[OperaHCI:18] COLUMBUS (or OMERO): Can you simply not figure out how you survived without it, or do you wish you could forget it existed?
Ghislain Bonamy
I am contacting you today to hear from users how much you like
Columbus.
For the one who already use Columbus:
1. Do you use it for large screens (>400K wells), medium size screens
(>20K wells), small screens (>100 wells), single plates, a bunch of
images, or do you have it but don't use it?
2. How many wells did you already collect in this database?
3. Do you store every image or merely pick and choose what to keep/
upload?
4. What is your favorite thing about Columbus?
5. What is your wish list (if any)?
6. Do you have issues with the data access or data presentation using
this tool?
7. Does the data structure fit your screening and informatics needs?
8. Are your external tools easily integrated with Columbus?
9. Do you have other devices (microscope, high content imagers etc.)
connected to Columbus?
10. Do you have other comments etc. not covered by the previous
questions?
If you are NOT lucky enough to have Columbus, but have evaluated it:
1. Are you planning to purchase it?
2. Are you waiting on something to be modified before jumping onboard
3. Is something deterring you from you from buying it?
4. Anything else you would like to add?
If you looked at/evaluated Columbus and decided NOT to purchase it:
1. Were there missing features you felt where critical?
2. Did you simply not see the use for it?
3. Was it too expensive for your budget?
4. Any other factor? Perhaps another solution presented itself?
If you are not using Columbus and/or never have heard of it:
1. Are you using something else that the file system provided through
the Evoshell?
Hopefully many of you will be able to take some time in a busy
schedule to share your experience.
Also if you have users, postdocs etc. who would like to share their
opinion and experience, please point them to this group.
Sincerely,
Ghislain
PS: Columbus is a very complex product trying to do a big job, so
please avoid being too passionate about any shortcomings it may have –
this forum is obviously focused on constructive feedback.
--
You received this message because you are subscribed to the Google Groups "Opera High Content Imaging" group.
To post to this group, send email to Oper...@googlegroups.com.
To unsubscribe from this group, send email to OperaHCI+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/OperaHCI?hl=en.
Matthew...@sanofi-aventis.com
I recently noticed this survey posted to the group and had some
questions. I would also be interested in hearing from other
high-throughput Columbus Gallery (CG) users.
- do you see CG as a good system for high-throughput HCS? What is
performance generally like?
- what is the best feature? What is the most critical gap? (Acapella
and its license type seems like a powerful feature)
- does CG capture well-level results from image analysis? (acapella and
other) field level results? Cell-level results?
- if well results are stored, are they accessible from the database
using SQL queries? (e.g. stored in a fully relational model). If
field/frame and cell level results are captured, how are they stored?
Can they be accessed from an API or SQL query?
- can datasets be managed flexibly by user groups and roles? (e.g.
delete data, rename certain fields etc)
- I noticed in slides from one of your presentations (I think SBS) that
you have demonstrated analysis of Acumen data with Acapella. This is
already very interesting, but I am curious if the images were captured
in CG. If yes, was it before or after analysis. Were image analysis
results captured?
- can annotations like compound/sample IDs be stored in the system? If
yes, how are they added (manually, automatically?). After they are
added can you search across many datasets by compound ID to find images?
- are there any out of the box web based views of data that can be
easily accessed through a URL-like API? (e.g. from Spotfire)
- is it possible to store and retrieve image views with segmentation
data from image analysis overlaid?
- how is support from PE? Do they customize the software if needed?
Are there public APIs that can be used to customize?
- do CG releases closely follow Omero releases?
- I am aware that recent Omero versions would duplicate image pixel
data. Does CG do the same? (if not for Opera data, what about other
image types?)
I also have one question for the group that is unrelated to CG. Do any
members have experience with using JPG 12 or 16-bit lossy image format
for image analysis (in matlab for example)? I am curious if this could
be a viable option since we generally write once and do not edit images.
The compression ratios are generally 10% and full bit-depth is retained.
Thanks in advance for any potential replies. Fyi, many of these
questions come from the challenges we currently face with our HCS
systems and workflow. I wouldn't expect CG to address all of these
challenges.
Best regards, Matt
Matt Smicker
Research Investigator - Information and Data Management
Scientific Computing and Data Management Department
sanofi-aventis
Bridgewater, NJ 08807-0800
Office: JR1-3329
Phone: (908) 231-3319
FAX: (908) 231-3686
eMail: Matthew...@sanofi-aventis.com
Dear all,
Sincerely,
Ghislain
avoid being too passionate about any shortcomings it may have - this
Ghislain Bonamy
Thanks for the interest. I have not collected much feedback yet and
will try to answer this to the best of my knowledge. Hopefully other
members will jump and answer some of our questions to in particular
the ones that I cannot answer. Please bear in mind that we do not
currently use OMERO or Columbus gallery for our screening systems at
GNF.
1) I think CG would be a great system for HCS, but this product is
not fully mature to address the needs of HT-HCI (ie imaging 1-2x10^6
wells for a single screen) in my opinion. This is mainly because of
the pixel duplication cf. points bellow. Perhaps there are users
outthere using CG or OMERO that can give their overall opinion.
2) Hopefully someone with experience with CG can answer CG
performance.
3) I believe that OMERO/CG collects well level measurements. Single
cell features where cleverly saved into outside HDF5 files (pytables)
in OMERO, though I see no reason why PE would have changed this
perhaps someone with inside knowledge can comment here. Also I believe
that CG can store cell outlines, but I am not sure how this is done.
4) There is a good API for OMERO/CG using ICE, which can be bound
with many different languages. I also know that data can be queried in
CG using SOAP/wsdl technology. Together there should be a good access
point to mine data in this database.
5) Hopefully someone with experience with CG can describe which tools
are available through CG to manage data.
6) As per mentioned we do not use Colombus though we are thinking
about getting it. The images acquired with the Acumen where not loaded
in Columbus, but images where converted into a format that can be
handled by Acapella and then analyzed using this tool.
7) For compound annotations etc. I do not believe that CG is designed
to be a compound management tool and allow you to annotate wells. That
being said, someone with experience with CG should probably give a
better answer.
8) I do not believe that there is such a tool. I have requested it
from PE and they mentioned that the API/SOAP services could be used
for this. I personally believe that it would be an easy thing to do.
We currently have a WebApp that allows you to grab images from
spotfire. It is not pretty but does most of what we want if you are
interested in it.
9) I believe retrieving images with segmentation is possible, though
I am not sure how it works behind the scenes or how they store the
ROI. http://trac.openmicroscopy.org.uk/omero/wiki/HcsPreview
10) I do not believe that PE would do much customization, but an end
user should probably answer this.
11) I believe that CG closely follows OMERO releases.
12) Yes, data duplication is still a very big issue. I have told PE
that we would not consider CG until it was resolved. I can only
imagine that for anyone running very large screen this will be an
issue to, especially if your archiving/storage pipeline is well
defined like ours. OMERO is supposed to release a version of OMERO.fs
that would alleviate this issue but I am not quite sure when this will
be and when it will be deployed in CG (http://www.openmicroscopy.org/
site/support/omero4/server/fs).
Regarding the JPEG storage, I think that using JPEG2000 would be a
better choice as it provides both lossless and lossy compression. I
also think that the lossy compression of JPEG2000 is much better and
provide images that are much closer to the original data (fewer
artifacts) than JPEG. Also according to Frans Cornelissen (please
correct me if I miss quote you), the Jpeg2000 lossy compression offers
a “minor” smoothing that improves image analysis. I have implement as
part of Bioformat a compression algorithm in the TIFF format (much
like one could use LZW…) This has the advantage of keeping the OME-
TIFF model while having a nice compression scheme. Perhaps something
you would find useful. The only disadvantage of using such compression
is that you have a compression/decompression overhead when reading
images/writing images.
Best,
Ghislain