This study describes the development of the NGS-QR app, a FHIR-based web app, to report, manage, and monitor the quality information from the process of producing genomic data. As shown in Table 1, the development was composed of the following phases: (1) requirement analysis, (2) design, (3) implementation, and (4) testing. In the requirement analysis phase, we defined use cases and selected data elements such as DNA purity and integrity, library input amount and size, and sequencing running quality, in accordance with standards and guidelines. In the design phase, we profiled a FHIR resource on the existing genomic resource to exchange quality information in a standardized manner. We also designed user interfaces and functions for the NGS-QR app. In the development phase, we developed the three following components: web app (NGS-QR app), FHIR server, and R Shiny server. The web app includes a FHIR resource handler to generate and parse FHIR resources. Finally, in the testing phase, the NGS-QR app was validated using real-world data collected from the NGS pipeline, a targeted sequencing platform.

We defined two use cases: quality information reporting and quality management for clinical NGS genomic testing. The first use case is related to the quality information reporting for clinical NGS genomic testing, as shown in Figure 1a. The laboratory reports the results of NGS genomic testing to the hospital along with quality information. This use case assumes that the hospital information system provides the FHIR APIs. The laboratory requested for the NGS genomic testing sends the results using the NGS-QR app, which acts as an FHIR client. When the user fills out the reporting form in the NGS-QR app, a FHIR resource is created in JSON format and then the resource is sent to the server via the FHIR API. Hospitals that request NGS genomic testing to multiple laboratories can receive information in the same format through this standardized method.

The second use case is related to the quality management for clinical NGS genomic testing, as shown in Figure 1b. Users such as health care providers and performers of NGS testing manage and monitor the quality status of the genomic data received from the laboratories. Through the NGS-QR app, users can retrieve experimental data from the FHIR server and view the summary such as the total number of genomic tests and the number of tests based on the year or specimen type for all genomic test results. Moreover, quality information such as DNA purity, integrity, and data quality for each sample can be compared through statistical analysis. This step can be used to determine the reliability of clinical NGS genomic testing.

The NGS workflow involves complex procedures consisting of several steps, which are broadly divided into sample preparation, library preparation, sequencing, and data processing. In this process, various types of experimental conditions and results are generated and captured. As shown in Table 2, we selected data elements based on the ISO 22692 [15] and guidelines such as those of the FDA [20], American College of Medical Genetics [30], and College of American Pathologists [31] for quality management and comparative analysis. For this study, approximately 30 data elements were selected.

We profiled a FHIR genomic resource by defining constraints and extensions for exchanging quality information with a standardized method. Since the results of NGS genomic testing include genomic variant information and its quality-related data, the FHIR R4 MolecularSequence resource [32] was used for tailoring the NGS data to the use cases in this study. We created a QcMetrics element (a FHIR extension) in the resource and added data elements for each process under QcMetrics. Textbox 1 shows an example of a FHIR genomics resource that includes the QcMetrics profile. We uploaded the profile to the public FHIR registry SIMPLIFHIR.NET [33].

We developed the following three major components: the NGS-QR app (FHIR client), FHIR server, and R Shiny server, as shown in Figure 2. The NGS-QR app was developed using Node.js (v12.13.1), which is composed of user interfaces (UIs), a FHIR resource handler, and a REST API module. The app has been deployed to the App Gallery of SMART on FHIR, which is an open platform for substitutable third-party health apps to connect to electronic medical record (EMR)/electronic health record (EHR) systems with appropriate security guarantees [34,35]. The source code of the app is available at GitHub [36]. The FHIR server was locally installed using the HAPI FHIR server (v4.0.0), which is an open-source Java implementation of the FHIR specification. The QcMetrics profile was added to the FHIR server. The source code for operating the R Shiny server was written in RStudio Cloud (R version 3.6.3) and the server was deployed to the R cloud platform, Shinyapp.io [37]. The R Shiny server fetched the data used for statistical analysis from the FHIR server. The R source code is available at GitHub [38].

FHIR has a set of security recommendations that identifies communications security, authentication, authorization, access control, and auditing [39]. We applied Transport Layer Security (TLS) and OAuth 2.0, which are industry-standard protocols for communications security and authentication [40,41]. The TLS was used for the encryption of FHIR resources transmitted between the NGS-QR app and the FHIR server. The OAuth 2.0 protocol was used to grant the NGS-QR app access to the FHIR server. The NGS-QR app requests an access token by authenticating with the Authorization Server. The Authorization Server authenticates the NGS-QR app and issues an access token. The access token was used for security credentials for the NGS-QR app to make API requests on behalf of a user.

Samsung Medical Center (SMC) has developed and utilized a cancer panel sequencing pipeline, namely CancerSCAN, to determine effective treatment methods for patients based on data from more than 15,000 panel sequencing studies since 2014 [17,42,43]. The SMC received clinical laboratory accreditation by the MFDS in 2017, and CancerSCAN is clinically used for the diagnosis and prognosis of cancer patients [17]. For this study, approximately 1000 data entries were collected from CancerSCAN. They contained the experimental conditions and results for FFPE specimens, fresh cells, and cell lines.

In the reporting UI, input data are converted to a FHIR genomic resource in JSON format using the FHIR resource generator, and the resource is sent to the FHIR server using the POST method, as shown in Figure 3. In the dashboard UI, genomic resources in the FHIR server are compiled using the GET method to display the summary of data such as the total number of genomic tests and the number of tests based on the year and specimen type for the complete genomic test results. In the statistics UI, the user can select the group of samples, and view the distribution and threshold of each experimental parameter. The user can also check the current quality status of each sample in the overall distribution, as shown in Figure 4.

We validated the utility of the NGS-QR app using real-world data from the NGS pipeline CancerSCAN in SMC. All data were converted to FHIR genomic resources and sent to the FHIR server using the NGS-QR app. Figure 5 shows the statistics of the experimental data used in this study. As expected, fresh cell samples had better quality than FFPE samples in almost every category.

NGS technology has been widely adopted in clinical practice, which will play a prominent role in precision medicine. Although many countries and international institutions have developed NGS guidelines to implement clinical NGS applications, the quality criteria for determining the reliability of NGS genomic testing have not yet been standardized due to the complexity of the NGS technology. Since the scope and purpose of NGS applications are very diverse, it remains challenging to evaluate the validity and utility of clinical NGS genomic testing. At present, most individual laboratories have their own processes and criteria; thus, it is important to share quality information from the process of generating genomic data.

In this study, we propose the NGS-QR app to exchange quality information for clinical NGS genomic testing. This study provides the following main contributions to the field. First, we demonstrated that the quality information managed only by individual laboratories could be shared using a standardized format. In the NGS workflow, the experimental conditions change depending on the purpose of the tests and the status of the samples. Thus far, it has been difficult to determine the conditions under which the genome data were generated. This study defined quality-related elements and profiled the FHIR genomic resource to report the experimental conditions and results produced in NGS genomic testing. Since our proposed method is based on ISO/TS 22692 and HL7 FHIR standards to interoperably share the quality-related data of clinical genomic testing, the app can communicate with any EMR/EHR systems that conform to these standards.

Second, this study facilitated the verification of the quality status of each sample as experimental data accumulate. The quality of genome data is determined through comparative analysis based on the characteristics of NGS technology. As shown in Figure 6, the results from different types of specimens cannot be directly compared because this may lead to incorrect conclusions. Therefore, it is important to select the target groups and determine the performance of individual samples within them. The NGS-QR app allows users to select various experimental conditions and compare the produced data with the correct target group to check the exact data quality.

This study successfully demonstrated how the quality information of clinical NGS genomic testing can be exchanged using a standardized method. As the demand for NGS genomic testing increases and genomic data accumulate, quality information can be used as reference material for improving the quality of testing. This approach can also motivate laboratories to perform diagnostic tests to provide high-quality genomic data.