Model Curation Edmund J. Crampin Auckland Bioengineering Institute

1
Model CurationEdmund J. CrampinAuckland
Bioengineering Institute

• CellML Curation
• MIRIAM
• Automated Curation
• What is a valid reference description?
• Going Further
• models and data
• community forum

2
Curation of CellML Models

• http//www.cellml.org/notes/curation-document/
• Level 0
• not curated
• Level 1
• encoded model represents published model
• Level 2
• encoded model reproduces published results
• Level 3
• encoded model satisfies domain-specific
  requirements
• Levels are not a hierarchy Level 2 compliance
  may break Level 1 compliance

3
Level 2 Validity Checking

• Validation
• Simulates reference set of results
• Simulation / Graphing Metadata (Andrew and Andre)
• Reference Description
• Script process of running model, generating
  plots, validating against published results

4
Domain-Specific Requirements

• Units
• Biophysics
• Mass balance
• Charge balance
• Osmotic balance
• Thermodynamic consistency
• Mathematics
• ICs
• BCs
• well-posed

5
MIRIAM

• Minimum Information Requested In the Annotation
  of (Biochemical) Models
• Le Novère et al., Nature Biotechnology 23
  1509-1515, 2005
• http//www.ebi.ac.uk/compneur-srv/miriam/
• Arose from ICSB 2004 Heidelberg
• Proposal in 2 parts
• Correspondence to a Reference Description
• Set of rules for dealing with encoding, structure
  and results when simulated
• Annotation scheme
• documentation of model

6
MIRIAM Reference Correspondence

• Encoding in public, machine-readible format
• Compliance with standard in which encoded
• Explicit relationship to a set of results
• Model structure should reflect biological
  processes
• Instantiation in a simulation (all quantitative
  attributes defined ICs, parameters etc)
• Instantiation of simulation reproduces set of
  results listed in 3.

7
MIRIAM Annotation Scheme

• Attribution Annotation
• association of the model with reference
  description and encoding process
• External Data Resources Annotation
• link model components to corresponding databases,
  ontologies (etc.)
• allows models to be queried

8
MIRIAM Attribution Annotation

• Name of model
• Citation (of Reference Description)
• Name(s) of creator(s) of encoding
• Time and date of encoding and modifications
• Statement of terms of distribution

9
MIRIAM Resources

• External data resources annotation
• Database and web services to map Unique Resource
  Identifiers (URIs) to worldwide open access
  bioinformatics resources
• Implemented in biomodels.net

10
Automating Curation

• What can be automated?
• What must be manually curated?
• Units (Jsim, Jonathan Cooper Oxford)
• Checking of biophysical properties? (mass,
  charge, osmotic balance?)

11
What is a valid reference description?

• MIRIAM
• Scientific article
• Any other unique publication, print or online,
  which describes precisely the structure of the
  model, parameters, and expected output
• Encoded model must correspond to a single RD.
• Better (?)
• Can an annotated CellML model be its own
  reference description?
• ModelML?

12
Going Further

• Relationship between model and data
• Databasing of data used to fit models
• Automated validation of model results against
  newly added data sets
• Community commentary
• Comments from authors
• Critiques from community
• Expert information
• Moderated forum?