Talk:DbNPFeatures

From BioAssist
Jump to: navigation, search

Study capturing

Editing of templates

I (Jildau) think that not all users should be able to edit a template, especially for the pre-defined templates I would propose to appoint a administrator. Suggestion (Jildau): In the editting page a list of suggested items should be included (in order to normalize the additions as much as possible).

Events on subjects or on groups

Should an event refer to subjects or to groups? Both seem plausible, yet we have to make a choice in the data model. We can let events refer to groups only, to subjects only, or make both possible. The latter seems the most flexible approach, allowing events to be defined on groups (such as treatments), and also on individual subjects (such as premature death). However, this choice also makes the data model more complex. Any thoughts on this? Jildau: I think an event should always refer to a subject. Most often a group is defined by the events. Therefore I am still in doubt whether groups are necessary.

Events and samples

The sampling of all subjects in a group (e.g. a blood draw) is one event. Yet, this one event leads generally to as many samples (biological materials) as there are subjects in the group. For that reason, it makes sense to define two different entities here: the sampling event, and the resulting samples. Jildau: I would think an event refers to a single subject. Therefore, the event sampling combined with a subject leads to a certain sample.

Study subjects/culture

Jahn: Two questions

  1. Is the "identifier" different from the "Species / identifier"?
  2. Does this require disambiguation between NCBI and NEWT terms; if so, how are the two ontologies matched; should two ontologies at the same time be allowed? (After all, the point of an ontology is to reduce ambiguity and not to introduce more.) I don't know in how far the two ontologies overlap but they appear to be different.

Furthermore, there seems to be a difference between this NEWT browser and the taxonomy given by the second link in the text which refers to uniprot.org.

Kees: You are right about the differences between those taxonomies. I think we will have to choose one, which one, is up to the biologists (testers) to decide. I put NCBI first, they also have a good web service (which we could query using GroovyWS via [WS-client http://grails.org/plugin/ws-client]), so I suggested to Jeroen that he should use that one in the study capture wizard.

Study protocols

Jildau: There are two types of protocols. The study protocols and the analytic protocols. The study protocols define what happened to the subjects and how sampling was performed. The analytic protocols define all steps after sampling and is stored in the omics parts of dbNP.

Jahn: This sentence is borken or in complete "Only protocols that apply to the study design should be stored with the study, protocols describing sample handling and ??? The following information should be stored about protocols: ". Please clarify.

Kees: Done.

Kees: Protocols are descriptions of methods that are applied to subjects or samples in the study. There are two types of protocols: protocols that describe study design, and protocols that describe specific omics assays. The first is stored with the study metadata, the latter in the metadata of the specific omics modules. Regarding the first type, they are stored in a central protocol database, and specific protocol applications are described as events.

Transcriptomics

Jildau: Besides microarrays how are qPCR data going to be stored?

Export function for Affymetrix Command Console Software

Question from Guido: It would be nice if dbNP had an export function for the Affymetrix Command Console Software, so that if we capture our study design in dbNP, we do not need to type the sample details (tissue, subject age, etc.) again. Otherwise, we could probably use the ID from the dbNP database to identify the sample in the transcriptomics sample processing workflow in the lab.

User Interface of dbNP

Jildau: I would start with a relative easy user interface (no free text), containing a browse function and some simple queries (e.g. give me all studies that included treatment x)

User groups

Jildau: I think that the role GroupXStudyAdministrator: should be linked to the study owner.

Browse studies

Jildau: The data should also be downloadable in csv-format.

Jildau: The study titles should be clickable, and link to the study overview of the selected study. Not only of one study, it should be possible to select several studies and see/export these data in an aligned way.

Jildau: It would be nice if the treatment (events) and species/culture of a study are also shown. This is most likely the most often used selection criterium.

Event view

Jildau: I think the event view should be general (not only focussed on sampling)

Group view

Jildau: I would exchange the group view. I do not understand why we need groups if we already have events (but I can understand that defining cross-over designs will be easier).

Creating studies

Jildau: comment on "The interface may assume that every subject will just be the member of one group". This depends for what reason groups are defined.

The biomarker layer

Kees: A hard nut to crack for the dbNP architecture design, is how to implement the request for generic queries on specific types of data, such as patterns of metabolite concentrations in metabolomics and differential gene expressions in transcriptomics. The solution that is now (as of version 1.0) implemented in the requirements, entails that each omics module should expose specific biomarkers, that are linked to assays stored in the study capture part. These biomarkers have a general output format, and the implementation is of course left to the module. Implementation of the biomarker layer in the different modules basically means rendering a biomarker view on the omics data that is present in the assay for the targeted samples. Currently, all modules are developed in the same development environment (Grails), and therefore the code for this can easily be shared. An additional idea here could be to create also a derived biomarker layer within the query module of dbNP, to enable combination of biomarkers, statistic tests, or even unsupervised learning methods such as clustering and pattern recognition on the subjects, but this is out of the current scope of the project.

Jildau (quoted by Kees): the name biomarker suggests a too specific meaning for biologists. Please rename. Kees: Suggest to use 'clean data layer'.

Subject metadata

Kees: Where should for example information about the cage that mice were kept in go? It would probably make most sense to put this as a field of the subject information in the mouse studies template. But it is also possible to store it in the transcriptomics module (and expose it as a 'biomarker'?).