BTW, we've explored doing similar manual curation of microarrays for microbes… for example, looking at the experimental descriptions from the french dataset. I can't really site definitive success, and I won't say it's worth doing, but having the conditions for the experiments is useful. What I will say is we are working on regulatory models that link gene expression to stimuli. Now, if this works and works well, one can imagine running the model in reverse: you tell me what genes were expressed, I'll tell you the conditions of the experiment. I would love to get to a place where we could do this…. then begin a systematic bi-dectional comparison of "actual" experimental conditions with those predicted by the model… this then becomes analogous to what we're doing with phenotype data in metabolic models. I'm curious about the ontology they're using. The same ontology should ideally be used for stimuli in our regulatory models. On Apr 22, 2013, at 11:28 AM, Thomas Brettin <[email protected]> wrote:
Manual annotation == a human reads text that describes the experiment, and creates structured metadata for the experiment and associated microarrays.
On Apr 22, 2013, at 11:25 AM, Robert Olson <[email protected]> wrote:
OK, that's good. I just don't understand what manual annotation of microarray samples means. Seems a strange concept.
On Apr 22, 2013, at 11:25 AM, Thomas Brettin wrote:
I've just been watching. He seems to be moving towards a position that is similar to what you, Ross, Terry et. al took when you originally created the expression data model in the CDM.
t
On Apr 22, 2013, at 11:22 AM, Robert Olson <[email protected]> wrote:
do you guys have any idea what he's talking about and if matters?
On Apr 21, 2013, at 4:45 PM, He, Fei wrote:
Hi All,
In this email, I want to talk about why I suggest to stop manual annotation of plant microarray samples.
Currently, our plant expression service provide a function, get_eo_samplelist. This function can retrieve all the microarray samples, which are detected under a specific environmental condition (i.e. environment ontology term). Obviously this is useful. For instance, I am interested in getting all the microarray experiments under drought condition, I can easily use this function to get them.
However, this is based on manual annotation. We have to read the description for each microarray sample, then assign a ontology term with it. After reading hundreds of description file of microarray samples, I found it is impossible to automatic this step. Can we assign a person with the task of annotate each microarray sample added in KBase?
On the other hand, a microarray experiment often contains a group of samples. For instance, sample A is under drought condition while sample B is under normal condition (control). The biological insights should be generated by comparing A and B (find differential expression). In fact, 'get_eo_samplelist' will only return all the samples (from different experiments) under drought condition. The corresponding control will not be returned.
As I understand, this is because, based on all the samples associated with an ontology term, we can build co-expression networks. But, here is another question, the co-expression network based on a list of drought samples(from different experiments) makes no sense to me(I might be wrong). To make things simple, I'll talk about co-expression networks in another email.
Certainly, we have spent a lot of time working on along this track. Myself is also involved in the manual annotation of microarray samples. After the Plants Science Retreat, I realized that, for the microarray samples, we should store the data and provide tools to handle, analysis and display the data. The manual curation is useful but easy to fail (Again, I might be wrong.).
More specifically, for the plant microarray, we should store: 1) normalized expression abundance; 2) description for each sample. We should provides tools: 1) to search and retrieve the data; 2) to identify differentially expressed genes; 3) to build co-expression network; 4) to display.
Everyone is welcome to convince me 1) we should not stop manual annotation for each microarray sample; 2) we should not stop store pre-computed co-expression networks for each ontology term.
Best,
Fei He
Research Associate
Dr. Sergei Maslov's Lab
_______________________________________________ Kbase-devel mailing list [email protected] https://lists.kbase.us/mailman/listinfo/kbase-devel
_______________________________________________ Kbase-core mailing list [email protected] https://lists.kbase.us/mailman/listinfo/kbase-core
_______________________________________________ Kbase-core mailing list [email protected] https://lists.kbase.us/mailman/listinfo/kbase-core