Re: BioRDF Telcon
Hi Helen, Please see my response below. Helen Parkinson wrote: Responses in line. 1. We have text mined much of the Affymetrix GEO data, curated it and imported it into ArrayExpress - there is now much better sample annotation than the native data in GEO. We also are running QC across all the data files so we know which should be excluded for future analyses. I think it's the right thing to do both to enrich data annotation and to enhance data quality. This will help data integration a lot. Currently, we are exploring query federation in the neuroscience context. It'd be great if we can use the neuroscience use case(s) to help drive your ontology development for text mining and data visualization. In addition to the NIH neuroscience microarray consortium, it may be possible to collaborate with the Neuroscience Information Framework (NIF) to see if we can utilize some of its resources (e.g., neuron ontology). Re-use of the neuron ontology is possible, but it depends on whether there is available data to annotate either in ArrayExpress or GEO. If you can get me a list of experiments accessions or pubmed ids I can see if this is feasible That's related to our current mciroarray use case where we're exploring a few examples (experiements) that hopefully contains enough neuron-related annotation. 3. We have summary level data of genes x conditions for ~30,000 hybs worth of data in our gene expression atlas with p values indicating relative under/over-expression. We are planning to export these as triples as soon as we publish the atlas - these may be of interest. www.ebi.ac.uk/gxa - there's an API at present, but it will be improved in the next month or so. It fits well with what we're currently exploring in terms of gene list representation and linking genes and samples to existing ontologies. It'd be great if we can download or fetch RDF triples from EBI atlas. We have a student starting work on this in a month, if you can produce concrete use cases for how you want to access these data we can do something. As part of the current biordf effort, we're in the process of coming up with questions that we can ask across neuroscience microarray experiments. 4. If neuroscience data is of specific interest we could do a themed atlas release where we add datasets for a given community or project and make these available. These can be identified by ArrayExpress or GEO accession or pubmed and we can re-annotate the genes vs Uniprot/Ensembl, add GO terms, etc and curate the sample attributes and experimental variables. These pipelines are already in place as part of our production workflow. I think it's a great idea to do a themed atlas (e.g., neuro-atlas). I just played with gxa a little bit. It's nice! For example, I could find genes that are over-expressed in the hippocampus brain region across different experiments. However, when I tried to do the same thing for neurons, there are only a few neuron types that I can select. It'd be nice if we can have more neuron types, for instance. This is probably as we don't have data - here's a list of human experiments with the term neuron - if any of these are useful, then I can prioritise their curation and inclusion in an atlas release http://www.ebi.ac.uk/microarray-as/ae/browse.html?keywords=neuronspecies=Homo+sapiensarray=exptype=pagesize=25sortby=releasedatesortorder=descending The experiment: E-GEOD-4757 compares gene expression prolifes of layer 2 stellate island neurons of the entorhinal cortex between normal and AD subjects The experiment: E-GEOD-9770 compares gene expression profiles of Layer II stellate neurons (entorhinal cortex) and layer III cortical neurons (hippocampus CA1, middle temporal gyrus, posterior cingulate, superior frontal gyrus, primary visual cortex) for subjects with mild cognitive impairment. These two experiments are also in the NIH microarray consortium. There is an overlap of neuron type, brain region, and phenotype (memory) between the two experiements. This may be a start ... and brain http://www.ebi.ac.uk/microarray-as/ae/browse.html?keywords=brainspecies=Homo+sapiensarray=exptype=pagesize=25sortby=releasedatesortorder=descending I'd be very happy to collaborate, and for this group to use our data, we spend a lot of time adding semantic value to it, so please let me know if this is of interest We are also looking into the possibility of establishing collaboration with the scientific discourse task force based on the microarray use case. We're planning to have a microarray-related presentation and discussion on Aug. 31 (Monday, 11 am EDT/5 pm CET). Details will be announced later. It'd be great if you can join the BioRDF call to participate in the discussion. Cheers, -Kei best regards Helen Kei Cheung wrote: The minutes for yesterday's BioRDF call are available at:
Re: Can RDFa be used on XML: pharma information
This may also be an interesting way of intersecting microarray (mageml) and semantic web (rdfa) ... -Kei Ivan Herman wrote: I am sorry if I come into this thread very late. Additionally to what Ralph just said, the RDFa distiller running on the W3C site: http://www.w3.org/2007/08/pyRdfa/ should actually work with an arbitrary XML file, although only SVG is 'announced' there (which is probably my mistake). If there is a problem then, well... it is my bug:-( Ivan Ralph R. Swick wrote: At 10:48 PM 6/23/2009 +1000, Rick Jelliffe wrote: I see that the 2008 draft http://www.w3.org/2006/07/SWD/RDFa/rdfa-overview says RDFa itself is intended to be a technique that allows for adding metadata to any (XML) markup document, including SMIL, RSS, SVG, MathML, etc. Note, however, that in the current state, RDFa is being defined only for the (X)HTML family of languages. The RDFa specification was designed with the intent that other languages than XHTML could take advantage of RDFa markup. (The terminology host language was used in some drafts to signal this direction.) The charter under which the group was operating was specific to XHTML, thus the wording in the W3C Recommendation. So I think I will go ahead and add some RDFa markup to the XML, By all means, reuse the RDFa vocabulary if it seems appropriate for your application.
RE: Can RDFa be used on XML: pharma information
hi kei, there is already something better than RDFa tags in MAGE-ML, the OntologyEntry tags. Their purpose is exactly to provide the information to link to the semantic web. The examples you provided from NIH are well annotated with those tags. cheers, michael Michael Miller Lead Software Developer Rosetta Biosoftware Business Unit www.rosettabio.com -Original Message- From: public-semweb-lifesci-requ...@w3.org [mailto:public-semweb-lifesci-requ...@w3.org] On Behalf Of Kei Cheung Sent: Friday, July 24, 2009 10:56 AM To: Ivan Herman Cc: Ralph R. Swick; Rick Jelliffe; public-semweb-lifesci@w3.org Subject: Re: Can RDFa be used on XML: pharma information This may also be an interesting way of intersecting microarray (mageml) and semantic web (rdfa) ... -Kei Ivan Herman wrote: I am sorry if I come into this thread very late. Additionally to what Ralph just said, the RDFa distiller running on the W3C site: http://www.w3.org/2007/08/pyRdfa/ should actually work with an arbitrary XML file, although only SVG is 'announced' there (which is probably my mistake). If there is a problem then, well... it is my bug:-( Ivan Ralph R. Swick wrote: At 10:48 PM 6/23/2009 +1000, Rick Jelliffe wrote: I see that the 2008 draft http://www.w3.org/2006/07/SWD/RDFa/rdfa-overview says RDFa itself is intended to be a technique that allows for adding metadata to any (XML) markup document, including SMIL, RSS, SVG, MathML, etc. Note, however, that in the current state, RDFa is being defined only for the (X)HTML family of languages. The RDFa specification was designed with the intent that other languages than XHTML could take advantage of RDFa markup. (The terminology host language was used in some drafts to signal this direction.) The charter under which the group was operating was specific to XHTML, thus the wording in the W3C Recommendation. So I think I will go ahead and add some RDFa markup to the XML, By all means, reuse the RDFa vocabulary if it seems appropriate for your application.
Re: BioRDF Telcon
Hi Michael, Thanks for your detailed description of mageml. For our use case, we probably don't need to use all the information captured in mageml. The types of information we are currently focusing on include experiment/sample annotation (including some provenance as you indicated) and gene lists and how they are linked to existing ontologies. A couple of convincing examples may be enough to start. I can relay your comments about the validity of mageml to the consortium, although I don't know whether they can address them. Cheers, -Kei Miller, Michael D (Rosetta) wrote: hi kei and helen, like helen, i've been following the HCLS working groups with great interest. as one of the designers, with helen, of the MAGE-ML and MAGE-TAB specs i might be able to provide a little technical insight into the formats. (from helen) This is probably as we don't have data - here's a list of human experiments with the term neuron - if any of these are useful, then I can prioritize their curation and inclusion in an atlas release kei, are the NIH Neuroscience Microarry Consortium exeriments you've cited and others like them in GEO or ArrayExpress? a set of those could be a good starting point for helen. My understanding is that the publicly visible mciroarray projects in the neuroscience microarray consortium should also be in geo and/or arrayexpress, although I don't know whether all the annotations are preserved. first, MAGE-ML is based on a DTD[1], not an XSD. in early 2002 as the OMG Gene Expression specification[1] was being finalized, XSD was still in its infancy so we weren't comfortable at that point generating a XSD. the MAGE-OM UML[2], in a very early XMI format from Rational Rose and UniSys, was used to generate the DTD with code we wrote ourselves[3]. the UML model was designed to capture the flow of a microarray experiment and how the resulting arrays were organized in the experiment based on how the samples were treated and/or on the samples' phenotypes for the purpose of a reviewer understanding the methodology and for a researcher replicating and/or re-analyzing the results. some of the details of the flow may not be of much interest, i.e. it might be worth simply connecting the BioSource elements with their gene expression data and not worrying about how the hybridization was performed. but that depends on what you want to do and you know that better than i. also, the data itself are specified in external files, typically in a white-space delimited format where the column headers are specified in the MAGE-ML file in the QuantitationTypeDimension element and the identifiers of the row specified in one of the three DesignElementDimension elements, Feature, Reporter, CompositeSequence, depending on how derived the data is. Also the data can be in a vendor specific format such as the Affymetrix CEL (since the CEL file internally specifies the dimensions often they are left out of the MAGE-ML document). the ExperimentalFactor elements are certainly relevant and if you've looked at some of the examples you will noticed that the BioSource elements, in particular, and other elements are annotated by OntologyEntry elements. from the gene expression specification: OntologyEntry A single entry from an ontology or a controlled vocabulary. For instance, category could be 'species name,' value could be 'homo sapiens' and ontology would be taxonomy database, NCBI. for the element an ontology entry element is annotating, we looked at it as a way of specifying something like the object identified by the element is an instance of the class/individual specified by the OntologyEntry so from kitm-affy-droso-176167 one sees that the BioSource is an instance of Drosophila, whole animal, whole head and an age of 3 days: BioSource identifier=arrayconsortium.tgen.org::biosource.181527 name=Oregon R head 3d Characteristics_assnlist OntologyEntry category=Organism value=Drosophila description=Drosophila OntologyReference_assn DatabaseEntry accession=#Organism URI=http://mged.sourceforge.net/ontologies/MGEDontology.php#Organism; Database_assnref Database_ref identifier=MO/ /Database_assnref /DatabaseEntry !-- snip -- /OntologyReference_assn /OntologyEntry OntologyEntry category=OrganismPart value=whole animal description= OntologyReference_assn DatabaseEntry accession=#OrganismPart URI=http://mged.sourceforge.net/ontologies/MGEDontology.php#OrganismPar t Database_assnref Database_ref identifier=MO/ /Database_assnref /DatabaseEntry /OntologyReference_assn !-- snip -- /OntologyEntry OntologyEntry
Re: Can RDFa be used on XML: pharma information
This is probably technically possible - but you'd need to process a lot of complex mage-ml to get out some quite simple information - there's a node-edge sample processing graph, plus all the external data files in there - mage-ml is mostly tags and the files are large. We've moved internally to MAGE-TAB format, we have a MAGE-TAB parser that's being used by a couple of groups. We will be developing a standalone parser/backend database which will allow users to build a standalone atlas. There may be more mileage in developing that parser further to support RDF than to persue MAGE-ML. thanks Helen Kei Cheung wrote: This may also be an interesting way of intersecting microarray (mageml) and semantic web (rdfa) ... -Kei Ivan Herman wrote: I am sorry if I come into this thread very late. Additionally to what Ralph just said, the RDFa distiller running on the W3C site: http://www.w3.org/2007/08/pyRdfa/ should actually work with an arbitrary XML file, although only SVG is 'announced' there (which is probably my mistake). If there is a problem then, well... it is my bug:-( Ivan Ralph R. Swick wrote: At 10:48 PM 6/23/2009 +1000, Rick Jelliffe wrote: I see that the 2008 draft http://www.w3.org/2006/07/SWD/RDFa/rdfa-overview says RDFa itself is intended to be a technique that allows for adding metadata to any (XML) markup document, including SMIL, RSS, SVG, MathML, etc. Note, however, that in the current state, RDFa is being defined only for the (X)HTML family of languages. The RDFa specification was designed with the intent that other languages than XHTML could take advantage of RDFa markup. (The terminology host language was used in some drafts to signal this direction.) The charter under which the group was operating was specific to XHTML, thus the wording in the W3C Recommendation. So I think I will go ahead and add some RDFa markup to the XML, By all means, reuse the RDFa vocabulary if it seems appropriate for your application.
Re: BioRDF Telcon
Hi I meant to comment on this, I would not attempt a mage-ml-RDF transform, I can probably do something more quickly with an rdf export n of transformed data analysed for over/under expressions plus factor values and genes and we'll have a student to work on this I hope Helen Miller, Michael D (Rosetta) wrote: hi kei and helen, like helen, i've been following the HCLS working groups with great interest. as one of the designers, with helen, of the MAGE-ML and MAGE-TAB specs i might be able to provide a little technical insight into the formats. (from helen) This is probably as we don't have data - here's a list of human experiments with the term neuron - if any of these are useful, then I can prioritize their curation and inclusion in an atlas release kei, are the NIH Neuroscience Microarry Consortium exeriments you've cited and others like them in GEO or ArrayExpress? a set of those could be a good starting point for helen. first, MAGE-ML is based on a DTD[1], not an XSD. in early 2002 as the OMG Gene Expression specification[1] was being finalized, XSD was still in its infancy so we weren't comfortable at that point generating a XSD. the MAGE-OM UML[2], in a very early XMI format from Rational Rose and UniSys, was used to generate the DTD with code we wrote ourselves[3]. the UML model was designed to capture the flow of a microarray experiment and how the resulting arrays were organized in the experiment based on how the samples were treated and/or on the samples' phenotypes for the purpose of a reviewer understanding the methodology and for a researcher replicating and/or re-analyzing the results. some of the details of the flow may not be of much interest, i.e. it might be worth simply connecting the BioSource elements with their gene expression data and not worrying about how the hybridization was performed. but that depends on what you want to do and you know that better than i. also, the data itself are specified in external files, typically in a white-space delimited format where the column headers are specified in the MAGE-ML file in the QuantitationTypeDimension element and the identifiers of the row specified in one of the three DesignElementDimension elements, Feature, Reporter, CompositeSequence, depending on how derived the data is. Also the data can be in a vendor specific format such as the Affymetrix CEL (since the CEL file internally specifies the dimensions often they are left out of the MAGE-ML document). the ExperimentalFactor elements are certainly relevant and if you've looked at some of the examples you will noticed that the BioSource elements, in particular, and other elements are annotated by OntologyEntry elements. from the gene expression specification: OntologyEntry A single entry from an ontology or a controlled vocabulary. For instance, category could be 'species name,' value could be 'homo sapiens' and ontology would be taxonomy database, NCBI. for the element an ontology entry element is annotating, we looked at it as a way of specifying something like the object identified by the element is an instance of the class/individual specified by the OntologyEntry so from kitm-affy-droso-176167 one sees that the BioSource is an instance of Drosophila, whole animal, whole head and an age of 3 days: BioSource identifier=arrayconsortium.tgen.org::biosource.181527 name=Oregon R head 3d Characteristics_assnlist OntologyEntry category=Organism value=Drosophila description=Drosophila OntologyReference_assn DatabaseEntry accession=#Organism URI=http://mged.sourceforge.net/ontologies/MGEDontology.php#Organism; Database_assnref Database_ref identifier=MO/ /Database_assnref /DatabaseEntry !-- snip -- /OntologyReference_assn /OntologyEntry OntologyEntry category=OrganismPart value=whole animal description= OntologyReference_assn DatabaseEntry accession=#OrganismPart URI=http://mged.sourceforge.net/ontologies/MGEDontology.php#OrganismPar t Database_assnref Database_ref identifier=MO/ /Database_assnref /DatabaseEntry /OntologyReference_assn !-- snip -- /OntologyEntry OntologyEntry category=OrganismPartRegion value=whole head description= !-- snip -- /OntologyEntry !-- snip -- OntologyEntry category=Age value=Age OntologyReference_assn DatabaseEntry accession=#Age URI=http://mged.sourceforge.net/ontologies/MGEDontology.php#Age; Database_assnref Database_ref identifier=MO/ /Database_assnref /DatabaseEntry