These are chat archives for opencobra/cobrapy

25th
Aug 2016
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 09:37
@pstjohn Yes, we'll set up a meeting shortly. We're probably going to send out a doodle first.
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 09:49
Regarding lines between cobrapy and cameo, we had discussions with @zakandrewking and Ali in the past. So we have a rough plan of potentially moving cameo.core into cobrapy.core, which would change the solver interface (not breaking compatibility with the old ones though in the short term), thus making it easier to define more complicated optimization problems in the future. On the other hand we don't want to aggravate the community by doing this, so it would be good to get some feedback from folks that are not at UCSD/CfB. Also, we need to make sure that performance of model formulation doesn't drop in order to keep cobrapy as the basis for building and running ME models (we're fairly certain now that this will not be an issue).
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 09:54
@pstjohn regarding pathway visualizations, if you're interested in the automatic layout of pathways, there are plans to add this capability to Escher in the long-term. Actually I anticipate Escher to become a lot more feature rich soon since both PDB (http://www.rcsb.org/pdb/pathway/pw.do) as well as the bioinformatics company Genialis (https://www.genialis.com/; they're involved in a joint EU project with us) will develop/use it further.
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 10:48
[Henning Redestig, Sequencing, Informatics & Modeling] For a planning meeting / discussion on the future of cobrapy, please fill-out this doodle to help picking a good time spot: http://doodle.com/poll/ierr2d996i87pyqw
Peter St. John
@pstjohn
Aug 25 2016 13:31
Honestly I was surprised when I first started using cobrapy that it bothered to maintain its own solver interfaces :), so thats definitely a change I would support. For one example, it would make things like parsimonious FBA much simplier to implement if we could just use absolute values in the objective function rather than converting / reverting from irreversible models.
Peter St. John
@pstjohn
Aug 25 2016 13:44
On the visualization front, that sounds great. my main thought is that these models can be really tough to understand, so it would be great if cobrapy provided some really clear, interactive summary and visualization methods. My main concern with escher is that the optimization -> visualization step seemed slow, which would stop me from using the visualization as a diagnostic tool.
The d3.js one I hacked together, while definitely not as advanced in functionality, had the nice feature that you could display the result directly in the jupyter notebook and store all the map data in the model json.
I can't say I'm optimistic on automated layouts... Mainly I don't really see them being as good as a manually generated one :). but certainly anything that speeds the process would be useful: if you could then edit it in the escher interface that would be nice.
This message was deleted
Peter St. John
@pstjohn
Aug 25 2016 13:51
Here's a basic example of plotting in the jupyter notebook:
blob
Christian Diener
@cdiener
Aug 25 2016 16:32
Hi. Does the doodle automatically adjust for timezone?
May also be good to put some prior info in Google Doc, Trello Board, etc. to get some idea what everybody considers most important and so on. This would make the Hangout much more productive
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 16:34
I am confused, one of the best features of Escher is that you can display maps in jupyter notebooks? And I totally agree with that interactive summary and visualization methods for diagnostics would be nice.
Henning said that the doodle adjusts for time zone but we'll double check
Christian Diener
@cdiener
Aug 25 2016 16:36
For me all times appear in the evening. Which would be really early in the morning for you guys.
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 16:37
We're using trello for our stuff already, how about creating a trello board for cobrapy?
Christian Diener
@cdiener
Aug 25 2016 16:37
Actually would be like midnight - 2am for you I think.
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 16:38
Ok, that sounds wrong. @hredestig can you fix this? Should be in the evening for us.
Christian Diener
@cdiener
Aug 25 2016 16:39
Ok thats fine. Just put in a comment that the times are Denmark Time.
Nikolaus Sonnenschein
@phantomas1234
Aug 25 2016 16:41
Also, we're organizing this conference here http://cph-bioscience.com/conferences/data-driven-biotechnology-bench-bioreactor-and-bedside all expenses paid (not sure about flights) @pstjohn @cdiener might be a good occasion to get together?
Christian Diener
@cdiener
Aug 25 2016 16:44
Sounds good but will have to check funding. Travel expenses are usually not paid for postdocs here...
Christian Diener
@cdiener
Aug 25 2016 16:55
We're also using Trello for project management here so I would be up for that option. With an idea board where everybody puts what he considers important (or comments on cards that already describe the topic)
Peter St. John
@pstjohn
Aug 25 2016 20:36
@phantomas1234 , didn't know that escher works in jupyter notebooks - neat. Its been a while since I used it, I'll have to look back into it.
Peter St. John
@pstjohn
Aug 25 2016 22:34
Also, thanks for the conference link! I think that could work for us actually - I'll put in an abstract. Michelle O'Malley was actually on my PhD committee.