These are chat archives for openworm/ChannelWorm

10th
Jun 2015
Travis Jacobs
@travs
Jun 10 2015 11:31
Hey guys
@VahidGh @slarson
Stephen Larson
@slarson
Jun 10 2015 11:32
hello!
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:32
Hi
Travis Jacobs
@travs
Jun 10 2015 11:32
I'm finally back after a busy weekend, so I've got a bit of catching up to do, especially with what has been going on in the cwfitter branch that @VahidGh is working on
How are you guys doing?
Stephen Larson
@slarson
Jun 10 2015 11:33
Good
btw @VahidGh I think your observation about Peak vs. Steady State I/V curve is a good one
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:34
@travs, great. I hope you had a good exam :)
Travis Jacobs
@travs
Jun 10 2015 11:34
Haha thank you Vahid!
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:35
@slarson, yes. we should discuss a bit more about this
Stephen Larson
@slarson
Jun 10 2015 11:35
indeed
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:35
And a good news, @miladjafary also finished the first version of his app
Stephen Larson
@slarson
Jun 10 2015 11:35
Woo hoo!
Travis Jacobs
@travs
Jun 10 2015 11:35
Sweet!
Stephen Larson
@slarson
Jun 10 2015 11:36
Where to begin?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:36
Tonight he will lunch it
Needs some more work and integration with the fitter
Stephen Larson
@slarson
Jun 10 2015 11:37
ok cool
excited to have a look
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:37
And we now have the built-in digitizer which stores digitize data along with other parameters within a DB
blob
Stephen Larson
@slarson
Jun 10 2015 11:39
Sweeeet!
Travis Jacobs
@travs
Jun 10 2015 11:39
Looks cool!
Stephen Larson
@slarson
Jun 10 2015 11:39
is that bootstrap?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:40
blob
@slarson, yes, in addition to a Meteor Theme
Stephen Larson
@slarson
Jun 10 2015 11:40
oh nice and the digitizer is built in there
great guys; so what’s next?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:43
Tonight we will integrate it with the fitter, so we can get the data from DB
Stephen Larson
@slarson
Jun 10 2015 11:44
Cool
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:44
And then we can add PyOW
Stephen Larson
@slarson
Jun 10 2015 11:44
great progress you guys!
OK in terms of our conversation today; were we planning to have a look at some more images we would queue up for the pipeline?
Travis Jacobs
@travs
Jun 10 2015 11:46
Yes
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:46
Yep
Travis Jacobs
@travs
Jun 10 2015 11:46
Stephen Larson
@slarson
Jun 10 2015 11:46
ok and we had said we’d start somewhere different this time, no?
Travis Jacobs
@travs
Jun 10 2015 11:47
Yeah, K channels
Kv1?
I believe?
Stephen Larson
@slarson
Jun 10 2015 11:48
sounds good
travs you want to try taking the lead this time like i did last time? :)
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:48
yes, slo-2 is also one with high priority
Travis Jacobs
@travs
Jun 10 2015 11:48
Sure!
So I'm over here making a new hierarchy for Kv1
We can start with the Fawcett paper
And on p5 we have a couple nice graphs already it seems
2C
Just reading the description now
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:51
yes, everything we need
Travis Jacobs
@travs
Jun 10 2015 11:52
And there are actually two ion channels on each of these
SHL-1 with the white markers
and SHK-1 with the black
Stephen Larson
@slarson
Jun 10 2015 11:52
yep
Travis Jacobs
@travs
Jun 10 2015 11:53
The left graph is from Xenopus oocytes, and the right graph is from C elegans myocytes
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:53
yes, actually this paper exists also in Kv4 directory, and I think has info about both of them
Travis Jacobs
@travs
Jun 10 2015 11:55
Ok
So in terms of using these graphs for digitization, I suppose both would be useful, but the C elegans would be higher priority
So I will note that in a comment on the sheet
Stephen Larson
@slarson
Jun 10 2015 11:56
cool
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:56
If I'm right Xenopus oocytes is a medium for expressing the channel
Stephen Larson
@slarson
Jun 10 2015 11:56
yes that’s right
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:57
and actually the channel which is expressing, is form C. elegans
Travis Jacobs
@travs
Jun 10 2015 11:57
OK
Stephen Larson
@slarson
Jun 10 2015 11:57
So this is looking pretty good
Travis Jacobs
@travs
Jun 10 2015 11:58
Ok
so it looks like there are no more IV curves in this paper
Stephen Larson
@slarson
Jun 10 2015 11:59
nope
Travis Jacobs
@travs
Jun 10 2015 11:59
Let's move on to this one
Stephen Larson
@slarson
Jun 10 2015 11:59
OK
oh actually that doesn’t direct link to the paper
Travis Jacobs
@travs
Jun 10 2015 11:59
Oh, one sec
Vahid Ghayoomie
@VahidGh
Jun 10 2015 11:59
@travs, let's also include I- or V- clamp graphs
Stephen Larson
@slarson
Jun 10 2015 12:00
its a good point
we might as well
2A & 2B
Travis Jacobs
@travs
Jun 10 2015 12:00
Ok
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:00
more info, more accurate model
I've considered the ability to merge info from IClamp, VClamp and IV plots
Stephen Larson
@slarson
Jun 10 2015 12:01
merge in what way?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:02
First of all the simulation is going to be done using one or both of the patch clamp experiments
Results would be compared with digitized figures
Stephen Larson
@slarson
Jun 10 2015 12:03
the fitting you mean?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:03
yes
If we had patch clamp experiments, results are much more accurate
Stephen Larson
@slarson
Jun 10 2015 12:04
no question
would be great
shall we continue?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:05
Also we should look for some parameters for the patch clamp experiment
Travis Jacobs
@travs
Jun 10 2015 12:05
Sure, but I don't know what we're getting any more
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:05
such as the duration, steps, etc
Travis Jacobs
@travs
Jun 10 2015 12:06
IV curves, I- clamp, V- clamp graphs
ok and parameters
and the parameters are going in the spreadsheet as well?
Stephen Larson
@slarson
Jun 10 2015 12:07
if we can; not a bad idea
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:07
duration, intervals, holding potentials of current stimuli (from, to, and steps)
  • or
I mean or current stimuli
Travis Jacobs
@travs
Jun 10 2015 12:08
Yes
So should we just make a note that those parameters are in the paper? I can see them in the description of the fawcett et al. fig
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:09
Yes, there are usually in the description of the figures
Travis Jacobs
@travs
Jun 10 2015 12:10
Alright
I'll make a new column for the parameters then
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:12
Also here we have G/V plots in addition to I/V in the same graph
this is useful too, I'll add this feature to the fitter soon
Travis Jacobs
@travs
Jun 10 2015 12:13
Yep
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:14
and one thing is that the Y axis is normalized for both
Travis Jacobs
@travs
Jun 10 2015 12:14
Right
I think that maybe we should just defer the reader to the description if we want to include everything
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:14
it means we have to consider some reconstruction in order to compare our results with this one
Yes, we have considered this during designing the interface
but we need more details and options
Travis Jacobs
@travs
Jun 10 2015 12:16
Yeah, I think noting the figures here is enough though, since we can refer back to the figs and their descriptions
When the time comes we can add detail, right now we want to know which figures contain the info we want to use
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:16
Ok
Travis Jacobs
@travs
Jun 10 2015 12:16
So I'll get rid of the parameters column and continue the way we were.
Your point makes sense though, we need more info, so we should raise an issue for that
to go through our list of figures and flesh out each one with the details of the experiment
parameters, normalization, etc.
How does this sound @VahidGh ?
and @slarson
:)
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:18
Sounds good
Travis Jacobs
@travs
Jun 10 2015 12:18
Alright, I'll continue with Liu et al
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:18
Also we can do the data entry, when the web app is ready
Travis Jacobs
@travs
Jun 10 2015 12:19
Absolutely
Ok, Liu et al. fig 1E
what are we looking at?
patch clamp IV correct?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:22
yes, but specifically for every ion channel
Travis Jacobs
@travs
Jun 10 2015 12:22
Right
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:22
here I think figures are for the whole cell patch clamp, including all the participating channels
Travis Jacobs
@travs
Jun 10 2015 12:22
Whole cell, I'll note that on the sheet
That's correct
So I'll note that and we can come back to the details at a later date
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:23
we can use these kind of figures to validate integration of the channels in one cell
Stephen Larson
@slarson
Jun 10 2015 12:23
that’s actually useful for the unit tests off the muscle model
Travis Jacobs
@travs
Jun 10 2015 12:23
That is true
This is a body wall muscle cell
Stephen Larson
@slarson
Jun 10 2015 12:24
in addition to the stuff in Boyle & Cohen
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:24
It would be great if we also mention which channels are participating
@slarson, right
but fig 7 would be useful
Stephen Larson
@slarson
Jun 10 2015 12:25
yes
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:26
The red and blue lines are useful
Travis Jacobs
@travs
Jun 10 2015 12:27
And black again for an integrated model I suppose
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:27
yes
one thing, in the Boyle&Cohen model, we had shk-1 and shl-1 as K channels
if I'm right!
or the slow one is slo-2 maybe!
Stephen Larson
@slarson
Jun 10 2015 12:29
right; they are more abstract channels if i recall?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:30
They've considered two K channels, slow and fast
but they did not mention the genetic aspect
maybe we can refer this as paper as the genetic basis of the model
Travis Jacobs
@travs
Jun 10 2015 12:31
Ok, where should we note that?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:31
so the k slow would be slo-2 and k fast is shk-1
Travis Jacobs
@travs
Jun 10 2015 12:31
Vahid, can you make a comment for that on the cell in the sheet?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:31
Yes, we should find somewhere in the app for this purpose
I'm not able to edit!
Travis Jacobs
@travs
Jun 10 2015 12:33
Oh no!
One sec
Try now?
It should be world editable, otherwise I can give your account permissions specifically
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:34
Yes, I did
Travis Jacobs
@travs
Jun 10 2015 12:34
Ok perfect
So it looks like there are no more figures for us on this paper
Agreed?
If so I'll move onto this one
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:35
yes
Travis Jacobs
@travs
Jun 10 2015 12:35
Ok
So Liu 2014
Stephen Larson
@slarson
Jun 10 2015 12:35
well unless the current traces...
Travis Jacobs
@travs
Jun 10 2015 12:35
Ah, right, so let's keep noting those as well
Just 7A?
Stephen Larson
@slarson
Jun 10 2015 12:37
yes i think so
Travis Jacobs
@travs
Jun 10 2015 12:37
Ok
Stephen Larson
@slarson
Jun 10 2015 12:37
great; onward
Travis Jacobs
@travs
Jun 10 2015 12:37
And in 1E as well; I'll note it
ok, yes so Liu 2014
1B
Whoa
Stephen Larson
@slarson
Jun 10 2015 12:38
hehe
Travis Jacobs
@travs
Jun 10 2015 12:39
So we've got several whole-cell current traces
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:39
This paper is one with high priority
based on 3 different motor neurons
Travis Jacobs
@travs
Jun 10 2015 12:40
Right
and 4 mutant types
this is juicy
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:41
One thing also should be considered is the cell type
we should add this if available
Stephen Larson
@slarson
Jun 10 2015 12:41
hmm point
Travis Jacobs
@travs
Jun 10 2015 12:41
Ok
What about when entering this into the app
Should we tag the data with cell type now, or can we do it later along with the other stuff (params, etc.)
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:42
yes, we should add this to the form
Travis Jacobs
@travs
Jun 10 2015 12:45
Ok, so can I leave that to you Vahid?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:45
fig. 5 is also useful
Yep, we have an issue open for this if you remember
I'll refer to that when completing the experiment form design
Travis Jacobs
@travs
Jun 10 2015 12:47
Ok sounds good
For now we can move to fig 5 as you say
I'll note the ion channels involved in fig1 and the plots that are there
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:49
We should close #7 by this implementation (both in data layer and UI)
Travis Jacobs
@travs
Jun 10 2015 12:51
Ok
Right, I remember that issue
So for figure 5 we have current traces
So voltage clamp, awesome. I'll note that on the sheet
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:53
G and I/V curves
Travis Jacobs
@travs
Jun 10 2015 12:54
Right,on bottom
and then the next graphs are also useful
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:56
Yes, and that's need some extra work to be handled
Travis Jacobs
@travs
Jun 10 2015 12:56
Is this paper in the Ca++ section as well?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:56
because of Cl
Travis Jacobs
@travs
Jun 10 2015 12:57
So figure 6, what is going on here?
Ca++ and Cl- concentrations
Vahid Ghayoomie
@VahidGh
Jun 10 2015 12:57
needs more research
I haven't consider the [Cl] in my calculations, and haven't seen before
Travis Jacobs
@travs
Jun 10 2015 12:58
Ok, I will make a note of this in the sheet
Let's finish up this paper and that will be it for this meeting
So you're saying figure 6 requires some more research because of the Cl- concentrations involved
What about figure 7 @VahidGh
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:00
Ok
We should compare with other figures, but seems to be useful
but not for fig. 8
Travis Jacobs
@travs
Jun 10 2015 13:03
ok, but is this paper mentioned in the Ca++ folder?
Because it seems like fig8 may be useful for that
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:05
I don't think so
It seems it means these ca channel are not being expressed, in these cell types
so mutations have no effect
Travis Jacobs
@travs
Jun 10 2015 13:08
Ah I see
So this would be useful for a whole cell integrated model of these three muscle cells?
But I suppose we have that data in the WT lines in the other plots as well
So this figure is not useful to us
Agree?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:09
I think we have this in previous figures
Travis Jacobs
@travs
Jun 10 2015 13:09
(I think that's what you said at first anyway Vahid :) )
Yes exactly
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:10
Yep
Travis Jacobs
@travs
Jun 10 2015 13:10
Ok cool
so we're done with this paper for now
I think we made good progress here today
We know that there is a lot more data that will come from these plots and their captions, but we can extract that when we're ready
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:11
yeah, maybe we can go deeper next time with the UI
Travis Jacobs
@travs
Jun 10 2015 13:12
What do you mean @VahidGh ?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:14
I mean having the web app up, we can do this with in detail
Travis Jacobs
@travs
Jun 10 2015 13:14
Oh, yeah totally!
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:14
investigation
Travis Jacobs
@travs
Jun 10 2015 13:14
So you said Milad will be putting that up this evening?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:15
Yep
Travis Jacobs
@travs
Jun 10 2015 13:15
Great!
So yeah, we can leave this growing list of figures for now, and focus on an investigation next time
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:16
Right
Travis Jacobs
@travs
Jun 10 2015 13:19
Ok, sounds like a plan!
I wanted to chat about your comment here Vahid, if you have time
Since this is integral to the fitting process, and validation more generally
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:21
Yeah, of course
Travis Jacobs
@travs
Jun 10 2015 13:22
Ok, so you're trying to minimize the sum of squared distances between the points of the two models
As the cost function
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:23
Actually the simulated model and digitize data for my purposes
Travis Jacobs
@travs
Jun 10 2015 13:24
Yes sorry, the other "model" I was talking about was the actual data, my mistake
I think we can use this in sciunit as well
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:24
But if I normalize and store the results we can use for comparing two models
Travis Jacobs
@travs
Jun 10 2015 13:24
And I think we should be using the same statistics in both sciunit and the optmization
Do you think we should implement the cost function and sciunit comparisons in the same place?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:26
Yes we can, but we don't need to implement it twice
Travis Jacobs
@travs
Jun 10 2015 13:26
Right, we can just implement it somwhere that both sciunit and neurotune and see it
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:27
We just need to agree on some measurement units
Travis Jacobs
@travs
Jun 10 2015 13:27
Yes
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:27
and then work based on that unit
Travis Jacobs
@travs
Jun 10 2015 13:28
Right, as long as both environments are expecting the same type of data we should be OK
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:28
for example, we should first define "how far" two model are from each other
Travis Jacobs
@travs
Jun 10 2015 13:29
Right
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:29
Then we can store this parameter while fitting each model and use it then
Travis Jacobs
@travs
Jun 10 2015 13:29
So the comparison will ultimately give us some number that indicates this distance
say 0 < n < 1 ?
1 is identical
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:30
yes, for example
Travis Jacobs
@travs
Jun 10 2015 13:30
or even -1 < n < 1
Ok, so it's clear that we need to find or modify some statistic that will give us this number
Is there an issue for this Vahid?
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:32
I don't know what is the approach and statistics for other measurements in SciUnit
Travis Jacobs
@travs
Jun 10 2015 13:32
They are not implemented yet, so we are starting from zero basically
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:32
If we knew this, we could follow the same approach
So it is up to us to implement the statistic
for comparing two curves in this way
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:34
So we need to define the reference in the first place
Travis Jacobs
@travs
Jun 10 2015 13:34
Yes
So we don't have an issue for this yet, I will create one and link it
But in any case sum of squared differences is a good enough start
I think we just need to agree on a place to implement this that sciunit and neurotune can both make use of, and not duplicate the cost and evaluation functions
Since we agree that they will be the same, whether sum of sq diffs or something else
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:39
yes, in addition to some supplements in order to cover the inactivation problem with I/V curves
We also need some weighting mechanism to improve the score for more accurate models
Travis Jacobs
@travs
Jun 10 2015 13:41
Ok, is there an issue for this?
If not, we should create it!
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:42
For example a model evaluated against two plots is much more accurate than the one with one plot
Travis Jacobs
@travs
Jun 10 2015 13:42
Oh I see what you mean
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:44
Lots of this should be handled by our app, because this app is tracking different modelings from different experiments/users
So we need to store the final cost for each model, by some measurements
SciUnit could be used to globally, to compare it with other models from here or other sources
Travis Jacobs
@travs
Jun 10 2015 13:47
Ok, so that's decided! SciUnit will be used for the cost function as well
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:48
Better saying for comparing different models
Travis Jacobs
@travs
Jun 10 2015 13:49
Ok, and what about for models vs observed?
Regarding what the cost function will be, I have raised #52
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:49
This is what I've implemented right now
We can use the SciUnit for normalizing or weighting different models as I said
Vahid Ghayoomie
@VahidGh
Jun 10 2015 13:56
And also comparing our models with models not generated by ourselves
Another thing could be done by Sciunit, is that if the generated model is "valid"
By using the final model in e.g. other projects and evaluating the results!
Vahid Ghayoomie
@VahidGh
Jun 10 2015 14:08
@travs, Would you please tell me if we can spend some time to working on some data model issues
By launching the app, we need to close some of those issues, but I'm not sure which one is going to be handled by pyOW
For example we have an Experiment class here, and one in the PyOW you've implemented.
It would be great if we could finish the data modeling issues first
Vahid Ghayoomie
@VahidGh
Jun 10 2015 14:14
Is it possible to develop required APIs for ion channels here, and then migrate to the PyOW when it is ready to use?
Travis Jacobs
@travs
Jun 10 2015 17:25
@vahidgh Absolutely
@VahidGh and if you'd like to spend some time on the data modelling issues in channelworm, we should set up a meeting. Would you like me to send you a poll so we can arrange that?
I suppose when Milad launches the app tonight we will have somewhere to stage these changes
Vahid Ghayoomie
@VahidGh
Jun 10 2015 18:39
@travs, yes, good idea.
It would be much better if @miladjafary could also join us :D