Participants: Mario, Michael, Eva, Vincent, Rainer, Iza, Tomek
We have agreed to make a pilot installation of openBIS in Biozentrum and import real data into it.
Besides images and well-gene mapping we will import some analysis results to make the discussion about possible iBrain-openBIS integration more concrete.
There are 3 main goals of the pilot:
– define a clear interface through which openBIS and iBrain should be integrated. Besides simplifying the architecture it will allow iBrain to work independently or to integrate with other data management systems easily.
– judge the usability of openBIS as a productive data management solution in InfectX project and in Biozentrum labs daily work.
– define a list of crucial openBIS extensions to become a fully usable and user-friendly system.
Iza Adamczyk is the Biozentrum contact point from CISD, Tomek Pylak should stay in the loop for some time.
InfectX project setup
- there will be a separate machine for openBIS with 16 GB of RAM
- 3 microscopes running in parallel
- every microscope produces 1 plate per 4h
- TIFF images from each plate take 17 GB uncompressed, half of it after compression
- there will be around 1000 plates in total, acquisition should take few months
- groups involved in the project:
- Lucas Pelkmans - screening performed individually
- Cecile Arrieumerlou, Pascale Cossart, Christoph Dehio - screening performed in Biozentrum
- Urs Greber, Wolf-Dietrich Hardt, Ari Helenius - screening performed in LMC
- Data miners: Niko Beerenwinkel, Dagmar Iber, Christian von Meering
- Bernd Wollscheid InfectX
- it is not clear how long the research needed to start the screening campaign will take
- all data produced in the project should be accessible to all project participants, but should not be accessible outside
- each project participant wants to have his data on their site. We could think about a solution where there is one instance of openBIS, but with separate data store servers for Zurich and Basel. We should discuss at some point how does this relates to the existing openBIS instances in LMC and Lukas Pelkmans lab.
– plates have 16x24 geometry, 3 or 4 channels are used
– data will be shared by creating accounts in openBIS for external users or by exporting them to CIFEX
– plate libraries are not available at the moment, biologists are in the phase of testing - they are pipetting the oligos manually and write down the layout in the notebook. We should think about whether openBIS could provide partial LIMS functionality, i.e. to annotate and organize plate layouts.
OpenBIS - iBrain2 Integration Options
We have identified 3 levels of iBrain-openBIS integration. The pilot will make it possible to demonstrate the level 1 and 2. The level 3 can be demonstrated only partially, since the API for retrieving metadata is not yet written.
Level 1: OpenBIS stores only the links to data sets, provides GUI access
- Image data sets registration: iBrain creates absolute symbolic links to the data sets and puts them into OpenBIS dropbox. Data sets are therefore stored in the place managed by iBrain - OpenBIS holds only the links. Assumption: the files will not be moved/renamed, so the links in the OpenBIS can work preperly.
- Image analysis data sets registration: iBrain creates copies of the data sets and puts them into OpenBIS dropbox.
On this level of integration iBrain is not dependent in any way on OpenBIS, but users can use OpenBIS to browse the data sets and metadata.
Experiments and plates identifiers are provided together with the data set and appropriate entities are automatically created in OpenBIS. Properties of experiments or plates can be later updated by the users using Web Browser.
Level 2: OpenBIS stores the actual data sets
Just like Level 1, but instead of storing links to the image data sets in OpenBIS, the real data are put into dropbox. In this scenario OpenBIS stores the data in it's own structure, but a mirror view is provided that allows access to the data sets through the meaningful directory structure (e.g /experiment/user/plate/data_set).
Just like Level 1 or Level 2, but additionally iBrain uses OpenBIS (via API) as a metadata source.
Metadata which are collected at the moment by iBrain are entered by the user into openBIS.
Short term action points
- find libraries with well-gene (and maybe oligo) mapping for several plates
- generate image analysis results with one number per well describing the "out-of-focus level" for the images which we want to import
- extend the amount of RAM to 4 GB on the openBIS "pilot" machine
- mount the share on openBIS machine and copy the screens which should be imported
- CISD (we will start at the beginning of August, we should finish till the middle of August, but it has to be checked with Bernd if there are resources available):
- 3-4h implement a plugin which adheres to iBrain images naming convention and allows to specify well geometry configured in service.properties as a list of tile numbers. The current well layout is:
2 1 x
5 4 3
8 7 6
x 10 9
The layout in future will be:
x 1 2
3 4 5
6 7 8
x 10 9
- 2h write a plugin which creates "assays" (experiment) and plates on the fly.
- 4-6h openBIS deployment in Biozentrum
- install openBIS
- import images from 2 screens, from directories: pizaaro (10x zoom only), muntwill
- import "out-of-focus" analysis results, use unique iBrain plate names as openBIS dataset identifiers to link image analysis and image datasets
- configure heatmap plots for imported analysis results
- import the genes library
- mirror view
- 1h meet with Christof Kasper (PhD student from Cecile A. group) to see what MetaExpress can do
Identified mid-term requirements
- mechanism to mask well images which are of bad quality (e.g. out of focus) and should be rescanned (aka A/B structure in LMC). There can be several rescans, openBIS should be able to merge all datasets on the fly and present one view of the best images.
- allow to provide description for feature names
- the well layout (position of each tile - mapping form tile number to position in the well) should be configurable per experiment or even per dataset from GUI.
- thumbnails should be generated in the background to allow quick dataset registration
- support 4 channels (we currently support up to 3)
- provide a clickable plate view with thumbnails from all wells
- store analysis results per analyzed image