Atlas Interoperablity I & II: progress to date, requirements gathering Session I: 8:30 – 10am Session II: 10:15 – 12pm
Interoperability requirements The big question: –Do the observed relationships hold across species (development phases, etc.) It is a component of community building What types of bridges we may build –Brain region homology (impossible?? Messy…) Topology, shape, metric relationships –Functional homology –Neurochemical homology –Developmental homology
How to build these bridges Ontologies: –Neuronames, UMLS, BIRNLex, BONFIRE… –Ontology alignments: a high priority action item –Standard ontology formats and shared ontology tools (BIRNLex) Coordinate systems –Absolute: Stereotaxic (which), Talairach –Coordinate translation services (??) working within species –Other types of location description (relative, ontology-based, expression) Standard formats and APIs –To access data –To query registries (metadata, ontologies, spatial, cross-walks) –To exchange data across atlases –To perform analysis: Find automatic segmentation tools (cells and tissues) and morphometric analysis tools (incl. cell counting and volumetric analysis) –To allow conceptual interoperability (across concepts used in different species)
Progress has been made… On the mouse brain atlas front end and query framework (MBAT’2007), 3D slicer, query atlas, human brain atlas On data preparation and upload: warping tools, HID/AID On the ontology front: –Formal management of ontologies and Bonfire translations; concept mapper, concept queries On the spatial front –Spatial alignment and registration (spatial registry), spatial query, multi-scale visualization Annotations –Combined spatial-semantic On cross-atlas interoperability (Atlas Interop API)
Concept Query Infrastructure UCLA & CC Term Source Database Search the DB at that column for results matching the query Mediator Gives the column and DB information matched to that TERM (BIRNLex/Bonfire) Holds some underlying “business logic” for the Query interface-categorizes data types and search criteria. Move functionality to this over time Example: User generates a query for calb1 in C57BL/6 in BIRN Microarray DB and GeneNetwork. (Metadata Database holds information for Interface to formulate Query) 1) Query Term Source DB for terminology from different sources calb1 and C57BL/6, and the output is: DATASOURCE = TABLE NAME FIELD NAME Gene Network = “Gene table” “genesymbol” Microarray = “UAD_probe_term” “key=GeneSymbol” 2) (Optional) Query Term Source DB for all fields of “Table Name” 3) To generate this query to mediator, “get all fields = Calb1 in the given Table of each DataSource” 4) Query mediator for all the matching results This infrastructure: Easier for the User Expandable More comprehensive searches of multiple sources Start migrating functionality from Interface to Server after Fall If on Server, easier for others to use our infrastructure Metadata Database Concept Query Interface
Spatial Query Infrastructure UCLA & UCSD Image Metadata Retrieve Images Spatial Registry ArcIMS Images Atlas API webservices Atlas Interoperabiity Server Atlas API User can query with ROI Uses Atlas Interoperability Server and API Visualize images using zViewer zViewer 2D images (integrated into a MBAT window) Spatial Query Interface
Information Sources Information Query UCLA, USC, CC BAMs BonFire JDBC webservices Mediator/webservices User can query two different information sources depending on needs of user 1)ontologies: useful for defining what a user means by a term and mapping data across data sources 2)BAMs information: connections, molecules and cells in different areas (It is unlikely we will have the needed time to make the necessary changes to expand this by the fall release-we will need to decide if we want to include it at all) Information Query Interface
Microarray Databases GeneNetwork Microarray Data Handling UCLA, UTHSC BIRN Microarray Upload Interface Microarray Upload Interface Barlow Database Smith Database MAGE XML New 2007: MAGE compatible-facilitates compatibility with other microarray sources Expanded Query of DBs More easily expandable to other sources More robust process More flexible queries Gene Expression Explorer (can be visualized in MBAT) URL access Mediator/webservices Microarray Annotation Database Concept Query Interface
Image Metadata Retrieve Images Spatial Registry 2D Image Data Handling UCLA, UCSD (CCDB, ArcIMS), Neurcommons (ABA), Stott Parker (Gensat) ArcIMS Images Atlas API ABA Images Gensat Images CCDB images Neuro- Commons Implement Stott’s Gensat DB CCDB webservices RDF/Sparql webservices URL access Mediator/webservices 2D Registration Workflow 2D Registration Workflow zViewer 2D images (integrated into a MBAT window) User can query by Concept or Spatial Query and visualize in zViewer Concept Query Interface Spatial Query Interface Atlas Interoperabiity Server
Handling multiscale images
Spatial-Semantic Annotation DEMO Spatial Registration DEMO
Arbitrary query of spatially distributed signals
State Exchange between SA-MBAT Atlas alignment problems… Transformation matrix wrapped in Coordinate Transformation Service
Additional desiderata Additional data types –Histopathology –Time series Anatomical Physiological –Behavior –Connectivity (wiring/microwiring) –Data from typical laboratory Standards in metadata, registration procedures, middleware tools, handling of ontologies, annotations, etc.
Human and Rodent Atlasing: what is in common, what is different Significant overlap in needs and functionality: Atlases in two roles: as the query/analysis framework, and as spatial/semantic data registration framework Handling large images, using specialized grid tools Creating, registering, managing and querying 3D reconstructions 2D Image registration: from common metadata registration to spatial registration to semantic registration Coordinate systems, and location exchange between atlases Image and 3D annotation Vocabularies/ontologies
Human and Rodent Atlasing: what is in common, what is different Differences: Data types: 2D vs 3D and reconstructions Upload and registration: Regular process of image acquisition and registration, vs multiple acquisition methods, metadata conventions and resolutions, multi- scale --> multiple atlas tools Multi-scale image registration, query and visualization Query: Queries defined by clinical needs; medical records connection Regulatory: de-identification Also: formats; ontology stores we connect to; diseases; Analytically-driven vs data registration/fetching-driven Canonical vs individual We come from different contexts – but let’s not duplicate where possible!
Panel discussion What is atlas interoperability in your domain? What are interoperability challenges and priorities? How tools and approaches from other testbeds can be re-used? What additional questions you would like to formulate once data and services from other testbeds become available? What could be immediate steps towards BIRN mashups? Now that we have MBAT, Slicer, Query Atlas, DTI & LDDMM – what is next?
Slicer Within the user interaction threshold – updating label and eventually concept in ontology – this could be an excellent semantic bridge to other atlases Connectivity data –Searching based on connectivity neighborhood Need a database of visual maps, and then pattern matching with the existing maps –Structural and functional connectivity
Canonical Atlas? In Human: atlas is a set of priors and not a canonical atlas Mouse: atlas is essentially a paper atlas over a new media, and a common framework Connecting atlases based on variation
Path forward Fetching data –How much to fetch, from control and target –Analysis of stat significance (from Slicer) –Disease models used in mouse – populate them with data; need to get human to mouse and mouse to human use cases (conceptua interperability?) For example: –AD: reduction of neuroactivation, esp. in temporal regions of hippocampus (Function BIRN so far focused on regions without strong homology to mouse): what genes in mouse (Genesat, ABA) – have expressions in ventral Hippocampus, and related connectivity differences; –APP expressions (overexpressed in ventral hippocampus) – AD –APOE as a potential cortical factor, in AD; what other genes co-vary with APOE Refinement of queries Visualization at different levels Adding analysis functions to atlases –E.g. counting axons, spines –E.g. comparing histograms, for different signals, different areas
The use case Search genes co-expressed with Apoe (highlighted in human studies) using ABA in hippocampus using NeuroBLAST Issues: –Poor data on connections Both nice graphics and interactive display –Ontology alignment –Adding analysis to mouse