SMART Cancer Navigator: A Framework for Implementing ASCO Workshop Recommendations to Enable Precision Cancer Medicine

Abstract

Purpose

Data standards and interoperability are critical for improving care for patients with cancer. Recent efforts by ASCO include the Data Standards and Interoperability Summit in 2016, which led to the Omics and Precision Oncology and Advancing Interoperability workshops. To facilitate improved patient care, several recommendations for data sharing and standardization were made to the community.

Methods

To address these recommendations, we developed SMART Cancer Navigator, a Web application that uses application programming interfaces to gather clinical and genomic data from 11 public knowledge bases ranging from basic to clinical content coverage; three (CIViC, ClinVar, and OncoKB) explicitly linked genomic variants to clinical factors such as prognosis and treatment selection. We illustrated the utility of this application by selecting one of the monthly case studies presented by the ASCO University Molecular Oncology Tumor Board: Ovarian Cancer (BRCA Mutation). We also performed analyses on information from the three clinico-genomic knowledge bases to corroborate previous work and illustrate the state of data sharing among publicly available resources.

Results

SMART Cancer Navigator aggregates and contextualizes data from 11 different knowledge bases and stores user queries in a lightweight Web application that can link into Fast Healthcare Interoperability Resources–enabled electronic health records. Potentially relevant clinical trials and/or approved treatments were identified for three mutations found in a hypothetical patient with advanced ovarian cancer. A comparison of the three clinico-genomic knowledge bases indicated substantial differences in coverage at the gene and variant levels.

Conclusion

SMART Cancer Navigator has immediate relevance to practicing oncologists and others. Additional knowledge bases can be added without undue effort. As a first step toward utility, we generalized and disseminated the resulting implementation (https://smart-cancer-navigator.github.io) and data sets.

INTRODUCTION

Precision medicine is a model for disease prevention and treatment that considers variability in genes, environment, and lifestyle across individuals.¹ To expand and broaden this model, the federal government invested $215 million in establishing the Precision Medicine Initiative in 2015, with a large portion of the funding devoted to precision oncology.^2,3 Advances in molecular profiling to identify actionable genetic aberrations provided a compelling rationale for this initial emphasis.^4-6 In May 2016, ASCO convened a Data Standards and Interoperability Summit to address the critical data sharing and standardization issues that could impede the realization of precision oncology.

In typical genomic workflows, oncologists request genomic analysis of a recurrent or metastatic tumor, and genomics laboratories return variably structured narrative reports of the genes and variants they identified, which are often scanned into the patient’s electronic health record (EHR). Some reports are more than 20 pages long, including characterization of genetic variants, potential targeted therapies, and relevant clinical trial information, whereas others may be brief and may lack any interpretation. Facing a lack of clear actionability (ie, translation of genomic findings into treatment recommendations), potential bias in report curation, and potentially outdated information, oncologists often have to access additional knowledge bases to obtain more comprehensive, up-to-date disease-gene-variant information. Inconsistencies among knowledge bases (eg, differences in querying syntax and graphical user interfaces) lead to inconvenience and inefficiency; furthermore, oncologists must re-enter patient data every time they need to update their queries.

Because of these issues, attendees at the 2016 ASCO Omics and Precision Oncology (OPO) workshop called for the development of new software applications to help community oncologists interpret the ever-increasing volumes of genomic data critical for selecting treatment.⁷ Here we describe SMART Cancer Navigator, a software application (app) we developed to facilitate knowledge base queries and identify actionable information for practicing oncologists. The purpose of this article is to describe the development of and demonstrate the utility of the SMART Cancer Navigator and to disseminate it to the oncology community for implementation and feedback.

METHODS

App Overview

SMART Cancer Navigator securely links patient-specific data from EHR and genomics laboratories to multiple independent knowledge bases to facilitate efficient and effective querying of disease-gene-variant information. The Web-based app was built by a team of software developers at the Biomedical Cybernetics Laboratory at the Boston Children’s Hospital Computational Health Informatics Program and Harvard Medical School who used an angular and bootstrap front-end framework. The app, prepopulated with relevant information from genomic laboratories and EHRs, provides an interface that oncologists can use to query evidence from several cancer-focused, genomic, clinico-genomic, and clinical trials knowledge bases. Through built-in feedback functionality, meaningful responses regarding the usability and efficacy of the app can be collected for continuing improvement.

SMART Cancer Navigator is a modular app built by using Substitutable Medical Apps Reusable Technology (SMART) on Fast Healthcare Interoperability Resources (FHIR) technology. The SMART on FHIR platform was chosen to ensure interoperability with as many EHR systems as possible.^8-10 SMART on FHIR facilitates the development of apps that can connect to health data systems (eg, EHRs, personally controlled health records, health information exchanges) without any custom configuration.^8,11,12 Through the development of application programming interfaces (APIs [apps]) developed on the SMART platform are substitutable: just as applications on a mobile technology platform such as Apple’s operating system (iOS) are separated from the core system and services (eg, camera, geolocation), apps on the SMART platform are separated from health data systems (the equivalent of the core system and services on iOS).

Login and Authentication

SMART Cancer Navigator can be linked to a patient’s API-enabled EHR. Because SMART on FHIR-enabled EHRs provides an OAuth2-based approach for secure login and authentication, the user can login and authorize the app for data access once the user has been authenticated with the EHR (ie, has provided a username and password linked to appropriate credentials).⁹ Once successfully authenticated and authorized, the app can be populated with relevant patient and practitioner information stored in the local environment.

Nomenclature, Drop-Down Menus, Data Entry, and Queries

After logging in and optionally linking to a patient’s EHR, the user is presented with an interface for querying gene-variant combinations (Fig 1). SMART Cancer Navigator uses a hybrid data entry field that allows filtering via text entry followed by selection from a drop-down menu, which ensures efficient and consistent querying of gene-variant combinations. When performing a query, the user inputs data specific to a variant or gene. As the user types in the search field, the complementary drop-down list is populated in real time with suggested gene-variant combinations from the linked knowledge bases (Fig 1). Using this real-time service guarantees access to the latest available gene-variant information.

Fig 1.

Query view with three Variants from ASCO Molecular Tumor Board Case Study. Name and age are randomly generated; any resemblance to an actual person is purely coincidental.

The user begins a query by typing in the search bar, and the app’s search algorithm efficiently sifts through thousands of known gene-variant combinations. Users can input variants and genes in their natural language (eg, “STK1 D835Y” is translated to “FLT3:c.2503G>T (p.Asp835Tyr)”). Users then select their choice from the drop-down menu. This approach enables users to select gene-variant combinations without having to adhere to formal nomenclature or having to know the latest gene names.

For executing queries, the app uses the following standards: genes are represented using HUGO Gene Nomenclature Committee–approved symbols and National Center for Biotechnology Information Entrez gene IDs, and variants are represented using the Human Genome Variation Society (HGVS) protein and genomic reference sequences.^13-16

Once the user has selected the appropriate gene-variant combination, SMART Cancer Navigator translates the combination into relevant API calls and populates an expandable view containing the results. These searches and results can then be saved to the patient’s EHR.

Knowledge Bases and APIs

SMART Cancer Navigator accesses 11 knowledge bases: (1) Clinical Interpretations of Variants in Cancer (CIViC), (2) ClinVar, (3) Catalogue of Somatic Mutations in Cancer (COSMIC), (4) Combined Annotation Dependent Depletion (CADD), (5) Database for Nonsynonymous SNPs’ Functional Predictions (dbNSFP), (6) Database for Single Nucleotide Polymorphisms (dbSNP), (7) Cancer Genome Interpreter (CGI), (8) Precision Medicine Knowledgebase (PMKB), (9) OncoKB, (10) Drug-Gene Interaction Database (DGIdb), and (11) ClinicalTrials.gov.^17-27 These knowledge bases were selected because at least some or all of their content was freely available through APIs in mid to late 2017.

Variant annotation information from CIViC is queried via standardized APIs that use the HGVS genomic reference sequence ID through the MyVariant.info RESTful service²⁸ and via a proprietary API format. Knowledge bases 2 through 7 are queried with MyVariant.info, and knowledge bases 8 through 11 are queried using their own API formats. If the knowledge bases fail to return data for a query, the app records a knowledge base gap.

Displaying Results

Upon selecting a gene-variant combination, the app queries the knowledge bases and then returns results in an expandable view. The knowledge bases are used to query for the results outlined in Tables 1 and 2. Within three expandable subsections (Gene, Variant, and Clinical Trials), users can interact with information and view context-specific links. For example, when interacting with the functional prediction (clinical significance) field in the Variant subsection, users can view evidence from ClinVar and relevant accession numbers.

Table 1.

Information Presented in SMART Cancer Navigator.

Table 2.

Information Available From Individual Knowledge Bases

Because SMART Cancer Navigator compiles information from many different data sources, there may be instances in which sources present conflicting information. To mitigate this potential issue, the application presents interpretations in line (with their respective source of origin) and prepends each additional interpretation with a slash. In fields with more than two values, the application presents a link with the message “conflicting information.” Upon clicking the link, the user can view all interpretations and respective data sources in a tabular format.

Interface With EHRs and Local Servers

SMART Cancer Navigator follows SMART on FHIR specifications to connect to EHRs and interpret patient context.⁹ Optionally, users can also manually enter patient data; by default, only non-protected health information (PHI) is exposed to the external APIs. Also by default, gene-variant information is stored in the local environment. Thus, when a user adds or removes variants, a local FHIR server is automatically updated to reflect these changes. The app also provides a toggle to disable this feature and instead have the user manually save when necessary (eg, for cases in which the user does not wish to have the changes made immediately to the patient’s EHR).

SMART Cancer Navigator saves gene-variant combinations as FHIR “Observation” resources, which are stored on a local FHIR server. In accordance with SMART on FHIR specifications, the “Observation” resource is separate from the “Patient” resource, but includes a link to the “Patient Resource ID” to ensure that this information remains linked. With this preserved link, gene-variant combinations can be stored locally and used to repopulate the app with updated patient information. During development, the following test systems were used: the Healthcare Services Platform Consortium Sandbox²⁹ to simulate EHRs and the FHIR Genomics Server³⁰ server to simulate a Genomics Archive Communications System.

Knowledge Base Analysis

In addition to developing SMART Cancer Navigator, we performed analyses to examine concordance among publicly available knowledge bases. To verify and corroborate prior work, we compared publicly available data from CIViC, ClinVar, and OncoKB as of August 5, 2017.^19,23,24 First, all data from each knowledge base were examined individually; then, each downloaded database was filtered to include only data from Genome Reference Consortium Human Genome Build 37.³¹ We used Venn diagrams to illustrate the current state of genes present in more than one knowledge base.

To verify and expand upon prior work,³² we sought to determine which variant origin classification was most common within each database. We used χ² to compare the distribution of origin classifications to a uniform distribution. For the two databases that provided clinical significance classifications, we sought to determine the most common classification within each database, again using χ². All performed tests were two-sided, and P < .05 was considered to indicate statistical significance.

ASCO OVARIAN CANCER CASE STUDY

To illustrate the utility of SMART Cancer Navigator, we selected one of the monthly case studies that was presented by the ASCO University Molecular Oncology Tumor Board: Ovarian Cancer (BRCA Mutation).³³ In July 2017, a hypothetical case of high-grade serous ovarian carcinoma (HGSOC) was presented. Briefly, a 62-year-old woman underwent optimal debulking for advanced-stage HGSOC and remained in complete remission for 3 years after first-line platinum doublet therapy. She then relapsed with platinum-sensitive disease and attained a second complete remission (see Data Supplement for full case description; materials used with permission).

Under current National Comprehensive Cancer Network Clinical Practice Guidelines,^34,35 this patient with HGSOC should have received genetic counseling, regardless of family history. Because testing at diagnosis frequently does not occur,^36,37 we present the following hypothetical results from next-generation sequencing panel testing conducted at the time of relapse: (1) BRCA1 p.Trp1815Ter at 50% allele frequency, consistent with a germ line variant; (2) BRCA1 p.Cys61Gly at 10% allele frequency, consistent with a somatic variant; and (3) an ETV6-NTRK3 fusion (Data Supplement).

In the de facto workflow, medical oncologists and tumor board coordinators face significant obstacles when they attempt to corroborate and enrich these genetic testing reports that use multiple knowledge bases.³⁸ For each search, data must be manually entered; in addition, nonstandardized nomenclature between reports and knowledge bases can lead to inconvenience and the omission of crucial information. Furthermore, search results are not captured in a format amenable to EHR integration, which may diminish rapid translation.^7,39 The SMART Cancer Navigator app removes these obstacles.

To begin, the oncologist logs in, chooses the appropriate patient from the EHRs, and launches SMART Cancer Navigator, with prepopulated patient data, including name, sex, age, and any previously queried gene-variant combinations. Next, the user is presented with the query view (Fig 1). A separate query would be made for each gene-variant combination; SMART Cancer Navigator automatically saves each query, allowing for subsequent re-querying of knowledge bases at a later time without the need for redundant data entry.

Immediately after entering each query, the user can interact with the results for each gene-variant combination. In Gene results (Fig 2), the user is presented with relevant summaries and descriptions as well as important details (gene symbols and IDs, gene type, aliases, genomic position, and pathways). The Variant results provide relevant descriptions, evidence items, and details, such as variant type, functional prediction, coordinates, tumor site, knowledge base IDs, mutation frequency, and suggested treatments (Data Supplement). Finally, the ClinicalTrials.gov results present entries relevant to the respective gene-variant combination by searching through the National Institutes of Health’s database via the appropriate HGVS protein and genomic reference sequences (Data Supplement).

Fig 2.

Results view for Gene search. A prose description of the gene and its function is provided on the left; structured information from the knowledge bases is presented in tabular format on the right.

With regard to the example case, there is evidence for potentially relevant effective drugs within the Variant results section. Listed under the BRCA1 p.Trp1815Ter variant is olaparib, a US Food and Drug Administration–approved⁴⁰ maintenance therapy. Under the BRCA1 p.C61G variant, rucaparib is listed as a potential therapy; this poly (ADP-ribose) polymerase inhibitor has led to frequent durable remissions among patients with relapsed HGSOC, regardless of whether mutations were somatic or germ line.^41-43 In the ClinicalTrials.gov results, NCT02855944 is listed as an actively recruiting (as of October 2017) clinical trial (ARIEL4: A Study of Rucaparib Versus Chemotherapy BRCA Mutant Ovarian, Fallopian Tube, or Primary Peritoneal Cancer Patients).

Similarly, under the ETV6-NTRK3 variant, larotrectinib (LOXO-101) is listed as a potential therapy.⁴⁴ The accompanying ClinicalTrials.gov results for this variant include numerous site-agnostic trials, including the NCI-MATCH basket trial (NCT02465060: NCI-MATCH: Targeted Therapy Directed by Genetic Testing in Treating Patients with Advanced Refractory Solid Tumors, Lymphomas, or Multiple Myeloma), which contains larotrectinib in arm Z1E. Thus, using SMART Cancer Navigator, oncologists could readily review all of these results and consider options to select the best care for their patients.

DATABASE ANALYSIS RESULTS

In our gene-level analysis, we found that only 2.2% (147 of 6,548) unique genes were common to all three analyzed knowledge bases (Fig 3A; Data Supplement). Comparing variants present in the two cancer-focused knowledge bases (CIViC and OncoKB), we found only 8.1% (432 of 5,317) unique variants in common (Data Supplement).

Fig 3.

Venn diagrams for a comparison of genes in (A) ClinVar, Clinical Interpretations of Variants in Cancer (CIViC), and OncoKB and (B) ClinVar, CIViC, and OncoKB that are involved in Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathways in Cancer.

Using the functional annotation tool Database for Annotation, Visualization and Integrated Discovery (DAVID⁴⁵), we performed gene enrichment analysis on data from CIViC, ClinVar, and OncoKB. The Kyoto Encyclopedia of Genes and Genomes (KEGG⁴⁶) Pathways in Cancer genes were identified in all three databases, with 4.1% of ClinVar genes, 30.8% of CIViC genes, and 34.1% of OncoKB genes associated with KEGG pathways (all P < .001 via DAVID’s EASE Score Threshold,⁴⁷ a modified Fisher’s exact P value). Notably, of all KEGG cancer pathway genes, 64 genes were present in all three knowledge bases, whereas 140 genes were not present in any of the three knowledge bases evaluated (Fig 3B; Data Supplement).

Analyzing 276,092 classified variants from 6,425 genes (approximately 94.5% of the variants in the ClinVar database) indicated that the distribution of origin types for variants differed significantly from a uniform distribution, with 92.6% classified as germ line (P < .001). Similarly, an analysis of 2,039 unique variants from 292 genes (100% of the CIViC database) with a germ line, somatic, or not applicable classification found significant variance from a uniform distribution (1,650 of the 2,039 variants [80.9%] were of somatic origin; P < .001). OncoKB was not analyzed this way, because all variants accepted to this knowledge base must be of somatic origin.²³

ClinVar classified variants as being benign, pathogenic, of uncertain significance, or with conflicting interpretations of pathogenicity.¹⁹ An analysis of 279,998 variants from 6,110 genes (approximately 92.8% of the variants in the ClinVar database), indicated that the distribution of clinical significance classifications for variants differed significantly from a uniform distribution (P < .001); 42.1% of the analyzed variants were of uncertain significance, followed by benign (30.9%), pathogenic (22.8%), and conflicting interpretations of pathogenicity (4.2%). OncoKB classified variants as inconclusive, likely neutral, likely oncogenic, or oncogenic. An analysis of 3,701 variants from 273 genes (100% of the OncoKB database) indicated that the knowledge base’s distribution of clinical significance classifications differed significantly from a uniform distribution (P < .001; 52.1% of the analyzed variants were of likely oncogenic, followed by oncogenic [28.6%], likely neutral [11.4%], and inconclusive [8.0%]).

DISCUSSION

SMART Cancer Navigator was built in response to the OPO recommendation that “a software application should be developed by or on behalf of ASCO, to help both community and academic oncologists integrate higher quality genomic data into clinical practice.”⁷ In particular, the app addresses four critical concerns identified by ASCO OPO workshop participants: a lack of structured, actionable genomic data; poorly standardized genomic nomenclature; inconsistency in interpretation between laboratories; and a lack of formal training among the majority of oncologists in interpreting genomic data.⁷ First, SMART Cancer Navigator overcomes a lack of structured data by storing patient and query data as FHIR bundles. Second, adopting several nomenclature standards with widespread use alleviates issues arising from poorly standardized genomic nomenclature. Third, inconsistency in interpretation is addressed by providing multiple interpretations from numerous knowledge bases, with the aim of presenting a more holistic approach on the basis of multiple up-to-date knowledge sources. Fourth, with simple drop-down interfaces in the Search/Query view and well-organized structured outputs in the Results view, oncologists can readily access disease-gene-variant information without any formal training. In addition, SMART Cancer Navigator links genomic information to relevant clinical trials, provided that they have the relevant HGVS protein or genomic reference sequences, thereby addressing another key OPO workshop recommendation.

Analysis of existing knowledge bases revealed fairly disparate coverage at the gene and variant levels. When reviewing these results, it is important to first consider the following issues. Although CIViC and OncoKB are knowledge bases with a clinical focus on cancer-related variants, ClinVar contains reports of all human variants and related phenotypes. Consequently, there are far more genes and variants recorded in ClinVar than in the other two knowledge bases. Both cancer-focused knowledge bases are enriched with somatic variants, a reasonable trend, because oncology today is rooted in the fundamental understanding that cancer stems from the accumulation of somatic mutations that elicit malignant transformation.^48-51 However, the fact that even sporadic cancers can be enriched for germ line variants means that cancer-focused knowledge bases should also have substantial germ line information.^52,53 Whether differences in coverage of the cancer-focused knowledge bases reflect differing priorities versus incomplete curation is unclear; active efforts such as the Variant Interpretation for Cancer Consortium⁵⁴ are underway to coordinate efforts and maximize interoperability. It is expected that more cancer-specific knowledge base content will become available via public APIs; our analysis was limited to CIViC and OncoKB because of their API availability.

Review of the ClinVar clinical significance results, which were enriched with variants of unknown significance, makes apparent the ongoing need for the interpretation and discovery of actionable variants. After the recent advances in next-generation sequencing technologies, it is clear that there is still much work to be done in studying new variants and providing relevant information regarding actionable variants to oncologists.^55-57

In addition, differences between proprietary and standardized APIs pose an obstacle to directly incorporating data from some knowledge bases. Although our Web application provides a level of abstraction useful for clinicians, clinical researchers may have additional needs. They would need to familiarize themselves with the structure of the APIs and would find that different combinations of operations and internal terminologies or IDs are often needed to access data across knowledge bases. Because data granularity, specificity, and nomenclature may differ, and evidence reliability ratings are not readily translatable across knowledge bases, this highlights a need for better representation of genomic data and adoption of uniform standards. Furthermore, the proprietary nature of some knowledge bases poses obstacles for peer review. Indeed, there is much work to be done to facilitate data sharing between different organizations; nevertheless, our application serves to mitigate these gaps in data sharing by linking multiple knowledge bases together through one convenient solution.

In its current state, SMART Cancer Navigator is focused on data aggregation, and it presents data in a simple tabular format. More sophisticated methods of data visualization are necessary, in particular to highlight the co-occurrence of mutations, which is increasingly clinically relevant.⁵⁸ We have previously shown that standard visual tools such as the pie chart do not scale when dealing with large numbers of genes and variants.⁵⁹ To address this challenge, the BioVis community recently convened a challenge workshop and will soon be formulating recommendations.⁶⁰

Although SMART Cancer Navigator allows users to select a relevant patient condition and then select a gene-variant doublet, true clinical decision support in oncology requires a triplet of disease-gene-variant information and, in many cases, further metadata such as prior treatments received. Current knowledge base API architecture does not meet this need, because most provide only end points for the querying of genes and variants or provide end points for other relevant clinical information that may not presently return data. The clinical trials example shown in the Data Supplement illustrates this challenge, because only one of the identified trials is directly relevant to the patient with relapsed HGSOC.

It must also be acknowledged that apps such as SMART Cancer Navigator may be seen as a security risk by some organizations. Although SMART on FHIR apps have a security structure that meets or exceeds the requirements set forth by the Health Insurance Portability and Accountability Act,⁹ genetic information is nonetheless highly sensitive. Whether used for clinical or research purposes, information entered into SMART Cancer Navigator is subject to the conditions of the Genetic Information Nondiscrimination Act. The ethical, social, and legal implications of genetic information exposed through APIs, with or without accompanying PHI, will be the subject of continued intense scrutiny.

CONCLUSION AND DISSEMINATION

As a first step, we have released SMART Cancer Navigator⁶¹ as an open-source application under Apache License Version 2.0 and provided our data sets (Data Supplement) to the community to enable synergy and further development. Future goals include working with ASCO and other partners to develop a mapping service (a meta-knowledge base) that can identify various nomenclatures for any gene-variant combination.⁷ In subsequent versions, we will further prioritize and contextualize results and create reports of gene-variant queries and respective results that are exportable as documents or FHIR messages.

We have made several assumptions that are critical to the success of this and similar apps. First, we designed the app to be simple to use, but we are mindful of the fact that clinicians are under intense time pressures that are not expected to abate anytime soon. Second, we assume that the app ecosystem will evolve rapidly^11,12,62; endorsement by large EHR vendors through activities such as Epic’s App Orchard⁶³ and Cerner’s participation in the HL7 Argonaut Project is encouraging. However, for the app ecosystem to reach its full potential, apps must be able to write to EHRs or to intermediary systems that subsequently communicate with EHRs, such as Genomics Archive Computer System.^10,39,64 Finally, we assume that the regulatory environment for apps, which are considered “Software as a Medical Device” by the US Food and Drug Administration under certain circumstances,^65,66 will remain on the track that has recently been indicated; expensive regulatory burdens would likely stifle the environment of open-source apps because costs could not be recouped.

In light of recent discussions regarding data sharing,⁶⁷ we believe that our framework acts as a model for facilitating a symbiotic relationship between clinicians, clinical researchers, and bioinformatics data scientists. Although SMART Cancer Navigator was developed primarily for clinical application, it also holds value as a tool for clinical researchers and knowledge base curators. Although concerns of policy, privacy, and practicality must still be worked out to ensure that data sharing is efficient, sustainable, and secure, SMART Cancer Navigator illustrates the benefits that are conferred when parties work together to share resources and knowledge.

Finally, SMART Cancer Navigator seeks to support oncologists in the interpretation of ever-increasing volume of patient molecular data.^68-70 By linking directly to patient EHR systems, genomic laboratories, and multiple authoritative knowledge bases, SMART Cancer Navigator can serve as an educational resource for clinical oncologists, trainees, patients, and researchers. Furthermore, SMART Cancer Navigator’s use of APIs, which are mandated by the 21^st Century Cures Act,⁷¹ provides for a truly modular application that can be easily integrated into many environments. With the ability to gather user feedback, SMART Cancer Navigator can continue to be developed and improved, which will help integrate genomic data into clinical practice and accelerate precision oncology.^72-74

AUTHOR CONTRIBUTIONS

Conception and design: Jeremy L. Warner, Ishaan Prasad, Makiah Bennett, Monica Arniella, Kenneth D. Mandl, Gil Alterovitz

Financial support: Gil Alterovitz

Administrative support: Kenneth D. Mandl

Collection and assembly of data: Jeremy L. Warner, Ishaan Prasad, Makiah Bennett

Data analysis and interpretation: Jeremy L. Warner, Ishaan Prasad, Makiah Bennett, Alicia Beeghly-Fadiel, Gil Alterovitz

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Jeremy L. Warner

Stock and Other Ownership Interests: HemOnc.org

Ishaan Prasad

No relationship to disclose

Makiah Bennett

No relationship to disclose

Monica Arniella

No relationship to disclose

Alicia Beeghly-Fadiel

No relationship to disclose

Kenneth D. Mandl

Stock and Other Ownership Interests: Accountable Care Transactions, Twine Health, Medal Pharmaceuticals

Consulting or Advisory Role: Medal Pharmaceuticals

Travel, Accommodations, Expenses: Quest Diagnostics

Gil Alterovitz

No relationship to disclose