These are chat archives for openworm/ChannelWorm

26th
Jun 2015
Travis Jacobs
@travs
Jun 26 2015 14:30 UTC
@VahidGh @a-palyanov hey guys! How's it going?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:33 UTC
Hi @travs
a-palyanov
@a-palyanov
Jun 26 2015 14:35 UTC
Hi guys!
Travis Jacobs
@travs
Jun 26 2015 14:36 UTC
Hey! Welcome
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:36 UTC
Hi @a-palyanov
Travis Jacobs
@travs
Jun 26 2015 14:36 UTC
How have your guys' weeks been?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:37 UTC
A good one, thanks :)
a-palyanov
@a-palyanov
Jun 26 2015 14:38 UTC
Quite well, thanks!
Travis Jacobs
@travs
Jun 26 2015 14:38 UTC
That's good! :D
I guess the meeting description kind of described why we're here, but to refresh our memory it
it's to do with this milestone
it's to do with this milestone
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:40 UTC
Yep
Travis Jacobs
@travs
Jun 26 2015 14:40 UTC
We are getting along pretty well with modelling ion channels that have patch clamp data, but without patch clamp data is a separate problem, right Vahid? :)
And Andrey, you expressed interest in helping with this milestone, so I think we can use your help in making some issues here today, on what next steps we need for this
a-palyanov
@a-palyanov
Jun 26 2015 14:42 UTC
Sure
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:42 UTC
Ok, I'm going to explain some previously on modeling with no patch clamp data
Travis Jacobs
@travs
Jun 26 2015 14:42 UTC
@VahidGh Please do!
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:42 UTC
As this is the subject of my thesis project
a-palyanov
@a-palyanov
Jun 26 2015 14:42 UTC
You've mentioned ion channels that have patch clamp data. Was it about C. elegans ion channels?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:43 UTC
Doing some structure-function relationship analysis using machine learning algorithms
Actually in the first approach we are modeling C. elegans ion channels for which patch clamp data is available
Travis Jacobs
@travs
Jun 26 2015 14:44 UTC
@a-palyanov Yes, we're tracking those in this milestone
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:45 UTC
But in this approach we need to estimate models (parameters) that we don't have patch clamp data for ion channels in C. elegans
Travis Jacobs
@travs
Jun 26 2015 14:45 UTC
Right, because many ion channels of this organism do not have patch clamp data
available
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:46 UTC
But we have models of homolog channels
a-palyanov
@a-palyanov
Jun 26 2015 14:46 UTC
That's reasonable
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:46 UTC
Ok
a-palyanov
@a-palyanov
Jun 26 2015 14:47 UTC
By the way, is there any data about how ion channels of various types are distributed over the surface of a typical C. elegans neuron?
In this spreadsheet we have some data regarding this issue
a-palyanov
@a-palyanov
Jun 26 2015 14:49 UTC
Thanks, @VahidGh!
Travis Jacobs
@travs
Jun 26 2015 14:50 UTC
@VahidGh Are the homolog models you mentioned available someplace already?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:51 UTC
The types in the 2nd and 3rd sheet describing some of these classifications
but let me open it a little more
Travis Jacobs
@travs
Jun 26 2015 14:51 UTC
Ok
a-palyanov
@a-palyanov
Jun 26 2015 14:53 UTC
I've read somewhere that typical C. elegans neuron has only about 50-200 ion channels distributed over the whole nervous cell. If there are so many types of channels, does it mean that each type of channel in each cell has a very small count?
Or it means that each cell has not so much different types of channes
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:55 UTC
@a-palyanov, yeah, that's true
a-palyanov
@a-palyanov
Jun 26 2015 14:55 UTC
This estimate is quite reasonable taking into accont the surface of a typical neuron (with axon / dendrites) and the size of a typical ion channel. Because cell membrane also contains other built-into-membrane proteins, as well as ion pumps / exchangers.
Vahid Ghayoomie
@VahidGh
Jun 26 2015 14:56 UTC
I'm trying to paste some images, the channel is stuck :D
@a-palyanov, right, although here the approach is dealing with single channel models
a-palyanov
@a-palyanov
Jun 26 2015 14:58 UTC
Ok
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:00 UTC
Let me first link to some useful papers
Travis Jacobs
@travs
Jun 26 2015 15:00 UTC
@VahidGh Ok :)
This is a good one, that the authors describe a method for simulating single channel kinetics using structural studies
Travis Jacobs
@travs
Jun 26 2015 15:01 UTC
@VahidGh This seems awesome
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:02 UTC
The problem with this approach is the hard task of dealing with MD
I'm going to take the other methods in this study and combining this with other approaches
This is an overview of predicting channel types based on protein sequences using machine learning
a-palyanov
@a-palyanov
Jun 26 2015 15:05 UTC
By the way, I know a guy here in Novosibirsk who is a specialist in MD, I can talk to him if it will be necessary
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:05 UTC
And this is some practical work considering functional domains
Actually I did this before in our university
And also some email contacts with others doing this
All them agree that this is a really time-consuming work
Travis Jacobs
@travs
Jun 26 2015 15:07 UTC
Thank you @a-palyanov
@VahidGh what would the pipeline look like?
We find a homolog which is sequenced, then?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:09 UTC
This approach is good for simulation with high-accuracy purposes
For example the first paper I linked are doing this to simulate channel activity in heart
This lab, trying to simulate heart pumping etc
Ok, but what I'm envisioning is trying to take the second method
(predicting channel types from its protein sequence)
And applying this to channel kinetics
Travis Jacobs
@travs
Jun 26 2015 15:14 UTC
Ok, so predicting kinetics from protein sequence?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:14 UTC
I've had some talk with Prof. Nowzari who is in our university and a co-author of this paper
He was optimistic for applying the same method for predicting kinetic parameters
Ok
let me talk more about what we have now, and where we are going
Travis Jacobs
@travs
Jun 26 2015 15:16 UTC
@VahidGh Ok :D
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:17 UTC
hmm, gitter is not uploading an image I want
I guess the problem is with it's size
Travis Jacobs
@travs
Jun 26 2015 15:18 UTC
Try this
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:20 UTC
The same error!!
BTW, I was going to paste the image of structural categorization of ion channels
Travis Jacobs
@travs
Jun 26 2015 15:21 UTC
Ahh, ok
a-palyanov
@a-palyanov
Jun 26 2015 15:21 UTC
ok
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:21 UTC
And in this link we have this mapping to channels in C. elegans
Travis Jacobs
@travs
Jun 26 2015 15:22 UTC
Yes
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:22 UTC
So with those sequence to structure prediction studies
We can do this prediction
Now we have a data-set from ChannelPedia
with ~100 channel models
from mammalian
for example we have kv1 in C. elegans
and Kv1.1-8 ! in mammalians
and we have model for most of them
And we have information about protein sequences for those channels
I've extracted some data about which region of the protein sequence is related to which property of the channel
In addition we can use an unsupervised algorithm for this step
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:28 UTC
And have a better prediction for relating regions of protein sequences to those domains
Travis Jacobs
@travs
Jun 26 2015 15:28 UTC
Ok, but where is the starting point? With a channel model from channelpedia?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:28 UTC
In addition to it's protein sequence
Which is easily accessible through PDB
Travis Jacobs
@travs
Jun 26 2015 15:31 UTC
Ok, so what about next steps? What do we actually need to do now?
Say we have a channel model/sequence
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:32 UTC
Most of the steps discussed in #25
I've done some good works with a knn algorithm
Travis Jacobs
@travs
Jun 26 2015 15:32 UTC
Ok, let's make some issues for those steps now
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:33 UTC
as a case-study
but need some improvements
Travis Jacobs
@travs
Jun 26 2015 15:33 UTC
What is meant by input types in #25 (1)?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:34 UTC
in the next step I have to combine the first results with an algorithm such as SVM to predict needed parameters for the HH model
I've started with conductance, V_half activation, etc
Those input types are channel sequences need to feed the algorithm
Which is protein sequences
Travis Jacobs
@travs
Jun 26 2015 15:36 UTC
Ok
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:37 UTC
And in FASTA format
Travis Jacobs
@travs
Jun 26 2015 15:38 UTC
Ok, and number 2 in that list
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:38 UTC
We need some scripts to retrieve those from PDB for channels in C. elegans and those in chnnelpedia with available models
I'll make some issues for each of them separately
I'm still doing some research for each of them
Still comparing some methods
To make sure about the overall work
Travis Jacobs
@travs
Jun 26 2015 15:42 UTC
Alright, but I'm still not exactly clear on the pipeline. We start with the mammalian channel and sequence, then what?
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:42 UTC
This depends heavily on those functional domains
that relate to the structural domains
I'm working on this part right now for a case study
when this is done we can have an algorithm with those predictors
Travis Jacobs
@travs
Jun 26 2015 15:46 UTC
Hmm, I'm not sure how this relates to the mammalian channel/sequence though
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:46 UTC
And then applying an algorithm such as SVM to predict parameters from HH model with the given protein sequence
The mammalian channels are being used for training and cross-validation purposes
When the predictor built using these data then we can use it to predict any kind of channel kinetic
The main task is to consider which part of the sequence for prediction of related parameter
Travis Jacobs
@travs
Jun 26 2015 15:50 UTC
Ok cool
Well @VahidGh if you can raise a few issues under that milestone that relate to next steps that would be super helpful
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:50 UTC
e.g. which region is related the channels pore that should be considered as the predictor for conductance estimation
Travis Jacobs
@travs
Jun 26 2015 15:50 UTC
But I have to go right now
@a-palyanov Is it becoming a little clear what we are trying to do here?
:)
a-palyanov
@a-palyanov
Jun 26 2015 15:51 UTC
Yes, thanks :)
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:51 UTC
@travs, Ok, we need to this together with more chats :D
@a-palyanov, sorry I talked too much
Travis Jacobs
@travs
Jun 26 2015 15:51 UTC
@VahidGh Yes totally :)
@VahidGh @a-palyanov Talk to you later guys!
a-palyanov
@a-palyanov
Jun 26 2015 15:53 UTC
@VahidGh, that's ok, this was a purpose of the meeting - to discuss what we are going to do :)
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:53 UTC
@a-palyanov, @travs, I don't know if you get exactly what I'm saying
@a-palyanov, What's your idea?
a-palyanov
@a-palyanov
Jun 26 2015 15:57 UTC
According to what I know about ion channels, if we have a sequence of unknown channel and its close homolog of known channel with PDB structure, than we can try to replace amino acids in PDB to our sequence of interest and then tune the new structure with MD and thus get a new model of the channel
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:59 UTC
Yes, this was the first approach I was going to work on
a-palyanov
@a-palyanov
Jun 26 2015 15:59 UTC
I need to read about that SVM algorithm to get a view about how it predicts the channel properties and how accurate this prediction can be
Vahid Ghayoomie
@VahidGh
Jun 26 2015 15:59 UTC
But we have two problems for this approach
  1. we don't have info for lots of channels assuming the same method in http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2888060/
  1. MD have good results for nano time-scale activities
here we have some kinetics with micro
a-palyanov
@a-palyanov
Jun 26 2015 16:02 UTC
I agree
Vahid Ghayoomie
@VahidGh
Jun 26 2015 16:02 UTC
and that's why the authors replaced MD with electrostatic energy calculations
And this also needs lots of work for each channel
@a-palyanov, great, any kind of help would be appreciated
Thanks for your attention
a-palyanov
@a-palyanov
Jun 26 2015 16:06 UTC
I was glad to participate and discuss all these today
That's really and interesting subject for research
Vahid Ghayoomie
@VahidGh
Jun 26 2015 16:07 UTC
@a-palyanov, yeah, that's really interesting for me too :D