Empowering personalized oncology: evolution of digital support and visualization tools for molecular tumor boards

Abstract

Background: Molecular tumor boards (MTBs) play a pivotal role in personalized oncology, leveraging complex data sets to tailor therapy for cancer patients. The integration of digital support and visualization tools is essential in this rapidly evolving field facing fast-growing data and changing clinical processes. This study addresses the gap in understanding the evolution of software and visualization needs within MTBs and evaluates the current state of digital support. Alignment between user requirements and software development is crucial to avoid waste of resources and maintain trust.

Methods: In two consecutive nationwide medical informatics projects in Germany, surveys and expert interviews were conducted as stage 1 (n = 14), stage 2 (n = 30), and stage 3 (n = 9). Surveys, via the SoSci Survey tool, covered participants' roles, working methods, and support needs. The second survey additionally addressed requirements for visualization solutions in molecular tumor boards. These aimed to understand diverse requirements for preparation, implementation, and documentation. Nine semi-structured expert interviews complemented quantitative findings through open discussion.

Results: Using quantitative and qualitative analyses, we show that existing digital tools may improve therapy recommendations and streamline MTB case preparation, while continuous training and system improvements are needed.

Conclusions: Our study contributes to the field by highlighting the importance of developing user-centric, customizable software solutions that can adapt to the fast-paced environment of MTBs to advance personalized oncology. In doing so, it lays the foundation for further advances in personalized medicine in oncology and points to a shift towards more efficient, technology-driven clinical decision-making processes. This research not only enriches our understanding of the integration of digital tools into MTBs, but also signals a broader shift towards technological innovation in healthcare.

Keywords: Annotation; Clinical decision support systems; Genomics; Molecular tumor board; Personalized oncology; Precision medicine; Requirements analysis; User-centered design; Visualization; cBioPortal.

Fig. 1

Design of iterative requirements analysis

Fig. 2

MTB process model. A molecular analysis is ordered when a patient is admitted to the MTB. Lab technicians sequence the tumor to produce the raw sequencing data in FASTQ format (https://compgenomr.github.io/book/fasta-and-fastq-formats.html) (Sequencing). Sequencing data is aligned to a reference genome, and differences between the sequenced genome and the reference genome and variants (mutations) specific to the tumor are identified (Alignment & Variant Calling). The alignment process on the reference genome generates a BAM file (Binary Alignment Map) and a corresponding BAI file (Binary Alignment Index). Variant calling tools generate files in the Variant Call Format (VCF). Mutation Annotation Format (MAF) is a tab-delimited text file with aggregated mutation information from VCF files and is generated on a project level. From this point on, cBioPortal is deployed. The research expert then creates a case in the MTB tool by bringing together all relevant case data from different sources. Variants are annotated to determine their potential impact on protein function and relevance to cancer biology (Annotation). The annotated variants are interpreted in the context of existing scientific knowledge, clinical guidelines, and patient-specific factors. The findings are summarized for presentation at the Molecular Tumor Board meeting (Interpretation & Preparation for MTB). Variants are categorized based on their therapeutic relevance according to external sources. The research expert selects the most relevant variants and assigns them to the therapies identified in the literature review. The recommendations and rationale behind the proposed treatment plan are documented for the patient's medical records. Additionally, the sequencing data and MTB deliberations are archived for future reference and research purposes (Documentation & Archiving)

Fig. 3

Visualizing longitudinal clinical and genomic profiles of a patient in GENIE BPC A. The timeline provides a comprehensive overview of biopsies, resections, diagnoses, treatment regimens, and diagnostic assessments from medical imaging and oncology. B. Detailed genomic event tables offer insights into mutations, structural variants, and copy-number alterations observed in each sample. Specifically, the mutations table includes information on protein effects and variant allele frequencies. The 'samples' column specifies the sample(s) where each event was detected, with a dash indicating non-profiling for a particular gene. Genomic events are annotated with data from external resources, such as OncoKB, represented by a blue target icon. Hovering over these icons provides users with additional information via tooltips. C. An example tooltip demonstrates the OncoKB annotation for ALK G1202R, indicating that this mutation detected in sample 4 is a known resistance mutation against crizotinib. Example data adopted from the public cBioPortal (https://www.cbioportal.org, https://genie.cbioportal.org)

Fig. 4

(in reference to Fig. 2 of Reimer et al. [42]): Proposal for integrating cBioPortal into the clinical infrastructure. Site-specific ETL processes are used to get the data from HIS in FHIR. FhirSpark serves as a mediation layer between cBioPortal and a FHIR-capable server. Rather than relying on a legacy database that lacks FHIR, which would require synchronization or mapping to FHIR, instead FHIR is implemented as the native storage format. In addition, there is the initial, direct way to get data into cBioPortal

Fig. 5

Distribution of applied systems / applications / websites for the preparation of the MTB. (Open question with two possible answers: 1) I do not use any systems / applications / websites for MTB preparation and 2) So far, I have used the following systems / applications / websites for MTB preparation.)

Fig. 6

Satisfaction with the clarity of the presentation of the data for the case review for preparing the MTB

Fig. 7

Certainty of being well-informed based on the available data for a case and having made the optimal interpretation

Fig. 8

Requirement for novel visualization methods along different data types

Fig. 9

Derived requirements for novel visualization methods R1 and R2. A showing R1: cBioPortal currently supports the annotation of variants with information from several different databases, including OncoKB and CIViC. These databases provide a variety of information including the clinical relevance of mutations, common cancer types, and potential drug targets. We aim to add additional annotations to enhance the preparation process for molecular tumor boards. For example, we intend to include PubMed and Google Scholar annotations that automatically provide search results based on patients’ mutation profiles and clinical data to improve the literature search for potential treatment strategies. B showing R2: We aim to improve cBioPortal’s patient timeline for its application in the preparation and presentation of molecular tumor boards. While the current implementation already provides good visualization of the longitudinal clinical data, we plan to implement changes based on the findings of the requirements analysis, to enhance the readability and improve the integration into clinical processes, e.g. the option to toggle between an absolute and relative date format. We also intend to provide custom visualization solutions for novel data types