New York (GenomeWeb) – Researchers from the Monash Biomedicine Discovery Institute at Monash University and elsewhere have developed a computational model that they claim can identify and prioritize the best combination of drug therapies to potentially combat resistant tumors.
According to a recent PLOS Computational Biology paper describing the approach, the researchers used the model to rank various combinations of drugs based on their likelihood of defeating triple-negative breast cancers by targeting signaling pathways that are involved in drug resistance in tumors. They also used the model to analyze genomic information from patients to identify those that are most likely to benefit from treatment with a particular drug combination, saving time and sparing patients from being subject to ineffective treatments.
The developers believe that their model has the potential to someday be used at the bedside to personalize treatments for patients based on their genomic profiles. Currently, there is no approved targeted treatment for triple-negative breast cancer cases, though there are several signaling pathways that have been considered as targets, such as the EGFR pathway, which is overexpressed in about 50 percent of breast cancer patients, according to Lan Nguyen, the head of the Integrated Network Modelling Laboratory at Monash Biomedicine Discovery Institute and one of the developers of the model.
However, treating patients with EGFR inhibitors alone does not work efficiently, he said in an interview. This is in part due to the fact that triple-negative breast cancer cells can develop resistance to a single targeted drug by rerouting the signaling pathways within the cells. Currently, patients with the disease are treated with chemotherapy, but there is no guarantee of success, Nguyen notes. Moreover, a pilot study performed by researchers from the University of Texas MD Anderson Cancer Center and Sweden’s Karolinska Institute found that resistance in triple-negative breast cancer cases can in some cases arise through acquisition of genomic mutations that occur as a result of chemotherapy.
Combination therapies offer a viable alternative for patients with these kinds of resistant tumors, and existing research shows that many in the oncology community are exploring the field. One study performed by researchers from Stanford University that compared combination-based clinical trials to non-combination-based clinical trials found a higher prevalence of combination therapy in oncology (25.6 percent used combination trials) compared to other disease trials (6.9 percent used combination trials). Furthermore, according to the American Society of Clinical Oncology, combination targeted therapies will be the standard of care for most cancers in the future and these treatment combinations will be tailored to individual patient's molecular profiles.
However, one of the key challenges of identifying combination treatments for cancer is that "we don't know how to predict the optimal drug combination that can overcome this kind of adaptive resistance," Nguyen said in an interview. Computational models, such as the one used by his team, help focus the search for therapies on combination candidates that are most likely to be effective. This way, researchers only have to test a subset of drug candidates and can connect patients to treatments faster.
Besides triple-negative breast cancers, using computational models to identify drug combinations is being explored in other contexts. For example, in 2014, researchers involved in the DREAM Consortium launched an open challenge that aimed to encourage the development of computational modeling methods for ranking pairs of drug compounds for use in cases of human diffuse large B-cell lymphoma. Although the focus of this study was triple-negative breast cancer, Nguyen's team's model can be adapted and used to determine effective drug combinations for other kinds of cancer such as lung and melanoma that also show drug resistance based on alterations in signaling pathways, he said.
As explained in PLOS Computational Biology, Nguyen and his colleagues built a model of the EGFR-PYK-c-MET interaction network in triple-negative breast cancer cells. They then tested six combinations of four small-molecule inhibitors — Gefitinib, PF396, EMD, and Stattic. When they tested the model in cell lines, their results showed that drug combinations that co-inhibit EGFR-PYK2 and EGFR-c-MET pathways had the strongest effect.
The researchers also looked at patient-derived gene expression data from 108 patients from the Cancer Genome Atlas project. Using the model, they were able to stratify patients into subgroups based on their likelihood of responding to combined therapies. "[It] gives us a more rational framework to work with rather than doing trial and error working with different combinations. The model gives us more rationalized decision making, and that is more likely to lead to something more tangible and shortcut the translation from the research lab to the [bedside]," Nguyen said
The researchers hope to begin testing the predicted drug cocktails in clinical trials within two to five years. They have already begun working on patient-derived xenographs in mice, which they will use to test the efficacy of the predicted drug combinations. If the suggested treatments work in mice, the next step would be to explore clinical trials that enroll patients based on particular molecular features that are similar to those used in the models, Nguyen said.
Ultimately, he hopes to see the model incorporated into an app that clinicians can use at the bedside to make more informed treatment decisions for their patients. But such an app is several years away from being a reality. The focus for now is on improving and refining the predictive power of the models.