October 1, 2016 (Vol. 36, No. 17)
Lisa Heiden Ph.D. Director of Business Development MyBioSource
Data Never Rests. It Is Constantly Being Updated, Aggregated, Analyzed, and Reassessed in New Clinical Contexts
Pharmacogenomic precision medicine can now be incorporated into routine clinical practice according to Martin Dawes, M.D., professor and head of family medicine, University of British Columbia (UBC). TreatGX software, developed at UBC, integrates a patient’s current physical state, medications, and pharmacogenomics.
“Within seconds, the software produces a list of safe and effective medication options for an individual patient,” exclaims Dr. Dawes.
Abigail T. Berman, M.D., assistant professor at the University of Pennsylvania and associate clinical director at the Penn Center for Precision Medicine, reflects on precision medicine from the patients’ point of view: “If something has been proven, even if it was just proven yesterday, clinicians have the job of rolling it out very quickly into clinical care,” she insists. Alessandra Cesano, M.D., Ph.D., chief medical officer, NanoString Technologies, concurs: “The revolution of precision medicine is not just in the new science; it is also in the application of these scientific discoveries to the clinic so the patient benefits.”
Speeding up the translational research continuum and implementing precision medicine initiatives were key themes at the Global Engage and GTC Precision Medicine Conferences held in London and Boston, respectively. Highlights included strategies and technologies to exploit the vast and ever-expanding amount of scientific data and information. Some of the outstanding conference presentations are discussed herein.
Interrogative Systems Biology
Traversing the translational research continuum has historically been a lengthy process. “BERG combines high-throughput molecular profiling and an artificial intelligence-based analytics system to expedite the process of understanding how molecules interact in human cells or the body,” says Leonardo Rodrigues, Ph.D., associate director, advanced analytics, BERG.
“The BERG Interrogative Biology™ platform operates with any sample-related data including clinical information, treatment outcomes, and high-throughput molecular data such as proteomics, lipidomics, and metabolomics,” details Dr. Rodrigues. Trillions of data points are generated from single samples that may include thousands of compounds such as proteins, lipids, and metabolites. Also, data points may reflect various kinds of biological activity.
Essentially, the BERG platform integrates systems biology with artificial intelligence and machine-learning algorithms to analyze data and generate unbiased multi-omics output. “Our platform can be used for the discovery of novel therapeutics and biomarkers,” explains Dr. Rodrigues. “We have used it extensively to build our own pipeline.”
Discovered by an Algorithm
BPM 31510 is the first drug developed through artificial intelligence, asserts Dr. Rodrigues. BPM 31510, a coenzyme Q10-containing proprietary formulation, may reverse the Warburg effect associated with altered lipid metabolism in cancer cells and elicit anticancer responses such as apoptosis.
BERG has clinical trials involving its BPM 31510 drug, including a Phase I trial open for all solid tumors and a Phase II trial for advanced pancreatic cancer.
The clinical trials incorporate precision medicine approaches through the BERG Interrogative Biology platform. For example, patients’ biological samples such as blood, urine, and tissue, along with clinical information, are analyzed, and integrative cause-and-effect maps are developed. The maps can help determine which patients might respond to a given drug or predict adverse events.
Causal Machine Learning
GNS Healthcare’s causal machine learning platform REFS™ (reverse engineering and forward simulation) turns healthcare data into actionable insights to guide precision medicine and scientific discovery. “We reverse engineer causal models from healthcare data,” explains Diane Wuest, Ph.D., associate director, precision medicine initiatives, GNS Healthcare. “Once these models are built, they can be forward simulated and queried to answer ‘what if’ scenarios on outcomes.”
The platform aggregates all data types found in healthcare environments, such as electronic medical records and clinical, mobile, and omics data (genomic, proteomic, metabolomic, etc.), and builds unbiased causal disease models.
“We input billions of data points that represent millions of variables and their interactions,” continues Dr. Wuest. “Then we conduct analyses to discern complex causal mechanisms and predictors across populations and on individual levels.” REFS models can be built in days to weeks depending on the data types and size available.
Models can predict patient outcomes. For diabetes, certain predictors of progression have been identified based on patient registries, making it possible to benchmark a new patient against the models to anticipate the progression of diabetes in that patient. Web-based dashboard tools help clinicians predict if the patient is at high or low risk for diabetes progression and inform treatment interventions.
REFS data-driven modeling technology is increasingly being used for translational scientific research and identifying clinically relevant information.
“With any of the models that we build, typically about one-third of the insights are previously known,” informs Dr. Wuest. “Another third, also known, may seem remote from the exact disease area and hence of questionable relevance, at least until they are given a bit of thought. The final third are novel pieces of information that need additional investigation.”
All three insights are represented in the results of a collaboration between GNS Healthcare and the Multiple Myeloma Research Foundation, which analyzed the CoMMpass Study™(NCT01454297). Results identified were the known RN7SK (7SK RNA), the partially known PDXP (pyridoxal phosphatase), and the new MIR3648-1 (MicroRNA 3648-1) disease-associated markers.
Implementing Actionable Knowledge
“It is our job as clinicians to make sure that every cancer patient who walks in the door is tested appropriately for actionable mutations,” states Dr. Berman. Given her position as a radiation oncologist, Dr. Berman has a particular perspective on the draft recommendations for lung cancer mutation testing that have been issued by the International Association for the Study of Lung Cancer (IASLC).
“There is ongoing debate about what individual institutions should be doing about their depth of testing and exactly how mutation testing should be done,” comments Dr. Berman. “This is why IASLC opened a paper for public comment. Mutation testing is a constantly changing dynamic because the research framework on what mutations are actionable or prognostic is constantly changing.”
The Penn Center for Precision Medicine’s next-generation sequencing (NGS) panel for lung cancer is based on a combination of actionable/druggable and prognostic mutations. Mutations have been identified in about 75% of over 800 lung cancer cases, with targeted drugs available for about 20% of the mutations. A retrospective study was done to find out if patients actually got available therapies. According to the study, not only did patients with common mutations get targeted drug therapy, but so did patients with more recently described mutations such as MET.
“The study indicates that doing advanced testing and being ahead of the curve is really critical for patient care,” remarks Dr. Berman. “Appropriate testing opens doors for patients that would otherwise remain closed,” remarks Dr. Berman.
Clinical Trials for All
The clinical trials landscape is changing. “Instead of the typical Phase I/ II/III trial design, umbrella and basket trials are emerging,” notes Dr. Berman. One umbrella trial might enroll patients with different tumor types, and multiple mutations and drugs may be included in a basket trial.
Dr. Berman strives to get all patients on a trial. One trial uses flaxseed (NCT02475330), which contains the active component secoisolariciresinol diglucoside, a compound that potentially reduces radiation damage to the lung.
“The role of radiation in stage IV lung cancer is highly debatable unless a patient is has having symptoms,” she informs. One project aims to determine if mutation status can help define who should get radiation in stage IV disease.
Evidence-Based Prescribing Systems
The world faces a population that is aging and likely to show increases in chronic diseases and medication complexity. According to Dr. Dawes, every year more than one-third of patients over 65 living in developed countries using current prescribing systems are prescribed ineffective or dangerous medications.
“Current prescribing systems are leading to adverse events that represent the fourth leading cause of death in the United States,” he warns.
The evidence-based prescribing decision support system TreatGX “not only deals with genetic information, but also other complexities that arise when clinicians attempt to identify safe and effective treatment options for patients,” asserts Dr. Dawes.
Pharmacogenomics in Routine Care
TreatGX currently includes 24 common conditions such as depression, diabetes, and hypertension. A cohort validation study conducted in primary care settings with 191 adults found 97% (185/191) had at least one actionable mutation for medications included in TreatGX. Strikingly, dose-corrected drug options and potential changes in prescription recommendations were indicated for 97% (179/185).
Patients and clinicians enter data and information into TreatGX, automatically incorporating it into complex algorithms and generating results. If a patient has a gene variant, TreatGX will likely present a different set of options compared to a patient who has the same clinical diagnosis but a different variant.
The system continually integrates new information into its capabilities. “We start with guidelines because they are the highest level of evidence,” details Dr. Dawes. “Then we consult Cochrane Library’s systematic reviews and Pub Med for randomized control trials.” For adverse events, the system draws upon monographs from company literature and published information from pharmacogenetics groups.
“Our system, concludes Dr. Dawes, “allows people for the first time ever to see the options for an individual patient based on their personal characteristics including genetics.”
Translational Medicine Takes a Village
Collaborations are essential for translating new technologies to the clinic. NanoString collaborates with pharmaceutical companies to develop companion diagnostics. Dr. Cesano says the pharmacogenomics assay for the application of the drug enzalutamide (an androgen receptor signaling inhibitor) in triple-negative breast cancer was first developed by Medivation/Astellas researchers using an RNA sequencing platform.
The test was moved to NanoString’s platform for companion diagnostic development because “our platform is robust, high-throughput, already used in clinical settings, and based on an FDA-cleared instrument (in the context of the Prosigna® assay),” she emphasizes.
The assay is based on the same gene set as NanoString’s already commercialized Prosigna, but uses a different algorithm to identify metastatic triple-negative breast cancer patients who could benefit from enzulutamide. The Prosigna Breast Cancer Prognostic Gene Signature Assay assesses risk recurrence of early-stage, hormone-receptor-positive breast cancer.
Different gene sets identify other clinically actionable targets. An assay in collaboration with Celgene looks at cell-of-origin in diffuse B-cell lymphoma to identify patients who could benefit from treatment with lenalidomide, an immunomodulatory drug, in combination with R-CHOP chemotherapy as first-line therapy.
Digital Technology Counts
NanoString’s digital technologies for identifying clinically actionable information are based on “counting nucleic acids in a digital format from research to clinical applications,” says Dr. Cesano.
“Detection relies on coordinated binding of two probes to a target,” she explains. One probe contains the digital barcode, while the other probe contains a biotin moiety. Binding of both probes to a target molecule allows for counting. The nCounter® instrumentation automates all steps including reporter/barcode purification, capture (via the biotin tag), and detection (via the barcode).
Unlike RNA sequencing, NanoString technology avoids amplification. “Amplification introduces a lot of potential errors and variability,” informs Dr. Cesano. “When you avoid amplification, you count only what is actually present in your sample.”
“When developing assays not just for research but also for clinical use, control of variability becomes extremely important,” she emphasizes.
The nCounter instrument carries out assays in a standardized and regulated manner limiting human interaction, which is also important for controlling variability and developing robust, reproducible assays.
“Unfortunately, the value precision medicine tools bring to the clinic has not been fully recognized from the reimbursement part of the world, ” sighs Dr. Cesano. Until that happens, the implementation of precision medicine will be spotty.
Autoantibody Discovery Using High-Throughput Screening
Diagnostic biomarkers are powerful decision-making tools that can enable more accurate diagnoses and better disease stratification for improved clinical management of patients. One class of biomarkers are autoantibodies (AAB), which are expressed early in a number of autoimmune diseases and can be indicative of disease stage and severity.
The human autoimmune profile covers a huge number of AABs, which provide an enormous resource to identify novel marker candidates. This resource can be utilized for the better identification of diagnostic biomarkers allowing for the improved stratification of patients based upon their disease profile.
Autoimmune diseases are influenced by a range of factors, meaning it is unlikely that a single given biomarker will be useful for patient stratification.
“Researchers, therefore, are utilizing multiple biomarker identification tools such as Protagen’s SeroTag® biomarker development engine for the high-throughput screening of AABs in autoimmune diseases such as systemic lupus erythematosus (SLE), systemic sclerosis (SSc), and rheumatoid arthritis (RA),” said Peter Schulz-Knappe, M.D., CSO of Protagen Diagnostics. “In one study, investigators used thousands of human autoantigens to develop a comprehensive landscape by which disease subgroups could be identified, based upon their highly differentiated autoantibody signature.”
Screening over 4,000 serum samples, a bead-based suspension array was coupled with advanced data mining to identify novel autoantibodies for SLE, SSc, and RA. Researchers also found that patients fell into one of two categories; either belonging to clusters defined by particular characteristic markers, or phenotypically overlapping with one another.
“Such technologies present a highly valuable tool for novel biomarker discovery, which will enable earlier diagnosis, differential diagnosis, and disease subgrouping,” added Dr. Schulz-Knappe.