Given a list of variants and the patient’s cancer type, the MTBP generates reports classifying their functional and predictive relevance for cancer.
The functionality of variants refers to their biological relevance for promoting disease phenotypes. This can inform –among others– the patient stratification to genomic-guided clinical trials (e.g. prioritize a FGFR inhibitor trial for those tumors with FGFR3 mutations estimated as activating) and the genetic counseling referral (e.g. if germline variants in cancer predisposing genes are estimated as pathogenic).
The predictive relevance of variants refers to their value for the disease diagnosis, prognosis and drug response according to the biomarkers reported by available clinical and preclinical studies. This analysis can reveal –among others– opportunities for investigational therapies or off-label prescriptions according to current knowledge.
Three distinct types of evidence are employed to classify gene variants as either putative functional or putative neutral in terms of the disease phenotypes.
First, the MTBP inspects whether a given variant has a curated effect compatible with a functional or a neutral role available across the queried knowledgebases (evidences type A). Assertions supported by lower quality data are filtered out as appropriate.
Second, variants without conclusive evidences of type A are evaluated according to bona fide assumptions (evidences type B). These include the identification of likely disrupting events in tumor suppressors following the guidelines for germline variants analysis in Mendelian disease genes.
Third, the remaining variants –evidences A not conclusive and evidences B not pertinent– are evaluated by computational-derived metrics. These calculations are based on methods supported by robust statistical evidences and benchmarked for the specific question addressed here.
The analysis is based on the biomarkers of disease diagnosis, prognosis and drug response currently described by clinical and preclinical studies and as gathered by expert-curated knowledgebases. The relevance of these biomarkers depends on the strength of the evidence supporting the biomarker effect (such as data from randomized clinical trials or case reports), the presence of co-occurring genomic events that can interact and modify that effect, and how the characteristics of the tumor (such as the cancer type) compares with that in which the biomarker has been originally described.
Tumor biomarkers are classified in distinct levels of clinical actionability following the recent recommendations of the European Society for Medical Oncology, which extends those previously published by American expert associations.
A number of biomarkers are currently described not by specific protein changes
(such as BRAF V600E and vemurafenib response) but by functional events (such as BRCA1/2 “loss-of-function” and olaparib response). These biomarkers are matched with the variants detected in the corresponding gene when these are classified as putative functional according to the MTBP analysis. In addition, the functional annotation is also used to identify mutation repurposing opportunities, which points out gene variants that may have an effect similar to known biomarkers.
The Cancer Core Europe (CCE, https://www.cancercoreeurope.eu/) was recently formed by seven leading European oncology centers to collectively implement new therapeutic strategies involving knowledge, data and technology sharing. Under this umbrella, we have developed the Molecular Tumor Board Portal (MTBP), a system employed by the CCE to interpret the germline and somatic tumor variants of advanced cancers and guide the treatment prioritization. The CCE is currently formed by the seven following members: Cancer Research UK Cambridge Centre, German Cancer Research Center & National Center for Tumor Diseases, Gustave Roussy, Instituto Nazionale dei Tumori di Milano, Karolinska Institutet, Netherlands Cancer Institute and Vall d’Hebron Institute of Oncology.
Data harmonization enables the compatibility of the different systems used to describe the variants detected in the germline and/or tumor material (input data) with that used by the resources employed to annotate them (metadata). Gene variants represented by existing standardized formats (e.g. VCF or HGVS) and those represented by arbitrary terms (e.g. EGFR “insertions at exon 20”) are converted to a common and unambiguous nomenclature. On the other hand, attributes of the metadata (e.g. the cancer type in which a certain variant effect is observed) are also harmonized to adapt and process their content. Data harmonization is supported by an array of existing bioinformatics tools as well as in-house methods and manual mapping efforts.
Having the data harmonization in place, the detected variants are then annotated with the results of clinical, population and/or experimental studies gathered by a comprehensive set of expert-curated knowledgebases. For variants with little or no information available in these knowledgebases, a number of bona fide biological assumptions and computational analyses are used to estimate their effect. The MTBP prioritizes true positive assertions; therefore, only higher quality studies (such as those supported by well-powered data), stringent assumptions and statistically robust predictions are used.
The annotation of the detected variants takes into account a number of considerations when compared to the metadata, as to intersect the genomic alterations at different levels of resolution or to match the cancer types depending on the disease ascendants/descendants classification.
Variants classified as putative functional, unknown significance and putative neutral are displayed in three separate summary tables. These tables include the main descriptives of each gene and variant as well as the evidence(s) supporting the classification. A detailed view (accessed by clicking the gene symbol box) includes additional information of interest, such as the comparison of the characteristics of the detected variant with those found in previous cancer cohorts, as well as curated information and computational estimation details that are not conclusive for the detected variant classification.
Of note, certain sequencing parameters (such as the germline and/or tumor origin of the variants or their allele frequency) cannot be retrieved by the generic variants file specifications used in the public version of the MTBP, and thus they are not incorporated in the report. Similarly, features that are project specific (as to create automatic alerts for variants associated with clinical trials of interest) are not included.
Biomarkers found in the tumor are displayed in separate tables depending on their current level of actionability following ESMO guidelines classification. Results are aggregated by gene variant and effect. Biomarker’s affected drug(s), cancer type(s) and level of supporting evidence are displayed as originally reported in the corresponding knowledgebase(s). The comparison of the detected variant and that reported as biomarker (e.g. whether is an exact nucleotide and/or aminoacid change coincidence or a variant fulfilling a broad genomic or functional event) as well as the comparison of the patient’s cancer type and that described for the biomarker effect (e.g. coincidence with the described disease or its subtypes or broad coincidence in case the biomarker is reported as pancancer) is also shown.