Genes in canine bone tumors predict drug response: Report

The rapidly increasing data on tumor genomics, coupled with information on drug response, have scientists all over the world building predictive models that might identify which patients are most likely to respond best to certain cancer drugs. But it doesn't always work so well. Recent clinical trials employing such predictive models in patients with lung and breast cancers had to be suspended because the positive results that were seen in lab animals weren't translating to people.

Dogs that develop cancer naturally may help scientists build better predictive models of cancer drug response. That's the conclusion of a new study led by the Flint Animal Cancer Center at Colorado State University. A team there used tumor samples from dogs with osteosarcoma to map out genes that can predict response to the commonly used chemotherapy drugs doxorubicin and carboplatin. The model might prove useful for people who develop the same form of bone cancer.

The researchers used a method developed at the University of Colorado Cancer Center called Co-Expression Extrapolation (COXEN), which combines data on gene expression--or the process by which a gene makes a protein or other product--with drug sensitivity data. Using cell lines from canine tumor samples, they were able to determine which human COXEN models could best predict response to the two chemo drugs in dogs.

They hit upon a model involving genes that are expressed in both canine and human osteosarcomas and showed that it could accurately predict response to doxorubicin in 73% of dogs. They also built a model for carboplatin sensitivity, though it was not as effective, they reported in the journal BMC Bioinformatics.

"It's actually cross-species because the models are being built by using human cancer cell line panels," said Dawn Duval, a professor at the veterinary teaching hospital at CSU, in a recent interview with FierceAnimalHealth. "We're then taking those genes from human cancer cell line panels and building models that can predict sensitivity in canine tumors."

There is a growing movement in the veterinary community to collect and analyze genomic data from dogs, in the hopes of improving treatment outcomes for both dogs and people with a range of diseases that are common in both species. The University of Massachusetts, for example, is recruiting 5,000 dogs for research aimed at linking genes with brain diseases such as anxiety, post-traumatic stress disorder and epilepsy. And last year, researchers at the Broad Institute of MIT and Harvard identified genes in dogs with lymphoma that are also pivotal in human blood cancers.

CSU's veterinarians now planning a clinical trial in dogs with osteosarcoma to test the predictive models they developed. "We're collecting samples from dogs in the clinic and running gene-array analyses on them," Duval said. "Then based on those same models, we'll predict whether they'll be more sensitive to doxorubicin or carboplatin for their osteosarcoma and then treat them with that drug to see if we can improve their response to therapy."

- here's the BMC Bioinformatics report