collect interface requirements

[ahwagner]

this issue is for bringing attention to specific search features or strategies that we want to incorporate into our search interface (for both the variant-centric and VTB views), to drive our data modeling and backend design choices.

Example: The OncoKB variant view includes important elements to consider for a variant, including a protein view and prevalence of the variant across multiple high-level disease types.

[bpitel12]

Sorry for the delay in posting this. Here is the feedback we've had so far on the user interface and what tools might be helpful. I've added a lot, so we'll need to prioritize/tackle accordingly. Some of these may evolve into their own issues at some point, but wanted to get it all written before I become unavailable for a while. :)

Interface Requirements, Wish List, and Suggestions based on large VICC-VTB calls to date:

• Ability to have multiple viewing modes: proposed therapy-matching mode; variant interpretation mode (cohort/whole genome profile/single variant); and data export mode.
• Collaborate with ClinGen Somatic working group to integrate Cancer Genetics rapid publication into VTB component. Maybe we could have a way of marking variants as not being represented in the knowledgebase cohort triggering a prompt for the rapid publication form (work with Subha)
• Populate a Search Results summary for variant interpretations and therapy-matching modes. To discuss in more detail with large group.
• Ability for interface to accept a VCF file or similar large profile-based input (.xls/.csv/.seg/.txt)
• Ability to search/query wild-type variants (pertinent negatives) in context of cancer profile
• Harmonization of variant synonyms (i.e. FGFR1 amp vs. FGFR1 amplification)
• Classification by Tier1/Tier2 evidence in addition to current evidence levels.
• Adopt a more automated disease-classification system (easier to do a comprehensive, disease-specific search)
• Ability to filter between pediatric vs. adult disease
• Information about resistance mutations that might negate effects of variant-drug matches could be available.
• Consider adding a "consensus metric" for variants with similar (or dissimilar) associated evidence across knowledgebases.
• Better way to search drug families (i.e. EGFR inhibitors)
• Expand searchable variant types to include: CNVs, TMB, epigenetic, expression profiles, immunotherapy biomarkers, biochemical pathways. Might be nice to integrate with diagnostic testing method as an "advanced query mode"
• Along with expanding searchable variant types - it might be handy to able to "relax" to a gene-level query. (use with discretion per clinicians on large call)
• Rename "Paper" as search criteria and maybe don't use as default for therapy-matching mode.
• Attribute names will need to be more intuitive in an outward-facing site.
• Find a way to annotate therapeutic/disease-course history to help refine therapy-matching.
• Add a statistical prediction for ongogenicty for rare variants. (https://github.com/cancervariants/metakb/issues/12)

[bpitel12]

Of note, St. Jude's has a visualization tool I really like called ProteinPaint. https://pecan.stjude.cloud/proteinpaint/etv6. Especially nice for gene fusions and ITDs, which are often not represented on this scale. It also allows the user to switch from Protein view to DNA view and toggle between transcripts.

[bpitel12]

Sorry - I left out two suggestions from the VICC calls.

• Addition of ClinVar star category to variants - related to evidence level
• Addition of European standards in addition to AMP/ASCO/CAP guidelines

[shrutir]

Clinical Trials

• The current state of the cross-KB doesn’t appear to have updated clinical trial information.
• How often should clinical trials get recycled through the KB?
• Do we want to only display clinical trials that are active/recruiting?
• Can we include the inclusion/exclusion criteria from clinical trials

[ahwagner]

Gordana emphasized the need for different preset search views:

• Variants + Tumor type
• Variants alone
• Tumor type alone

And said primary source linkouts were very helpful for context.

[bpitel12]

FYI - St. Jude's recently launched a new visualization tool called GenomePaint, which is a really great complement to their ProteinPaint feature mentioned above. GenomePaint displays intronic and extragenic mutations as well as indels/fusions in genomic context not just in the context of the resulting protein. Lots of other useful information available on the website too. Not sure if we'll use it for VICC/VTB, but wanted to make this group aware. https://proteinpaint.stjude.org/genomepaint/

[ahwagner]

Thanks @bpitel12! Their code isn't open-source, but it's possible that they would be open to collaboration when we start building our interface. We'll keep this in mind, although hopefully an open-source solution presents itself between now and then and makes this easier for the rest of the community!