mixMC framework

[Kim-Anh Le Cao]

As discussed, functions to add or improve (pull from DEVEL)

Framework is:

filter -> TSS -> CLR/ILR -> multilevel -> PCA or sPLSDA

• mixMC.filter

• mixMC.TSS

• pca(logratio = c('CLR', 'ILR')

• splsda(logratio = c('CLR')

• multilevel(logratio = c('CLR', 'ILR', method = c('splsda', 'spls', 'pca'))

• cladogram.input()

• add most diverse data as data.raw, indiv and taxonomy (only those 3!)

When you update those files, please comment on this issue point by point.

Note: make sure splsda can only have CLR transformation!

[fbartolo]

add new option log-ratio in pca

[fbartolo]

add new option log-ratio in splsda

[fbartolo]

add option logratio and add pca to multilevel

[Kim-Anh Le Cao]

NOTE: disregard my comment about V = NULL. We still need it to be called from multilevel so OK. You need to merge ilr.transfo and the V calculation in a single function. It will avoid copy paste in PCA and PCA multilevel. See next ocmment.

Hi Francois, Thanks for that. For any choices, e.g. Logratio in your pca function, I would prefer something like:

logratio = c('none', 'ILR')

In the argument call of the function, instead of:

pca <-

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT25

+function(X,

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT26

• ncomp = 2,

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT27

• center = TRUE,

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT28

• scale = FALSE,

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT29

• max.iter = 500,

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT30

• tol = 1e-09,

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT31

• logratio = 'none',

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/pca.RT32

• V=NULL)

It just makes things more readable when we write the help file.

Also, in PCA, the V should not be an input but should be calculated directly inside the pca function when logratio = 'ILR', as you then did.

Make sure once you have included V inside that you test thoroughly to obtain the same results as before!

Regards,

Kim-Anh

Too brief? Apologies, I need to restrain my typing due to a computer elbow injury.

--

Dr. Kim-Anh Lê Cao

NHMRC Career Development Fellow

ResTeach Affiliate (School of Chemistry and Molecular Biosciences)

The University of Queensland Diamantina Institute

Translational Research Institute

Level 5, 37 Kent Street | Woolloongabba QLD 4102

T: [phone]| F: [phone]

http://www.tri.edu.au/staff/kim-anh-le-cao

mixOmics: http://mixomics.org/

[Kim-Anh Le Cao]

For multilevel: you cannot do a logratio ILR with sPLSDA, I propose you reset it to CLR instead (not 'none')

• if(logratio == 'ILR')

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT182

• {

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT183

• warning("reset 'ILR' to 'CLR' for sPLSDA.",

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT184

• call. = FALSE)

#KA: and so, where is your reset?

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT185

• }

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT186

• else if (logratio == 'CLR')

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT187

• {

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT188

• X.clr=clr.transfo(X)

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT189

• rownames(X.clr) = rownames(X)

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT190

• X=X.clr

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT191

I think you can include the clr.backtransfo function directly inside the ilr.transfo. That will avoid you to repeat in all functions (PCA and multilevel PCA).

And now I understand why the V is necessary inside the PCA function :) OK!

V = clr.backtransfo(X)

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT241

• dim(V)

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT242

• 2 - PCA on ILR data

https://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#add-commenthttps://bitbucket.org/klecao/package-mixomics/pull-requests/124/mixmc/diff#LmixOmics/R/multilevel.RT243

• res = pca(X = X.ilr, ncomp = ncomp, logratio = 'ILR', V = V)

KA: note: your loratio is hardcoded here.

Regards,

Kim-Anh

Too brief? Apologies, I need to restrain my typing due to a computer elbow injury.

--

Dr. Kim-Anh Lê Cao

NHMRC Career Development Fellow

ResTeach Affiliate (School of Chemistry and Molecular Biosciences)

The University of Queensland Diamantina Institute

Translational Research Institute

Level 5, 37 Kent Street | Woolloongabba QLD 4102

T: [phone]| F: [phone]

http://www.tri.edu.au/staff/kim-anh-le-cao

mixOmics: http://mixomics.org/

[fbartolo]

Hi Kim-Anh,

To splsda reset, in fact, it's a 'none' reset also I don't need to add other thing, just a warning. I fixed pca.

[Kim-Anh Le Cao]

I'd like the reset to be set to CLR instead of 'none'. It fits better with the mixMC philosophy (compositional data) Thanks

[fbartolo]

change reset. I add 2 new functions: filter and TSS.normalisation

[Kim-Anh Le Cao]

Better even: mixMC.filter mixMC.normalisation

[Kim-Anh Le Cao]

Can you call it normalisation.TSS? It might sound trivial but it is a matter of alphabetical order and how a user can go through the names of a thousands of functions in mixOmics :) Thanks!