Mia A. Levy

Computational Phenotype Discovery Using Unsupervised Feature Learning over Noisy, Sparse, and Irregular Clinical Data

Inferring precise phenotypic patterns from population-scale clinical data is a core computational task in the development of precision, personalized medicine. The traditional approach uses supervised learning, in which an expert designates which patterns to look for (by specifying the learning task and the class labels), and where to look for them (by specifying the input variables). While appropriate for individual tasks, this approach scales poorly and misses the patterns that we don’t think to look for. Unsupervised feature learning overcomes these limitations by identifying patterns (or features) that collectively form a compact and expressive representation of the source data, with no need for expert input or labeled examples. Its rising popularity is driven by new deep learning methods, which have produced high-profile successes on difficult standardized problems of object recognition in images. Here we introduce its use for phenotype discovery in clinical data. This use is challenging because the largest source of clinical data – Electronic Medical Records – typically contains noisy, sparse, and irregularly timed observations, rendering them poor substrates for deep learning methods. Our approach couples dirty clinical data to deep learning architecture via longitudinal probability densities inferred using Gaussian process regression. From episodic, longitudinal sequences of serum uric acid measurements in 4368 individuals we produced continuous phenotypic features that suggest multiple population subtypes, and that accurately distinguished (0.97 AUC) the uric-acid signatures of gout vs. acute leukemia despite not being optimized for the task. The unsupervised features were as accurate as gold-standard features engineered by an expert with complete knowledge of the domain, the classification task, and the class labels. Our findings demonstrate the potential for achieving computational phenotype discovery at population scale. We expect such data-driven phenotypes to expose unknown disease variants and subtypes and to provide rich targets for genetic association studies.

Clinical Analysis and Interpretation of Cancer Genome Data

The scale of tumor genomic profiling is rapidly outpacing human cognitive capacity to make clinical decisions without the aid of tools. New frameworks are needed to help researchers and clinicians process the information emerging from the explosive growth in both the number of tumor genetic variants routinely tested and the respective knowledge to interpret their clinical significance. We review the current state, limitations, and future trends in methods to support the clinical analysis and interpretation of cancer genomes. This includes the processes of genome-scale variant identification, including tools for sequence alignment, tumor-germline comparison, and molecular annotation of variants. The process of clinical interpretation of tumor variants includes classification of the effect of the variant, reporting the results to clinicians, and enabling the clinician to make a clinical decision based on the genomic information integrated with other clinical features. We describe existing knowledge bases, databases, algorithms, and tools for identification and visualization of tumor variants and their actionable subsets. With the decreasing cost of tumor gene mutation testing and the increasing number of actionable therapeutics, we expect the methods for analysis and interpretation of cancer genomes to continue to evolve to meet the needs of patient-centered clinical decision making. The science of computational cancer medicine is still in its infancy; however, there is a clear need to continue the development of knowledge bases, best practices, tools, and validation experiments for successful clinical implementation in oncology.

The IGNITE network: a model for genomic medicine implementation and research

BackgroundPatients, clinicians, researchers and payers are seeking to understand the value of using genomic information (as reflected by genotyping, sequencing, family history or other data) to inform clinical decision-making. However, challenges exist to widespread clinical implementation of genomic medicine, a prerequisite for developing evidence of its real-world utility.MethodsTo address these challenges, the National Institutes of Health-funded IGNITE (Implementing GeNomics In pracTicE; www.ignite-genomics.org) Network, comprised of six projects and a coordinating center, was established in 2013 to support the development, investigation and dissemination of genomic medicine practice models that seamlessly integrate genomic data into the electronic health record and that deploy tools for point of care decision making. IGNITE site projects are aligned in their purpose of testing these models, but individual projects vary in scope and design, including exploring genetic markers for disease risk prediction and prevention, developing tools for using family history data, incorporating pharmacogenomic data into clinical care, refining disease diagnosis using sequence-based mutation discovery, and creating novel educational approaches.ResultsThis paper describes the IGNITE Network and member projects, including network structure, collaborative initiatives, clinical decision support strategies, methods for return of genomic test results, and educational initiatives for patients and providers. Clinical and outcomes data from individual sites and network-wide projects are anticipated to begin being published over the next few years.ConclusionsThe IGNITE Network is an innovative series of projects and pilot demonstrations aiming to enhance translation of validated actionable genomic information into clinical settings and develop and use measures of outcome in response to genome-based clinical interventions using a pragmatic framework to provide early data and proofs of concept on the utility of these interventions. Through these efforts and collaboration with other stakeholders, IGNITE is poised to have a significant impact on the acceleration of genomic information into medical practice.

Validating drug repurposing signals using electronic health records: a case study of metformin associated with reduced cancer mortality

Objectives Drug repurposing, which finds new indications for existing drugs, has received great attention recently. The goal of our work is to assess the feasibility of using electronic health records (EHRs) and automated informatics methods to efficiently validate a recent drug repurposing association of metformin with reduced cancer mortality. Methods By linking two large EHRs from Vanderbilt University Medical Center and Mayo Clinic to their tumor registries, we constructed a cohort including 32 415 adults with a cancer diagnosis at Vanderbilt and 79 258 cancer patients at Mayo from 1995 to 2010. Using automated informatics methods, we further identified type 2 diabetes patients within the cancer cohort and determined their drug exposure information, as well as other covariates such as smoking status. We then estimated HRs for all-cause mortality and their associated 95% CIs using stratified Cox proportional hazard models. HRs were estimated according to metformin exposure, adjusted for age at diagnosis, sex, race, body mass index, tobacco use, insulin use, cancer type, and non-cancer Charlson comorbidity index. Results Among all Vanderbilt cancer patients, metformin was associated with a 22% decrease in overall mortality compared to other oral hypoglycemic medications (HR 0.78; 95% CI 0.69 to 0.88) and with a 39% decrease compared to type 2 diabetes patients on insulin only (HR 0.61; 95% CI 0.50 to 0.73). Diabetic patients on metformin also had a 23% improved survival compared with non-diabetic patients (HR 0.77; 95% CI 0.71 to 0.85). These associations were replicated using the Mayo Clinic EHR data. Many site-specific cancers including breast, colorectal, lung, and prostate demonstrated reduced mortality with metformin use in at least one EHR. Conclusions EHR data suggested that the use of metformin was associated with decreased mortality after a cancer diagnosis compared with diabetic and non-diabetic cancer patients not on metformin, indicating its potential as a chemotherapeutic regimen. This study serves as a model for robust and inexpensive validation studies for drug repurposing signals using EHR data.

Translating genomic information into clinical medicine: Lung cancer as a paradigm

We are currently in an era of rapidly expanding knowledge about the genetic landscape and architectural blueprints of various cancers. These discoveries have led to a new taxonomy of malignant diseases based upon clinically relevant molecular alterations in addition to histology or tissue of origin. The new molecularly based classification holds the promise of rational rather than empiric approaches for the treatment of cancer patients. However, the accelerated pace of discovery and the expanding number of targeted anti-cancer therapies present a significant challenge for healthcare practitioners to remain informed and up-to-date on how to apply cutting-edge discoveries into daily clinical practice. In this Perspective, we use lung cancer as a paradigm to discuss challenges related to translating genomic information into the clinic, and we present one approach we took at Vanderbilt-Ingram Cancer Center to address these challenges.

Beyond Histology: Translating Tumor Genotypes into Clinically Effective Targeted Therapies

Increased understanding of intertumoral heterogeneity at the genomic level has led to significant advancements in the treatment of solid tumors. Functional genomic alterations conferring sensitivity to targeted therapies can take many forms, and appropriate methods and tools are needed to detect these alterations. This review provides an update on genetic variability among solid tumors of similar histologic classification, using non–small cell lung cancer and melanoma as examples. We also discuss relevant technological platforms for discovery and diagnosis of clinically actionable variants and highlight the implications of specific genomic alterations for response to targeted therapy. Clin Cancer Res; 20(9); 2264–75. ©2014 AACR.

Current and Future Trends in Magnetic Resonance Imaging Assessments of the Response of Breast Tumors to Neoadjuvant Chemotherapy

The current state-of-the-art assessment of treatment response in breast cancer is based on the response evaluation criteria in solid tumors (RECIST). RECIST reports on changes in gross morphology and divides response into one of four categories. In this paper we highlight how dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and diffusion-weighted MRI (DW-MRI) may be able to offer earlier, and more precise, information on treatment response in the neoadjuvant setting than RECIST. We then describe how longitudinal registration of breast images and the incorporation of intelligent bioinformatics approaches with imaging data have the potential to increase the sensitivity of assessing treatment response. We conclude with a discussion of the potential benefits of breast MRI at the higher field strength of 3T. For each of these areas, we provide a review, illustrative examples from clinical trials, and offer insights into future research directions.

Current and future trends in imaging informatics for oncology.

Clinical imaging plays an essential role in cancer care and research for diagnosis, prognosis, and treatment response assessment. Major advances in imaging informatics to support medical imaging have been made during the last several decades. More recent informatics advances focus on the special needs of oncologic imaging, yet gaps still remain. We review the current state, limitations, and future trends in imaging informatics for oncology care including clinical and clinical research systems. We review information systems to support cancer clinical workflows including oncologist ordering of radiology studies, radiologist review and reporting of image findings, and oncologist review and integration of imaging information for clinical decision making. We discuss informatics approaches to oncologic imaging including, but not limited to, controlled terminologies, image annotation, and image-processing algorithms. With the ongoing development of novel imaging modalities and imaging biomarkers, we expect these systems will continue to evolve and mature.

Optimizing Personalized Bone Marrow Testing Using an Evidence-Based, Interdisciplinary Team Approach

OBJECTIVES To address the overuse of testing that complicates patient care, diminishes quality, and increases costs by implementing the diagnostic management team, a multidisciplinary system for the development and deployment of diagnostic testing guidelines for hematologic malignancies. METHODS The team created evidence-based standard ordering protocols (SOPs) for cytogenetic and molecular testing that were applied by pathologists to bone marrow biopsy specimens on adult patients. Testing on 780 biopsy specimens performed during the six months before SOP implementation was compared with 1,806 biopsy specimens performed during the subsequent 12 months. RESULTS After implementation, there were significant decreases in tests discordant with SOPs, omitted tests, and the estimated cost of testing to payers. The fraction of positive tests increased. Clinicians reported acceptance of the new procedures and perceived time savings. CONCLUSIONS This process is a model for optimizing complex and personalized diagnostic testing.

Integrated Information Systems for Electronic Chemotherapy Medication Administration

INTRODUCTION Chemotherapy administration is a highly complex and distributed task in both the inpatient and outpatient infusion center settings. The American Society of Clinical Oncology and the Oncology Nursing Society (ASCO/ONS) have developed standards that specify procedures and documentation requirements for safe chemotherapy administration. Yet paper-based approaches to medication administration have several disadvantages and do not provide any decision support for patient safety checks. Electronic medication administration that includes bar coding technology may provide additional safety checks, enable consistent documentation structure, and have additional downstream benefits. METHODS We describe the specialized configuration of clinical informatics systems for electronic chemotherapy medication administration. The system integrates the patient registration system, the inpatient order entry system, the pharmacy information system, the nursing documentation system, and the electronic health record. RESULTS We describe the process of deploying this infrastructure in the adult and pediatric inpatient oncology, hematology, and bone marrow transplant wards at Vanderbilt University Medical Center. We have successfully adapted the system for the oncology-specific documentation requirements detailed in the ASCO/ONS guidelines for chemotherapy administration. However, several limitations remain with regard to recording the day of treatment and dose number. CONCLUSION Overall, the configured systems facilitate compliance with the ASCO/ONS guidelines and improve the consistency of documentation and multidisciplinary team communication. Our success has prompted us to deploy this infrastructure in our outpatient chemotherapy infusion centers, a process that is currently underway and that will require a few unique considerations.