Novel drivers of clinical outcomes uncovered in multiple myeloma
NORWALK, Conn. & CAMBRIDGE, Mass.—The Multiple Myeloma Research Foundation (MMRF) and GNS Healthcare (GNS), a precision medicine company that applies causal machine learning technology to match health interventions to individual patients and discover new target intervention pathways, have announced initial results from a multiyear collaboration that leverages an unprecedented longitudinal observational study to speed the discovery of innovative treatments for patients with multiple myeloma. The organizations presented these findings in a poster session at the American Society of Hematology (ASH) Annual Meeting.
Potential drivers of clinical outcomes and their associated molecular pathways, including some that may be novel, have begun to emerge from the largest and most comprehensive computer models of molecular and clinical interactions in multiple myeloma. These drivers and molecular pathways may represent targets for drug discovery and development, leading to new pharmaceutical strategies that prevent progression of disease and address continued unmet treatment needs of multiple myeloma patients.
The computer models are the product of large-scale, multimodal patient data from the MMRF’s landmark CoMMpass Study and the revolutionary GNS Bayesian causal inference learning and simulation platform REFS (Reverse Engineering and Forward Simulation). Results reflect an analysis by REFS of the CoMMpass Interim Analysis 7 dataset, which is composed of extensive clinical and genomic data for a population of almost 800 enrolled patients.
The CoMMpass study follows 1,000 newly diagnosed patients with active multiple myeloma for eight years. Its objective is to map to clinical parameters each of these patients’ myeloma cells genomic profiles, generated from specimens collected at first presentation and at progression events, to develop a more complete understanding of patient responses to treatments. The study is designed to show which treatments are used most often as first and subsequent lines of therapy, and to correlate this information with critical therapeutic response criteria, including best responses achieved, overall survival, time to disease progression and quality-of-life measures. It is also powered to track treatment data to correspond with genetic information such as mutations and translocations (the movement of a chromosomal segment from one position to another, a phenomenon that often occurs in cancer).
According to Dr. Iya Khalil, co-founder and executive vice president of GNS, Bayesian causal inference aims to learn the probabilistic causal relationships between variables. REFS begins by reverse-engineering the underlying mechanism that gave rise to the data, to find the most likely explanation, then calculates all possible combinations and outcomes from the data and quickly finds causal relationships and patterns that would otherwise take years to discover. The second component of REFS—forward simulation—enables new insights from the models created in the reverse-engineering step by simulating “what if” questions. The focus is on where the answers converge, which allows predictions about which answers are most likely, and knowledge of the level of certainty around those predictions.
“Insights from our work with the MMRF are guiding researchers toward the most likely causes of the disease and targets for drug discovery and development, and are revealing new knowledge about the underlying disease biology in multiple myeloma,” says Khalil. “This work will ultimately improve the ability to match more patients with an effective treatment, to do so sooner, and with more confidence.”
The MMRF and GNS collaboration officially began in 2014, according to Dr. Daniel Auclair, senior vice president of research at MMRF. Auclair tells DDNews, “The first transfer of CoMMpass data occurred in December 2014, which was followed by the formation of a working group in support of the initiative. The initial analysis was delivered this summer, and included data on 452 subjects, which was presented as a poster at ASH. With GNS, we are already working on an updated tranche of molecular and clinical data that should be fully analyzed in early 2016.”
GNS leveraged REFS to reverse-engineer the molecular pathways that affect treatment outcomes in the CoMMpass population and to assess the significance of these pathways in treatment response. REFS employs a hypothesis-free approach, simulating every possible combination and outcome from large, heterogeneous datasets. The hypothesis-free approach of REFS identified a broad range of known disease drivers and biomarkers of response, in addition to the discovery of novel drivers of clinical outcomes and patient response, giving the research team more confidence in the importance of predicted drivers of clinical outcomes and patient response.
“We knew the MMRF was uniquely positioned to gather the immense dataset needed—large-scale, longitudinal data, including outcomes, from patients exposed to a range of interventions,” says Khalil. “We were confident that, once the data was gathered, we could unravel the underlying biological mechanisms of myeloma disease and that we could then use that new knowledge to find novel molecular pathways that can be targeted by pharmaceutical intervention strategies.”
The GNS ensemble of 256 network models built around the 28,200 variables considered for the first modeling captured elements connected to initial response to treatment and identified the cyclin-dependent kinase (CDK) pathway as important, explains Auclair. “Our clinical network, the Multiple Myeloma Research Consortium, has conducted a number of trials with CDK inhibitors, such as palbociclib (Pfizer) and AT519 (Astex). However, these trials did not enrich for patients based on specific biomarkers, which could be a disadvantage in design, given what we are learning. This current data, if confirmed in later, more complete analyses, may suggest new ways to identify patients for trials using CDK inhibitors. This is but one example, with other genes and pathways about which little biology is known, have also been revealed in the dataset. CoMMpass researchers are currently working on validating those findings and will publish those results at the nearest opportunity available.”
There’s another interesting outcome from this research, according to Auclair: “Initial GNS modeling suggests that stem cell transplant (SCT) favorably affects various outcome variables, even with the newer combination regimens that CoMMpass subjects are exposed to. Similar conclusions were also reached by researchers from another prospective study (DFCI-IFM) presented at this last ASH meeting. It is hoped with this data and the results that the application of SCT will be deployed more uniformly across the whole myeloma patient population, and that its contribution towards the ultimate goal of curing myeloma will be fully leveraged.”
The MMRF and GNS will continue to investigate the findings by using future CoMMpass Interim Analyses to validate the significance of predicted novel drivers while refining REFS models. Soon, MMRF and GNS plan to release the computer models for use by associated researchers, clinicians and partners, to facilitate future discoveries.