NEW YORK (GenomeWeb) – One of the demonstration projects that is being funded by the California Initiative to Advance Precision Medicine seeks to use bioinformatics infrastructure and analysis capabilities developed by researchers at the University of California, Santa Cruz and elsewhere to help clinicians locate effective therapies for pediatric cancer patients.
The so-called California Kids Cancer Comparison effort is being led by David Haussler, a professor of biomolecular engineering at the University of California Santa Cruz, and Theodore Goldstein, a research associate also at UCSC, and is receiving half of the $2.4 million funding earmarked for projects by the initiative. The UCSC team is working in collaboration with investigators from UC Irvine, UC San Francisco, and Stanford University as well as with industry partners, namely NuMedii, Cisco Systems, and DNAnexus.
The goal of the project is to match patients whose cancers come back, who proved resistant to treatment, or who have little or no treatment options, to novel and previously untried therapies. They'll identify these treatments by analyzing data from individual patients in the context of a much larger pool of data gleaned from other pediatric and adult cancer cases in hospitals and centers within California and abroad. By looking across genetic and treatment information from multiple cancers, the researchers believe that will be able to identify therapies that could potentially work for difficult to treat cancers.
Among other tools, this project will use a web-based platform called MedBook, which combines social media principles and bioinformatics tools, and which connects patients, clinicians, advocates, and researchers and enable them to exchange information and work collaboratively. MedBook was initially developed by researchers at UCSC as part of a three-year prostate cancer research initiative funded by Stand Up to Cancer, the Prostate Cancer Foundation, and the American Association for Cancer Research in 2012.
In the context of this project, MedBook will be used to collect tumor data including mutation, pathway, gene expression, and copy number change data from participating investigators, UCSC's Goldstein told GenomeWeb. Once in the system, researchers will analyze the data in the context of existing tumor datasets gleaned from multiple patients and will then share their findings with clinicians and collaborators across sites. In addition, researchers will create portals to Medbook specifically for patients, families, and advocates so that they have access to analysis results and suggested treatments, and can participate in conversations about their care.
The pediatric cancer project builds on the UC Santa Cruz Treehouse Childhood Cancer Project, which was established to enable the sharing of genomic data from various kinds of pediatric cancer from a variety of hospitals and cancer centers. The system provides a single access point to tumor data from various tissues of origin including whole-genome, exome, and RNA sequencing datasets from the National Cancer Institute's projects as well as from pediatric clinical sequencing trials.
Besides the NCI, contributions come from investigators at the Seattle Children's Hospital; Children's Hospital Los Angeles; British Columbia Cancer Agency; Hospital for Sick Children; Washington University School of Medicine; Texas Children's Hospital; BC Cancer Research Centre; UC Irvine; and Stanford University. Although the emphasis is on pediatric cancer datasets, the Treehouse platform also offers access to data from adult cohorts as well.
Treehouse datasets are accessible through Sage Bionetwork's Synapse platform or the UCSC Cancer Genomics Browser, which provides access to processed datasets from over 200,000 patients including gene expression information, mutation data, and copy number information, Olena Morozova, a postdoctoral scholar in UCSC's Center for Biomolecular Science and Engineering and one of the project leads for Treehouse, told GenomeWeb.
One recent move that aims to ease the burden of moving large datasets from one site to the next involves packaging up the bioinformatics pipelines that UCSC uses to analyze datasets in Docker containers so that it can share them with collaborators, she said. Partners can then run the pipelines internally and transfer processed datasets, which are much smaller than all the raw sequence, to the Treehouse platform. This ensures that no protected patient information leaves the hospital and the data comes in a format that can be combined with existing datasets within the platform.
Combining information from multiple sources is crucial to studying pediatric cancers because they are quite rare compared to the incidence of adult cancers. About 42 pediatric cancer cases are diagnosed per day in the US, according to Morozova — that translates to over 15,000 or so cases per year. In California alone, about 1,800 children are diagnosed per year and treated in multiple hospitals across the state.
Treehouse unites otherwise siloed datasets and makes it possible to combine them with other sources of data. It's a combination that makes sense in light of the current shift towards molecular classification of cancers, away from tissue or age-based classifications, she said. This pan-cancer analysis approach helps researchers identify common pathways and shared mechanisms of action that point to possible therapies that might work for sick children.
In fact, a case where a suggested therapy was actually implemented in practice marked the starting point for the California Kids Cancer Comparison, UCSC's Haussler, also one of the Treehouse project leads, told GenomeWeb. "We found that participants wanted to use it in an immediate clinical setting in addition to a long-term research setting, [and] this inspired us to think ... instead of just having the research program, we should initiate a clinical program, and that's what [the] California kids comparison is all about."
The case in question involved a patient who was part of a clinical trial in British Columbia that sequenced patients' tumors and tried to match them to treatments, Morozova told GenomeWeb. An analysis of genomic and RNA sequence from the patient revealed an offending gene fusion but no existing therapies that could help the patient. Using their bioinformatics tools and expertise, the Treehouse team generated maps that combined the patient's tumor data with data from samples collected from other patients. These maps ultimately showed similarities in sequence from the British Columbia patient's tumor and a subset of neuroblastoma patients, which pointed to a treatment available for the neuroblastoma patients that the researchers believed might work in the BC case.
"We were surprised when the clinician actually decided to act on this observation and actually tried the treatment," Morozova said.
In this case, the treatment only partially worked for the patient — the cancer had metastasized and some metastatic sites got better with the treatment while others worsened — but it represented a paradigm shift for Treehouse, which was initially established as a purely research effort, according to Morozova. "We are experts in bioinformatics and big data, and here we are making suggestions and recommendations that could be then followed up by our clinical colleagues in the clinics."
As a result, the UCSC group now has its sights on clinical applications of its technology, and that's why it launched the cancer comparison project. "We want to do [the analysis] in a more controlled way," Morozova said, adding, "We want to capture information [from patients] so that it helps the next child."
For their first steps under the new initiative, the UCSC researchers will work with collaborators running three clinical trials — they plan to add more trials later. These trials are led by researchers at UCSF, Stanford, and UC Irvine.
The UCSF-led trial — which is currently recruiting patients at 11 sites — focuses on diffuse intrinsic pontine glioma (DIPG), a lethal brain tumor in children that affects the brainstem. According to the DIPG registry website, this particular iteration of cancer accounts for 10 to 15 percent of all brain tumors in children, and there are about 100 to 150 new diagnoses per year in the United States. It is mostly diagnosed in children between 5 and 7 years of age, and fewer than 10 percent of children diagnosed with DIPG survive two years post diagnosis.
Currently, there are no treatment options for patients other than radiation. For the trial, the researchers will biopsy and sequence samples from 15 patients, and then run their pan-cancer analysis approach to find similarities between DIPG and other cancer subtypes, and identify possible alternative therapies to radiation, Morozova said.
The other two trials selected for this initial stage of the project focus on pediatric and young adult patients with relapsed and refractory tumors of various types. The Stanford study expects to have about 100 patients, while the UC Irvine study expects to recruit about 40 patients in the next year but already have data from 90 patients from previous studies, which they hope to include as well.
As with the UCSF trial, researchers in these studies will collect and sequence tumor samples from participants, and then the UCSC team will analyze them in the context of other datasets in the Treehouse database and try to make new treatment recommendations based on their findings.
"We want to improve the yield of all these trials by adding this pan-cancer analysis component," Morozova said. "It's very significant because actually in the pediatric space, analyzing every patient in isolation only helps about 10 percent or less of patients. It's not like adults where actionable mutations are more common. We need this approach to help more patients."
Besides academic partners, the California Kids Cancer Comparison project is working with three companies: NuMedii, Cisco Systems, and DNAnexus. Cisco, for its part, is providing telecommunications systems that will support the group's activities, UCSC's Goldstein said. NuMeddi, a company that has built a business out of identifying new indications for existing drugs, will apply its mathematical models to identify new therapies that might not otherwise be obvious choices for treatment, while DNAnexus is offering its cloud-based bioinformatics infrastructure for processing datasets from the respective sites, he said.
The funds from the California Initiative are expected to last 18 months. The researchers involved in the pediatric cancer project are already seeking an additional $3 million to $5 million in funding, according to Goldstein, to support their activities beyond that timeframe and also to increase the pool of patients that can be helped with the tools.