From: Josh G. <jog...@in...> - 2018-01-11 15:52:33
|
Hi Bradford, Yes, for the data we currently have, the image_cvtermprop approach would work. However, we are starting to see (and we expect to see a lot more) image data of Drosophila embryos at single-cell transcriptome resolution ( http://science.sciencemag.org/content/sci/358/6360/194.full.pdf). It would be very useful to take embryo images with single cell expression data coordinates and cvterms, store that data in Chado, and then take advantage of the geometric functions/operators in PostgreSQL to perform some interesting queries. That said, I'm happy to go the image_cvtermprop route as an initial phase and keep geometric points in mind for possible future implementation. Josh FlyBase On Thu, Jan 11, 2018 at 7:06 AM Bradford Condon <bra...@gm...> wrote: > Hi Josh, > > Thanks so much for your input. > If you’d like the GitHub issue for the eImage module is here, so feel free > to chime in there. > > https://github.com/GMOD/Chado/issues/31 > > It looks like your polygon data consists of a cvterm to map the polygon > to, and a set of coordinates ie `139,116 70,117 73,270 135,212 139,116`. > If thats all correct then it sounds like image_cvtermprop would work, > unless using a table withthe polygon type would work better for your > database? > > > pBradford Condon > Postdoctoral Scholar, University of Tennessee Knoxville > www.bradfordcondon.com > > > > > > On Jan 10, 2018, at 11:50 PM, Josh Goodman <jog...@in...> wrote: > > > > > > * Are you using the eimage table? No > > * have you written eimage linker tables? No > > * Do you have strong images about changing the eimage table itself to > image? No > > > > I would like to request that an image module allow for storing polygon > annotations associated with images and linking those annotations with CV > terms. FlyBase has around a thousand or so images that have been annotated > but are not in our production Chado instance. For example, on > http://flybase.org/reports/FBim0000898, you can mouse over the image and > see the anatomy CV terms associated with particular areas or vice versa. > > > > One open question is whether to store the coordinate data in a prop type > table (e.g. image_cvtermprop) or opt for a table that uses the polygon > geometric type. > > > > I can throw this into a new or existing github issue if you want. > > > > Josh > > > > On Wed, Jan 10, 2018 at 5:37 PM Bradford Condon < > bra...@gm...> wrote: > > Dear Chado community, > > > > We are working on improvements for Chado 1.4 at PAG! You can see the > issues we are working on here. One of the improvements we are making is an > image module from the expression eimage table. > > > > We would like to hear from Chado users who use the expression module's > eimage table. We have a few options, but we are leaning towards keeping > eimage base table as is, and creating the linker tables as _image. This is > the least disruptive option, as the eimage table already exists (even if it > would hold all images, instead of just expression images.) > > > > * Are you using the eimage table? If so, to hold what sort of images? > > * have you written eimage linker tables? If so have you named them > _eimage or _image? > > * Do you have strong images about changing the eimage table itself to > image? > > > > Cheers > > > > Bradford > > > > > > > > Bradford Condon > > Postdoctoral Scholar, University of Tennessee Knoxville > > www.bradfordcondon.com > > > > > > > > > > |