Figure: haplotype structure about ag1000g-phase1-vgsc-report

[alimanfoo]

@cclarkson here's some food for discussion tomorrow:

...this is a clustering of haplotypes carrying L995S. Only distinct haplotypes are used to construct the tree. Haplotype frequencies are shown in the adjacent bar plot. Shaded regions show the 5 largest clusters by a method I'll explain tomorrow - I'm thinking we show one network for each of these 5 clusters.

[alimanfoo]

Hey @cclarkson, maybe Python is an option after all, I made this with graphviz:

cc @hardingnj.

[cclarkson]

That's great, will be cool if we could keep most of it in python. Looking forward to seeing the code.

[alimanfoo]

Making some progress, here's a network with some real data, the Kenya/Uganda L995S cluster:

Will turn this into a usable function and share asap.

[alimanfoo]

Still not sure how to figure out which edge has which mutation, that may be tricky...

[cclarkson]

Hi @alimanfoo ,

Just to document the TCS comparison, almost all clusters produced identical networks to your function, except for 995F8, which TCS did not circularise and 995F16, which added a number of loops compared to the one you generated - see below (I tried to lay it out the same to compare but TCS kept crashing, this is best I managed, the bottom right hand corner is a bit messy!).

[alimanfoo]

Thanks Chris. Can you compare with the networks I made using median-joining? They should be in the (recently updated) hapclust_demo notebook, look out for the calls to graph_haplotype_network(..., network_method='mjn'). What you have above in TCS looks very similar to what I made with MJN...

[cclarkson]

@alimanfoo

Yep, I'm pretty sure that is the same as the TCS one, interesting.

ION - having a problem getting graphviz to work due to neato - how did you fix that problem?

[cclarkson]

No worries, think I've cracked it.

[cclarkson]

Hi @alimanfoo ,

Now I've got this all up and running again, I was just trying to plot multiple networks on one figure but I'm struggling. I can't see how to get it to use an 'ax', is this because it is just a wrapper for external code and can't talk to matplotlib?

[alimanfoo]

Hi Chris, sorry yes, I haven't figured out the best way to get the networks into a mpl figure. I'll have a play around...

…

On Thu, Mar 9, 2017 at 4:19 PM, Chris Clarkson ***@***.***> wrote: Hi @alimanfoo <https://github.com/alimanfoo> , Now I've got this all up and running again, I was just trying to plot multiple networks on one figure but I'm struggling. I can't see how to get it to use an 'ax', is this because it is just a wrapper for external code and can't talk to matplotlib? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#25 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAq8QjyTXWPUdlCPk95wE9ytk49Yuzurks5rkCZ_gaJpZM4MQYub> .

-- Alistair Miles Head of Epidemiological Informatics Centre for Genomics and Global Health <http://cggh.org> The Wellcome Trust Centre for Human Genetics Roosevelt Drive Oxford OX3 7BN United Kingdom Email: [email protected] Web: http://purl.org/net/aliman Twitter: https://twitter.com/alimanfoo Tel: [phone]

[cclarkson]

Hi @alimanfoo ,

I'm trying to find a nice way to bring the dendrogram together with the clusters but I'm struggling to work out how to give the fig_haplotypes_clustered function (from haplclust_utils) an ax to plot on in my composite figure because the dendrogram function is already a composite figure. Any pointers would be gratefully received!

Cheers, C.

[alimanfoo]

Hi Chris, I would suggest building each panel of the figure as a separate matplotlib figure, saving out to jpeg or png, then composing the whole figure in libreoffice draw, at least for now.

…

On Tuesday, April 11, 2017, Chris Clarkson ***@***.***> wrote: Hi @alimanfoo <https://github.com/alimanfoo> , I'm trying to find a nice way to bring the dendrogram together with the clusters but I'm struggling to work out how to give the fig_haplotypes_clustered function (from haplclust_utils) an ax to plot on in my composite figure because the dendrogram function is already a composite figure. Any pointers would be gratefully received! Cheers, C. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#25 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAq8QgYwfaX7Z2JqA_q4dTDmXNZEVZTQks5ru1qIgaJpZM4MQYub> .

-- Alistair Miles Head of Epidemiological Informatics Centre for Genomics and Global Health <http://cggh.org> The Wellcome Trust Centre for Human Genetics Roosevelt Drive Oxford OX3 7BN United Kingdom Email: [email protected] Web: http://purl.org/net/aliman Twitter: https://twitter.com/alimanfoo Tel: [phone]

[cclarkson]

Been flailing at this for a while and I think I am finally getting somewhere! I'm now trying to work out how to colour the network's backgrounds like the dendrogram (and align/scale things in a less awful way).

[alimanfoo]

Cool. If you think it isn't going to work to fit the tree and networks in the same figure, please feel free to rethink.

…

On Tue, Apr 11, 2017 at 6:13 PM, Chris Clarkson ***@***.***> wrote: [image: test_fig] <https://cloud.githubusercontent.com/assets/13287471/24921332/57756eda-1ee2-11e7-8c2b-fa4a4bc6662b.jpg> Been flailing at this for a while and I think I am finally getting somewhere! I'm now trying to work out how to colour the network's backgrounds like the dendrogram (and align/scale things in a less awful way). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#25 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAq8QrAij8TQYoxCQAkYeHYATY28HGqYks5ru7TRgaJpZM4MQYub> .

-- Alistair Miles Head of Epidemiological Informatics Centre for Genomics and Global Health <http://cggh.org> The Wellcome Trust Centre for Human Genetics Roosevelt Drive Oxford OX3 7BN United Kingdom Email: [email protected] Web: http://purl.org/net/aliman Twitter: https://twitter.com/alimanfoo Tel: [phone]

[cclarkson]

As far as I can see, the importing of the graphviz image as an image array means that you can't overlay or 'draw' on top of those arrays. I'm going to try an kick out the graphviz stuff in a vector format so that I can build something a bit like this:

I think this figure could then be quite small, losing the fine network details, but we could have individual figures for the clusters of interest. What do you think?

[alimanfoo]

Hi Chris, You should be able to plot over the top of an image if you want to. I just tried it, did imshow() then plot() to plot a line over the top, seemed to work. What did you want to overlay on the image? Re the networks being small, that may be OK for some, I guess the key thing is to be able to see the topology, the colours of each node (including singletons), and labels for any non-synonymous edges. I think F1 will need it's own figure though. But do some playing around and see what works.

…

On Wednesday, April 12, 2017, Chris Clarkson ***@***.***> wrote: As far as I can see, the importing of the graphviz image as an image array means that you can't overlay or 'draw' on top of those arrays. I'm going to try an kick out the graphviz stuff in a vector format so that I can build something a bit like this: [image: inkscape_example_vgsc_netclustfig] <https://cloud.githubusercontent.com/assets/13287471/24960804/d779e7d0-1f8e-11e7-86c9-6f0593736589.png> I think this figure could then be quite small, losing the fine network details, but we could have individual figures for the clusters of interest. What do you think? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#25 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAq8Qu9qpuiIwQOMZmaJZCCEvMcfFxrVks5rvNargaJpZM4MQYub> .

-- Alistair Miles Head of Epidemiological Informatics Centre for Genomics and Global Health <http://cggh.org> The Wellcome Trust Centre for Human Genetics Roosevelt Drive Oxford OX3 7BN United Kingdom Email: [email protected] Web: http://purl.org/net/aliman Twitter: https://twitter.com/alimanfoo Tel: [phone]

[cclarkson]

Hi @alimanfoo,

This was what I was trying to do with just code, but I resorted to a bit of Inkscape in the end. Looks okay, but we would need to make it clear that the network nodes are not to scale across networks, they are just relative to that cluster. Just going to make the stand alone for cluster 16.

-Should I push on with this format for 995S?
-What do you think about leaving the colony haplotypes in this analysis?
-At some stage I saw the network function able to add the size of node (number of haps) to the figure, I'm trying to find this to make a size key, is it still there?

[cclarkson]

Hi @alimanfoo

Not sure how I feel about the overlapping nodes, but they can all be re-arranged nicely in Inkscape if we decide to use this type of figure...

[cclarkson]

Figure 2 (draft) - haplogroup distance and frequency

[alimanfoo]

Looking good.

…

On Thursday, May 4, 2017, Chris Clarkson ***@***.***> wrote: Figure 2 (draft) - haplogroup distance and frequency [image: figure2_dendfreq_draft] <https://cloud.githubusercontent.com/assets/13287471/25711347/9ba8e1c0-30e6-11e7-8f74-e645bdb33155.jpg> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#25 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAq8QgxRsZeUea786MIXWs8WqzcMU1d0ks5r2e6WgaJpZM4MQYub> .

-- Alistair Miles Head of Epidemiological Informatics Centre for Genomics and Global Health <http://cggh.org> Big Data Institute Building Old Road Campus Roosevelt Drive Oxford OX3 7LF United Kingdom Email: [email protected] Web: http://a <http://purl.org/net/aliman>limanfoo.github.io/ Twitter: https://twitter.com/alimanfoo

[alimanfoo]

Superseded by #69 and #73.