Novel Approach Uses Genomic Analysis to Predict Side Effects of Cancer Treatment

Publication
Article
Oncology & Biotech NewsAugust 2012
Volume 6
Issue 8

A new paradigm that relies on identifying networks or clusters of single nucleotide polymorphisms can accurately predict which patients will develop treatment-related mucositis and other side effects.

Stephen Sonis, DMD

A new paradigm that relies on identifying networks or clusters of single nucleotide polymorphisms (SNPs) can accurately predict which patients will develop treatment-related mucositis and other side effects, explained Stephen Sonis, DMD. Sonis discussed the new application of genomics and bioinformatics at the recent 2012 MASCC Annual Symposium.

“How cool would it be if you could predict risk of developing treatment-associated side effects at the time of cancer diagnosis? This is becoming a reality. The stars are now aligned. We have a better appreciation of biological complexity of mucosal injury, and with advances in the human genome project, we are at the precipice of being able to identify patients at risk for mucositis and other treatment-induced side effects,” said Sonis, a founder of InformGenomics and chief of Oral Medicine at Brigham and Women’s Hospital, Boston, Massachusetts.

Single nucleotide polymorphisms are the most common variation in the genome, he explained. There are many more SNPs than genes: There are about 25,000 genes, but over 10 million SNPs.

“Unlike genes, SNPs may or may not be associated with function. This means we are not restricted to looking at genes only involved in coding,” Sonis said.

The old model assumed that a master gene or individual SNP was associated with risk. This model focused on drug metabolism, direct cell response to drug, and bystander biological effects of drugs. The new approach is based on the assumption that clusters or networks of genes or SNPs cooperatively define toxicities.

“Advances in bioinformatics, specifically advanced Bayesian networks, made it possible to take disparate pieces of information to identify clinically applicable findings. The Bayesian network is a novel way to select ‘players’ for a winning team. It allows analysis of one to two million individual SNPs to determine how each of them interacts with another, instead of analyzing what appear to be the best candidates,” he explained. “This method is not restricted to analyzing a predetermined number of ‘players.’”

The old paradigm was a bit arrogant, in that it presumed that we already knew what we were looking for. Findings using the new paradigm are more robust and much more likely to be valid.”

—Stephen Sonis, DMD

“The old paradigm was a bit arrogant, in that it presumed that we already knew what we were looking for. Findings using the new paradigm are more robust and much more likely to be valid,” Sonis said.

He described two recent unpublished studies using SNPs to predict which patients will develop mucositis and other side effects on chemotherapy.

Severe mucositis occurs in about 40% of stem cell transplant patients who receive conditioning regimens with chemotherapy, and there are effective interventions for treatment. “Until now, we didn’t know who would develop mucositis,” he said.

A retrospective study at Dana-Farber Cancer Institute used Bayesian networks to identify which patients would develop mucositis. The investigators examined charts of myeloma and lymphoma patients slated for transplant from 2006 through 2011, identified 153 subjects, and classified them as oral mucositis (OM)-negative (102 patients) and OM-positive (51 patients), defined as having 2 consecutive days of WHO grade >2 OM.

DNA was extracted from patients’ saliva specimens, and 82 SNPs were found to identify OM-positive patients with an accuracy of 99.3%. The investigators then sought to determine if this method was applicable to other patient populations and other treatments.

The OnPART development study, currentlybeing conducted at the West Clinic in Memphis,Tennessee, looked at a broader range of solidtumors, including breast, colorectal cancer,non—small cell lung cancer, and ovarian cancer,in patients treated with at least three cycles ofchemotherapy. Six toxicities are being analyzed:nausea/vomiting, mucositis, fatigue, peripheralneuropathy, cognitive dysfunction, and diarrhea,and these were classified as “none” and “mild”versus “moderate to severe.”

“Let me share preliminary data–that we onlyreceived yesterday–on our ability to predictthe risk of chemotherapy-induced diarrheaamong 30 breast cancer patients treated withadriamycin/cyclophosphamide plus taxane,”Sonis said. “The Bayesian network that wedeveloped predicted which patients woulddevelop diarrhea with an accuracy of 96.7%.”

“These findings are remarkably consistentwith the first study at Dana-Farber in transplantpatients with myeloma,” Sonis said.

“We believe this [method] will be a gamechanger and will let us prospectively evaluateand understand who is at risk for side effectsfrom cancer therapy. This can have a majorimpact on how we address side effects and howwe optimize outcomes for our patients. Thisaffects not only doctors, but patients as well.If you can have a conversation about the risksof side effects, it can help you select the bestregimen for that patient.”

From an economic perspective, pre-selectingpatients at risk for specific side effects can helppayers optimize their resources, and from aresearcher’s standpoint, this can improve patientselection for clinical trials, shrinking the size oftrials to achieve a robust endpoint. “This methodcan make us study things faster,” Sonis said.

Related Videos