Grounding Assist: Capture hints in paper

[jvwong]

Description

Q: What is the name of the feature?

A: Grounding Assist

Q: What does this feature enable the user to do?

A: Indirectly, disambiguate a name for a bioentity (e.g. gene) more accurately

Q: What information must the user provide to use the feature?

A: (1) Article information (2) names of bioentities

Q: What are the applicable constraints, e.g. compatibility or performance?

A: There main cases to consider:

1. Default: No prior information is available
2. Bioentity database identifiers are available
3. Species information is available

Q: How does this feature affect each class of user (persona)?

A: Synonyms and orthologues account for a large proportion of observed errors (30%). It is conceivable that other types of errors could be mitigated (e.g. spelling issues) and that hints would enable features such as a true "type-ahead" autocomplete.

• Uses

  • curation: in normalization
  • post-submission: in an automated error flagging system (even if not available at curation time)
  • triage: e.g. classifier to more accurately identify potential articles and authors
  • information extraction: e.g. context provided by authors

• Users

  • Biologist: Eventually, better search across deposited data, better discovery
  • Editor: Increased quality and trust in the accuracy of Biofactoid data
  • Computational biologist: Increased fidelity of Biofactoid data, better data integration
  • Curator: Increased fidelity of Biofactoid curation

Specification

Sources of bioentity information

• Considerations
  • Entity types
  • Consistent concepts (gene product, family)
  • Compatible Identifiers
  • Scope
  • Accuracy (curated vs NLP)
  • Format (file, web service)
  • Latency (seconds)
  • Hardware (GPU)
• Providers
  • Curated
    • PubMed
  • Natural Language Processing
    • PubTator3
    • Reach

Scoring algorithm

This is to be determined. Should consider:

1. Location: Prioritization based on mention in title vs abstract vs body
2. Type: Local hint (e.g. entity database IDs) vs global (e.g. species)
3. Reliability of source

Tasks

The factoid project should be responsible solely for obtaining bioentity hints for a given article:

• 1. Define a Hint model
• 2. PubTator3 Hint provider
• 3. Organism Hint ranking
• 4. General Hints API
• 5. Retrieve and store Hints on create/update of Document
• 6. Augment grounding-search query with Hints

At least for network curation, grounding-search should be responsible for scoring search hits in light of hints.

References

• Entity normalization
  • Chen, L., Liu, H. & Friedman, C. Gene name ambiguity of eukaryotic nomenclatures. Bioinformatics 21, 248–256 (2005)
  • Gyori, B. M. et al. Gilda: biomedical entity text normalization with machine-learned disambiguation as a service. Bioinform Adv 2, (2022)
  • Wei, C.-H. et al. GNorm2: an improved gene name recognition and normalization system. Bioinformatics 39, btad599 (2023)
• Entity identification
  • Luo, L. et al. AIONER: all-in-one scheme-based biomedical named entity recognition using deep learning. Bioinformatics 39, (2023)
• Species
  • Pafilis, E. et al. The SPECIES and ORGANISMS Resources for Fast and Accurate Identification of Taxonomic Names in Text. PLoS ONE 8, e65390 (2013)
  • Wei, C.-H. et al. SR4GN: A Species Recognition Software Tool for Gene Normalization. PLoS ONE 7, e38460 (2012)
  • Luo, L. et al. Assigning species information to corresponding genes by a sequence labeling framework. Database 2022, baac090 (2022)
• Applications
  • Wei, C.-H. et al. PubTator 3.0: an AI-powered Literature Resource for Unlocking Biomedical Knowledge. arXiv (2024)

[jvwong]

• Organism ranking edge case examples
  • Intentionally mixed species
    • The rapidly evolving X-linked MIR-506 family fine-tunes spermatogenesis to enhance sperm competition. eLIfe 13 2024
  • Parallel pathways in orthologues
    • A CDK1 phosphorylation site on Drosophila PAR-3 regulates neuroblast polarisation and sensory organ formation. eLife 13 2024