design of cancer therapies and for truly individualized cancer treatment,” said Dr. Kornelia Polyak, a breast cancer researcher. The model was developed by Polyak and Dr. Franziska Michor, a computational biologist at Dana-Farber.

The study analyzed breast cancer samples from 47 patients who underwent pre-operative chemotherapy to shrink the tumour so it could be removed more easily. The biopsy samples, representing the major types of breast cancer, included specimens taken at diagnosis and again after the chemotherapy was completed.

As has been increasingly recognized, a tumour contains a varied mix of cancer cells and the mix is constantly changing. This is known as tumour heterogeneity. The cells may have different sets of genes turned on and off – phenotypic heterogeneity – or have different numbers of genes and chromosomes – genetic heterogeneity. These characteristics, and the location of different types of cells with the tumour, shape how the cancer evolves and are a factor in the patient’s outcome.

In generating their predictive model, Polyak and Michor integrated data on the genetic and other traits of large numbers of individual cells within the tumour sample along with maps of where the cells were located within the tumours.

“We asked two questions – how heterogeneity influences treatment outcomes and how treatment changes heterogeneity,” said Polyak.

The computer model cranked out some general findings. For one, the genetic diversity within a tumor, such as differences in how many copies of a DNA segment are present – didn’t change much in cancers that had no response or only a partial response to treatment.

Another result: Tumours with less genetic diversity among their cells are more likely to completely respond to treatment than are tumours with more genetic complexity. “In general, high genetic diversity is not a good thing,” commented Polyak. “The results show that higher diversity is making you less likely to respond to treatment.”
While the genetic diversity of tumour cells was not strongly affected by chemotherapy in patients with partial or no response to treatment, the study revealed that certain types of cells – those more likely to grow rapidly – were more likely to be eliminated, and the locations of cell populations changed.

“Based on this knowledge,” said Polyak, “we could predict which tumour cells will likely be eliminated or slowed down by treatment, and how this may change the tumor overall.” She said this information might help design further treatment strategies for patients who didn’t respond well to the initial therapy.

In the future, said the researchers, cancer doctors may use models of this type to analyze a patient’s tumour at the time it’s diagnosed; the results could help tailor specific drugs and plan treatment strategies matched to the tumour's predicted behaviour.

Publication: Inference of Tumor Evolution during Chemotherapy by Computational Modeling and In Situ Analysis of Genetic and Phenotypic Cellular Diversity. Vanessa Almendro, Yu-Kang Cheng, Amanda Randles, Shalev Itzkovitz, Andriy Marusyk, Elisabet Ametller, Xavier Gonzalez-Farre, Montse Muñoz, Hege G. Russnes, Åslaug Helland et al. Cell Reports (January 23, 2014):

This article originally appeared on