**1. Introduction**

Advances in next-generation sequencing (NGS) technology and the resulting decrease of costs hold large promises for personalized medicine, currently revolutionizing cancer diagnostics in particular. The sequencing of whole tumor exomes, genomes and transcriptomes of patients allows physicians to make individual molecular-guided decisions. However, the complex nature of cancer and its large interindividual heterogeneity require an interdisciplinary board composed of medical and scientific experts to review and interpret the equally complex analysis results. Recent data sugges<sup>t</sup> that such molecular tumor boards (MTBs) have the potential to improve therapy and care for patients that have run out of guideline-based treatment options or have rare tumors [1,2].

Several German medical centers have already started to implement MTBs in their clinical environment, all working with various types of omics data from, e.g., NGS and other technologies [3,4]. To handle the many results from a large amount of omics data there is a high need for a standardized toolset that supports clinicians in analyzing and interpreting these data and creating high-quality presentations of complex multi-dimensional data effectively. Moreover, it requires both the integration of clinical with molecular and genomic data and the visualization of joint analysis results. However, experiences with the implementation and establishment of information technology (IT) and bioinformatics support for MTBs are still rare in Germany and probably world-wide, thus in need of improvement and optimization [5].

To address this issue, the MIRACUM consortium (Medical Informatics in Research and Care in University Medicine) has established the Use Case 3 which focuses on the provision of IT and bioinformatics support for translation and visualization of data analyzed in MTBs. As part of this use case, we will establish a generic, open-source framework that supports the analysis, interpretation and visualization of both clinical and omics data [6]. Data analysis is handled through MIRACUM-Pipe [7] which provides automatic, parameter-controlled processing of omics data with alignment, variant calling, annotation and analysis. The second aspect of the framework will be the data visualization and documentation of the results of the MTB. Both, the pipeline and the visualization, will be provided as separate and open-source components developed in a user-centered design process.

The cBio Cancer Genomics Portal (cBioPortal) was selected as a suitable platform to visualize the generated data supplied by the MIRACUM-Pipe [7] and to support the decision-making processes in an MTB. cBioPortal provides an extensive set of tools for exploring, visualizing and analyzing multi-dimensional and large-scale cancer genomics data sets [8,9]. Within the context of an MTB, cBioPortal may support case preparation, case review, and the documentation and communication of treatment recommendations in the near future. Therefore, it is well suited to replace the current practice at some German university hospitals of managing complex mutation data in huge spreadsheets by providing comprehensible visualizations [10,11]. At the participating clinics, cBioPortal could optimize the processing of up to 200 cases per year with sometimes hundreds of identified but not necessarily relevant mutations, and thus improve the decision making [5].

The integration of cBioPortal into the workflow of MTBs requires adjustments regarding di fferent functionalities and needs (requirements). For instance, to meet some of our requirements the user needs to have write access to the data stored in cBioPortal. Therefore, we must find a proper solution to accomplish this in line with the concepts cBioPortal currently pursues as a (read-only) data warehouse.

The objective of this work is to provide a requirements specification for an IT platform based on cBioPortal that supports processes in molecular tumor boards to find and document a therapy recommendation. To our knowledge, there is so far no systematic assessment of such requirements from the point of view of MTB participants in di fferent hospitals, even though there are already existing tools that integrate numerous data sources and even support the documentation process in a uniform MTB tool [12,13]. Our work could serve as a blueprint for the development of further tools based on cBioPortal for MTBs in Germany and worldwide.

#### **2. Materials and Methods**

To identify the requirements, we conducted a qualitative research to assess a set of potential requirements in two consecutive rounds of interviews. The second round of interviews became necessary because the first round was developed iteratively and therefore not all participants had the chance to comment on all mentioned potential requirements.

In preparation, we reviewed the literature published between 1997 and 2017 (scoped review) for existing systems, tools and knowledgebases that support molecular tumor boards (see Figure 1), resulting in an interview guideline for the first round of interviews.

**Figure 1.** Outline of the process of requirements analysis.

Based on almost all assessments from the first round of interviews, for the second round, we created screenshot mockups for better understanding and visualization of possible implementations in cBioPortal.

#### *2.1. Details about Scoping Review*

We conducted a review of literature for potential MTB tools following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [14,15] as far as appropriate for the requirements analysis. Therefore, we searched MEDLINE and Web of Science (all databases) for articles focusing on MTBs or equivalent clinical decision-making structures published between 1997 and 2017. We captured several features of different potential MTB tools which are either described in the literature or used by physicians at our MIRACUM sites. We used these findings to prepare for the first round of interviews. A detailed description of the methods we used can be found in Supplementary File 1.

#### *2.2. First and Second Round of Interviews*

Based on this prior knowledge, we conducted one group interview per partner site and per round from a constructivist point of view. We hypothesized that each site would have different views and visions on requirements for supporting a local MTB. Therefore, we took all suggestions regarding the demands of the participants into account.

The interviews were conducted in local focus groups, in which the interviewers served as moderators. This approach allowed discussions between the participants and thus as many requirements and their potential variants as possible could be identified. At each site, all participants were interviewed together in one session. The interviewers had an interdisciplinary background: medicine (Philipp Buechner, Melanie Boerries), medical informatics (Jan Christoph, Marc Hinderer), bioinformatics (Jan Christoph, Melanie Boerries) and biology (Melanie Boerries).

All interviews, except the one at the first author's local university hospital, took place as web conferences with transmission of voices and screen contents. In addition to the option of easily recording the session, the major argumen<sup>t</sup> for this setting was that the participants at the various university hospitals were spread all over Germany.

All interview participants were members of the MIRACUM Use Case 3 and thus known to us. The members responsible for the use case at each site arranged an appointment and invited additional local experts, who all were also members of the use case. All participants had to have experiences with the processes related to an MTB in order to join the interviews.

#### 2.2.1. Structure and Purpose of the First Round of Interviews

Philipp Buechner, Marc Hinderer and Jan Christoph conducted the interviews of the first round together as members of MIRACUM's Use Case 3 between June 2018 and August 2018. It comprised a short guideline with questions (see Supplementary File 2) we developed from the results of the scoped review. We also demonstrated the main functionalities (see Supplementary Files 3 and 4) of the following potential MTB tools:


We collected all mentioned requirements cBioPortal must meet (including details about potential options for implementation) that were mentioned during the meetings. The interview process was developed iteratively and the information gained was immediately incorporated into the subsequent interviews with other partner sites during this first round. We used the web conference system "Adobe Connect" to conduct, record and subsequently analyze these interview sessions.

#### 2.2.2. Structure and Purpose of the Second Round of Interviews

The second round of interviews was performed by Philipp Buechner, Melanie Boerries and Jan Christoph between November 2018 and December 2018. In order to make optimal use of the limited time during the interviews, Philipp Buechner developed a comprehensive interview guideline describing the requirements and their potential options identified in the first round with text and screenshot mockups (see Supplementary File 5). However, this round did not cover the requirements mentioned in the first round of interviews, that are already implemented in cBioPortal or are generally out of the scope of MIRACUM Use Case 3. To familiarize the participants with the requirements, this guideline—once it was finally validated by Melanie Boerries, Jan Christoph and Philipp Unberath —was handed out to them prior the meetings.

Since some requirements had more than one potential option regarding implementation and visualization, sites were asked to select one during this round of interviews. In case they had different opinions, they were encouraged to find a compromise.

For the final software specification—after all interviews have been conducted and analyzed - we grouped related features into larger meta-categories to account for individual requirements and ye<sup>t</sup> to keep the assessment structured. For example, the term "sample metadata", comprises six (individual) data features:


When calculating the total number of requirements identified by us, we only counted those combined meta categories. Therefore, the above-mentioned example of "sample metadata" counts as one requirement instead of six individual ones.

We used the web conference system "Zoom" to conduct, record and subsequently analyze these interview sessions.

#### *2.3. Low-Fidelity Mockup Demonstrator*

We created 54 descriptive screenshot mockups for almost all options of the identified requirements from the first round of interviews using the image-editing tool GNU Image Manipulation Program (GIMP), version 2.10.8. These low-fidelity mockups are based on full-screen screenshots of the cBioPortal graphical user interface and have been manipulated to give the realistic appearance of providing the respective functions. To quickly direct the viewer's focus to the part of the image that represents the demanded function we indirectly highlighted the area by darkening the rest of the image with a black overlay (opacity: 20%).

#### *2.4. Consultation with Main Developers of MSKCC*

After all interviews have been conducted, we discussed the requirements with the main developers of cBioPortal from the Memorial Sloan Kettering Cancer Center (MSKCC) in New York, USA, in an online audio conference. Prior to that, we had detailed the most important and far-reaching changes in a letter including excerpts from our mockups.

The aim of this was to increase the chances of merging our planned implementations into the main development branch of cBioPortal and to maintain contact with the main developers right from the beginning.

## *2.5. Ethical Approval*

This study was ethically approved by the ethics committee of the Friedrich-Alexander-University Erlangen-Nürnberg (FAU) (see Supplementary File 6).
