Experimental Functional Impact Annotation Definition and Scope

[mbrush]

We will initially proceed with our initial decision to split 'Predicted' from 'Experimental' Functional impact annotations - and model these as separate VA types. Our rationale was that:

• they have very different provenances and evidence types (computational vs experimentally validated)
• they make statements about functionality at different levels of granularity
• there is a fairly clean separation in terms of the processes, tools, and sources for these VA types
• there is a clear understanding in the community w.r.t. how they are different and how they can be used

The proposals/notes below are derived from the initial requirements work for this VA type here.

---

Definition: a statement about the impact of a variant on gene product function, as supported by experimental evidence.

Scope/Comments:

• These statements may describe the impact in more general terms (e.g. ‘loss of function’, ‘gain of function’), but they often describe more precisely how a specific function or activity of the gene product is affected (e.g. increased enzymatic activity, decreased localization to the nucleus)
• The key feature distinguishing these from "Predicted Functional Impact" annotations is their experimental basis (i.e. some in vitro or in vivo experiment was done to show the impact).
• Additionally, there is typically greater specificity in defining the impact (typically indicating impact on a particular function, as opposed to simply ‘damaging’ or ‘tolerated’).
• These distinctions are important because consumers of functional impact annotations trust and apply them differently than predicted derived annotations.

[mbrush]

Some examples of terms used to describe experimental impact:

More General Terms:

• CIViC Oncogenic mechanism terms describing functional impact: see Oncogenicity mechanism value set #28, and the example here.

  • gain of function, loss of function, neomorphic, unaltered function, other

• Muellers Morph terms: https://en.wikipedia.org/wiki/Muller%27s_morphs

  • Loss of function (Amorph, Hypomorph).
  • Gain of function (Hypermorph, Antimorph, Neomorph, Isomorph)

• DDG2P terms: https://www.clinicalgenome.org/site/assets/files/2757/fitzpatrick_ddg2p.pdf

  • loss of function, missense in frame, dominant negative, activating, increased dosage, cis-regulatory promoter mutation, uncertain

• Wormbase allele functional effect terms:

  • see Add allele properties monarch-initiative/GENO-ontology#3 (comment)
  • see Add allele properties monarch-initiative/GENO-ontology#3 (comment)

• Also of relevance: the 'Protein Function' section on p 7 of the ENIGMA article here: https://jmg.bmj.com/content/56/6/347.full

  • This set of generic 'levels/types' of impact (Complete, Partial, Gain of function, Dominant Negative) could be used as on component of a composed description of the complete impact ({level} + {function/activity} + {cellular location})

More Specific Terms

• NextProt: https://www.nextprot.org/entry/NX_P16422/phenotypes

  • defines a standardized set of impact terms/statements, e.g. gains localization in endoplasmic reticulum, decreases protein abundance

• Uniprot: https://www.uniprot.org/uniprot/P38398

  • often more specific but in free text, e.g. for M → R in BC: alters protein stability and abolishes ACACA and BRIP1 binding.

• ClinGen: http://dataexchange.clinicalgenome.org/interpretation/entities/AlleleFunctionalImpactStatement.html

  • e.g. desensitized Ca2+ binding to the thin filament
  • e.g. reduced Ca2 sensitivity of activation
  • e.g. the mutation lowered the Ca2+ affinity of the reconstituted thin filaments
  • e.g. Functionally Damaging
  • e.g. abnormal/deceased ERK activation
  • e.g. unchanged from reference

[mbrush]

Elements to capture in a statement model: (based on notes from initial requirements work here)

• variant - typically a precise variant instance, can be at genomic, transcript, or protein level
• functional impact - a description of the immediate effect of the variant the function or behavior of its gene product
• affected gene product - use in cases where a genomic-level variant is specified, and the impact applies to a particular transcript or protein isoform

---

The statement semantics here seem fairly simple, and map cleanly to an ACM-based model as follows:

• subject: Variation (1..1)
• predicate: code {has_functional_impact / has_phenotype} (1..1)
• object: string (1..1) ## impacts too diverse here to codify, as there is no existing relevant CV - but can provide recommendations on how to form strings in the documentation. See notes below.
• affectedFeatureQualifier: [ Transcript | Protein ]

Issues/Questions:

• As for predicted functional impact, understanding diversity of impact terms here, and deciding how to frame/constrain their entry, is key challenge.
• Consider the idea that these impacts can be framed as 'molecular phenotypes', such that this VA type becomes a specific kind of G2P annotation.
• And evidence and provenance information for this VA type could be particularly rich and important to capture in detail - given broad use of this VA type as evidence for clinical interpretations, the diversity and nuance of experimental design and execution, and the history of inconsistent interpretation and application of functional data in variant interpretation efforts (e.g. CSER studies)
• Look at what the ClinGen Functional Data Pipeline effort is doing here, and elicit requirements from them? Maybe have them present on a call - with focus on their use cases, requirements, and data
• Consider creative ways to add structure and consistency to the descriptor for this VA type. Will be important for integration and search of this data. Values will be diverse, esp when very precise impacts are describe (e.g. decreased binding to ER membrane) . . .but there may be a way to provide a recommended value set that covers most values, and/or define a 'Molecular Phenotype' object where the impact value can be composed from more fundamental terms (e.g. an entity-attribute like pattern as used in most phenotype ontologies). Efforts like NextProt that produce experimental functional impact data may already do something like this.

---

[ahwagner]

Relevant discussion on griffithlab/civic-docs#46 (comment). In short, this model is a subset of the functional evidence statements (EIDs) from CIViC, though it excludes the disease field (which will, going forward, always be set to "cancer"). Since there is (or more precisely, will be) no variability in the value set for that attribute, I think it is safe to model functional evidence from CIViC under the proposed Experimental Functional Impact Annotation.