Making (not only life sciences) data resources more Interoperable and Discoverable on the Web

Alasdair J.G. Gray �Bioschemas Steering Council Chair�Heriot-Watt University – ELIXIR-UK

​

​

NFDI InfraTalk

4 April 2022

(Bio)schemas:

Bioschemas: Markup for the Life Sciences

Picture: Carole Goble, Turing Lecture 2018

Schema.org: Enhanced Search Results

Picture: Carole Goble, Turing Lecture 2018

Global, lightweight vocabulary of terms

What we can say about those things

• Properties�(1,447)

What we are talking about

• Types�(792)

Google Search

http://bioschemas.org

5

Google Search

http://bioschemas.org

6

Oct 2020

Google Dataset Search (Nov 2021)

http://bioschemas.org

7

https://datasetsearch.research.google.com

https://www.blog.google/products/search/making-it-easier-discover-datasets/

Datasets

Schema definition:

• Dataset: A body of structured information describing some topic(s) of interest�http://schema.org/Dataset
• ~120 properties including:
• name
• description
• isFamilyFriendly

​

8

Google Dataset Profile

Google Dataset Profile

• 2 required properties
• Used for Google Dataset Search
• 10 recommended properties
• Link to DataCatalog
• Link to DataDownload

​

Other profiles: Events, Jobs, ...

https://developers.google.com/search/docs/data-types/dataset

Bioschemas: Schema.org for the life sciences

Profile over Schema.org model + Bioschemas extensions

Layer of constraints + documentation

​

• Aim: Improve data discoverability and interoperability in the Life Sciences
• Approach:
• Add Life Science types to Schema.org
• 6 Merged into Pending
• Provide usage guidelines and examples
• 6 Minimal properties
• Link to domain ontologies
• Support software

​

Data model

Minimum information

Controlled vocabularies

Cardinality

Documentation

New (properties | types)

Bioschemas �Profile

Data model

Marginality (Minimum | Recommended | Optional)

Controlled vocabularies

Cardinality (ONE | MANY)

Documentation

Examples

New types and properties

Bioschemas: Lightweight semantics

• Many domain ontologies
• Designed to model biology
• Bioschemas focus on search!

​

• Globally unique identifiers
• Community defined enriched metadata
• Indexable by search engines

• JSON-LD/RDFa
• Link to controlled vocabularies
• Links to other resources

• License
• Provenance

• Retrievable
• HTTP

Bioschemas Community

bioschemas.org/�liveDeploys

bioschemas.org/�liveDeploys

bioschemas.org/�liveDeploys

83

​

​

Sites

Markup

Live Deployment List�bioschemas.org/liveDeploys

Existing Deployed Markup

MolecularEntity profile

• ChEMBL
• Guide to Pharmacology
• MassBank Europe
• Scholia

​

ChemicalSubstance Profile: https://bioschemas.org/profiles/ChemicalSubstance

MolecularEntity Profile: https://bioschemas.org/profiles/MolecularEntity

Adding Bioschemas to a Resource

1. Decide on page type: e.g. Gene, Protein, ...
2. Map to Bioschemas profile
• Use generator
3. Embed JSON-LD�into pages
4. Register live deployment

http://tiny.cc/bs-live-deploy

Profile Creation Process

Mapping

Profile

Use cases

Mockup

Adoption

Testing

Application

Bioschemas: Profiles & Deployments

Released Profiles

• ChemicalSubstance
• ComputationalTool
• ComputationalWorkflow
• DataCatalog
• Dataset
• Gene
• MolecularEntity
• Protein
• Sample
• Taxon

​

​

Picture: Carole Goble, Turing Lecture 2018

100+ Deployments: Logos of some sites with Bioschemas markup

Exploiting Bioschemas Markup

Specialised Search: TeSS

http://bioschemas.org

23

29 November 2018

• contact
• description
• endDate
• eventType
• hostInstitution

• location
• name
• startDate
• …

​

Bioschemas Course:

• contact
• description
• endDate
• eventType
• hostInstitution

• description
• keywords
• name
• about
• abstract

Bioschemas CourseInstance:

Bioschemas TrainingMaterial:

toxicology

No concept merging

Bioschemas Profile for Workflows�

270+ workflow management systems

Bioschemas profile

- Minimum Information for Registering a Computational Workflow

- Creators, inputs, outputs, WfMS type ….

Working with WfMS providers to extract and add Bioschemas markup

Workflow Registry for discovery

Workflow Package

for exchange & portability

Search

Mark-up,

Validation

https://bioschemas.org/profiles/ComputationalWorkflow/

Community Registry: IDPcentral

http://bioschemas.org

25

29 November 2018

FAIR community registry of Bioschemas metadata

Concept merging

IDP Data Sources

• Curated from publications
• Structure disorder
• Functional annotation
• 2,038 protein entries

• Experimental and predicted
• All known protein sequences
• 189M protein entries
• 2,074 included in sitemap

• Deposition database of protein structural assemblies
• 172 protein assemblies
• 90 distinct proteins

26

Bioschemas Markup for IDP

• Protein
• SequenceAnnotation
• SequenceRange
• Taxon
• Dataset
• Scholarly Article

Not shown:

• DataCatalog

27

Legend

Red: Schema.org

Blue: Pending Schema.org

Green: Bioschemas

BMUSE: Bioschemas Markup Scraper and Extractor

• Data harvester
• List of URLs and sitemaps
• Extracts markup
• JSON-LD or RDFa or both
• Static or dynamic
• Returns markup with provenance
• Where
• When
• Tool version
• File per page of harvested markup

​

https://github.com/HW-SWeL/BMUSE

https://app.swaggerhub.com/apis-docs/swel/BMUSE/

Harvested Markup

• Markup harvested through standard API
• HTTP Get requests
• Saves time as common API and model (Bioschemas) for all sites
• Harvested markup is page centric
• Multiple sites represent the same concept
• Sites use their own IDs

Identity Reconciliation

• All sources include cross-reference to UniProt (schema:sameAs)�
• Differences in UniProt URL
• https://www.uniprot.org/uniprot/
• http://purl.uniprot.org/uniprot/

​

• Queries extract UniProt accession�
• IDP-KG IDs generated as�https://idpcentral.org/id/<accession>

IDP-KG: Merging Entries Options

1. Pick canonical source�Useful for values like name
• Supplement with other entries when values are missing
2. Concatenate entries from all sources�Useful for values like synonym
3. Keep all values with provenance of source�Wikidata approach

Keep all values with provenance �of source

Data Verification

• Process tested with small samples of data with known properties:
• Number of proteins per source
• Overlap of proteins
• Full harvest run and overlaps verified against sources
• Query to analyse the number of proteins by dataset groups

32

IDP-KG Interfaces

SPARQL Endpoint

Annotation count per protein

REST API

Querying IDP-KG

34

Querying IDP-KG

35

Count by type

Protein information

Annotation count per protein

Annotation information

Annotation count by term code

36

Bioschemas Data Harvest

https://github.com/HW-SWeL/COVID-19

Project 29:

#29_bioschemas

Alban Gaignard, Leyla Jael Garcia Castro, Alasdair J. G. Gray / Online

Pages harvested

413,748

Sites harvested

25 (partially)

https://swel.macs.hw.ac.uk/data/

Global Research Graph: OpenAIRE

http://bioschemas.org

39

29 November 2018

Markup needs to have:

• Sitemap: Currently only 28/73
• Link to
• Publication
• Dataset

EOSC-ENHANCE Bioschema use-case

VALIDATE

REGISTER

Added-value services

Slide credit: Paolo Manghi, ISTI, CNR, Pisa, Italy

Project 29:

#29_bioschemas

Alban Gaignard, Leyla Jael Garcia Castro, Alasdair J. G. Gray / Online

Connecting TeSS and bio.tools

https://github.com/bio-tools/biotoolsShim/blob/master/json2rdf/BioSchemas-Tess-Bio.Tools.ipynb

Work by Alban Gaignard �(ELIXIR-FR)

Automated Data Curation

http://bioschemas.org

• description
• keywords
• name
• provider
• url

Bioschema DataCatalog:

• alternateName
• citation
• dateCreated
• licence
• …

​

Data Exchange: Without an API�MarRef → BioSamples

https://github.com/EBIBioSamples/bioschemas_marref_demo/blob/master/Summary.md

Bioschemas �Scraper

Rich snippet generation

Summary

Bioschemas: Markup for the Life Sciences

Picture: Carole Goble, Turing Lecture 2018

Bioschemas

What?

• Exploiting schema.org to make Life Sciences resources more discoverable
• Search engines will index and understand markup

How?

• Extending Schema.org vocabulary for life sciences
• Guidelines on how to markup resources
• Software to support deployment and consumption

Approach can be followed by other domains!

Future Challenges

• Harvesting is compute and time intensive
• Both provider and consumer
• Solutions around data dumps being discussed
• Data quality is very variable
• Requires validators

49

Acknowledgements http://bioschemas.org/people

http://bioschemas.org

Join Bioschemas: http://bioschemas.org/howtojoin/

Bioschemas�Community Call

4th Monday of the month

17:00 CET, 25 April 2022