New harmonized knowledgebases update (November 9th)
The MTBP annotates the relevance of the analysed gene variants by using a variety of in-house tools and publicly available resources (referenced in the MTBP reports as appropriate). Among the latter, the portal employs a number of knowledgebases gathered by international efforts and open for the (non-commercial) access and feedback of the community.
These knowledgebases are under continuous updates, and we aim for the public version of the portal to be aligned with those. However, aggregating the content of these external resources requires first to harmonize their different variant representation and data models, which in the MTBP is implemented by an in-house semi-automatic pipeline. However, manual efforts to map certain terms are still required, which thus impedes doing the process in real-time. Instead, updating the knowledgebases content in the MTBP is performed via periodical harmonized (external) data releases, which we try to do regularly.
Note also that these updates have been halted for certain knowledgebases due to recent changes in their respective license policies, which we are currently working to address. Importantly, the version/date of each of the resources employed for the MTBP analyses is under version control, and the details are provided in the top right corner of the given report (see image).
-- firstname.lastname@example.org , November 6th 2020
Failed run when larger number of variants are passed via the free text box solved (October 12th)
We have recently updated one of the variant mapping tools employed by the public MTBP system (the Ensembl Variant Effect Predictor). This new version raises a default error when a larger number of variants is inputted if they are not ordered by chromosome position. However, we did not have this in place for variants passed via the ‘free text’ box of the public MTBP interface, which therefore caused a run error. This issue has now been fixed.
-- email@example.com , October 12th 2020
New report to provide the public MTBP analyses results (September 29th, 2020)
The results of the functional and predictive variant analyses provided by the public MTBP portal are now both included in a single document (two different html reports for each were used before).
On the one hand, the evidence supporting a functional (or neutral) role of a given variant to (putatively) acquire tumor traits is shown in the ‘Functional relevance evidence’ column. Of note, the functional interpretation is variant/allele centric and thus no further considerations (e.g. the sample’s cancer type) are taken into account here (e.g. an allele loss-of-function is always classified as so regardless of the context in which it is observed). Biomarker information is also integrated for the functional relevance interpretation, under the premise that a specific variant acting as a cancer biomarker is a functionally relevant event by definition.
In addition, variant information that is not conclusive (or relevant) for the functional classification but has been curated by one the employed knowledgebases (such as in an inconclusive ClinVar assertion or reported with a low allele frequency in population genomics databases) is now also included in this column (previously, such information was only shown in the advanced gene-variant view).
On the other hand, information about the predictive relevance (cancer prognosis, diagnostic and/or drug response) of the variant is now included in a new column named ‘Reported biomarker(s)’. In the public version of the MTBP, this analysis is solely based on the biomarkers reported by several public knowledgebases (CIViC, OncoKB and CancerGenomeInterpreter), which are updated in the portal on a regular basis (but not accessed in real time). Please note that this information is automatically aggregated by the annotation pipeline, as opposed to the MTBP instances employed by clinical projects such as Cancer Core Europe, in which the main actionability flags are manually curated based on in-house criteria. The public MTBP report provides a summary of the available information and the access to the corresponding knowledgebase entry for each given assertion via interactive popups (see image below).
Of note, the biomarker interpretation is a phenotype-centric analysis, and thus considerations such as the cancer match (i.e. whether the biomarker has been described for the sample’s cancer type or a subtype thereof) and the current clinical relevance of that biomarker (such as ‘reported by early clinical trials’) are taken into account here. In detail, the matched biomarkers are automatically tiered following a modified version of the ESMO-ESCAT scale of actionability.
-- firstname.lastname@example.org , September 29th 2020