Question: converting mutual information matrix to gene regulatory network
0
gravatar for F
2.1 years ago by
F3.4k
Iran
F3.4k wrote:

hi,

I made a mutual information matrix of 192 miRNAs and mRNAs derived by TCGA data by parmigene R package like below

> head(MI[,1:4])

      hsa-let-7a-1 hsa-let-7a-2 hsa-let-7a-3    hsa-let-7b
ACTB   0.012524427  0.026290733  0.012733996 -0.0006921249
ACTG1 -0.019429116 -0.010771804 -0.002583545  0.0072491626
ACTN4 -0.014522811  0.001822348 -0.005929466 -0.0138257202
AGR2   0.010171113  0.024482873  0.018627891 -0.0141422661
ALDOA  0.007940716 -0.007355786 -0.005625332  0.0027172067
ANXA2 -0.005766246 -0.003621098  0.011360110 -0.0636632984

>

when I tried to infer a gene regulatory network by another function, all of miRNAs disappread from resulted matrix

> grn <- aracne.a(MI, 0.05)
> head(grn[,1:4])

               ACTB        ACTG1        ACTN4         AGR2
ACTB   0.0125244268  0.000000000  0.000000000  0.010171113
ACTG1  0.0000000000 -0.010771804  0.000000000  0.024482873
ACTN4  0.0000000000  0.000000000 -0.005929466  0.018627891
AGR2  -0.0006921249  0.007249163 -0.013825720 -0.014142266
ALDOA  0.0000000000  0.000000000  0.000000000  0.008910742
ANXA2  0.0000000000  0.000000000  0.000000000  0.000000000
>

what happened for miRNAs?

R parmigene mirna grn • 988 views
ADD COMMENTlink modified 2.1 years ago by theobroma221.1k • written 2.1 years ago by F3.4k
1

I looked at your screenshots. If your miRNAs have "disappeared" how is it that both MI and grn are the same size? I expected grn to be "smaller" than MI.

ADD REPLYlink written 2.1 years ago by theobroma221.1k

yes the dimension is the same but in uploading grn in Cytoscape there is no miRNAs in network even if grn has been calculated of a mutual information of both miRNAs and mRNAs. if I transpose MI and place miRNAs in row then in grn all will be miRNAs. :(

look please

https://l42i.imgup.net/grn681b.jpg

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by F3.4k
1

Ok. Scroll down to the bottom of grn and send me a screenshot. And scroll all the way to the right and send me a screenshot. According to you all 192 observations in grn are miRNAs?

ADD REPLYlink written 2.1 years ago by theobroma221.1k

yes

I uploaded grn in Cytoscape all were miRNAs

please look

https://v58i.imgup.net/miRNAd937.jpeg

ADD REPLYlink written 2.1 years ago by F3.4k

This is not an answer - you're asking for clarification. Please do not add answers unless you're actually answering the question.

I'm moving this to a comment this time, please be more careful the next time.

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by RamRS21k
1
gravatar for theobroma22
2.1 years ago by
theobroma221.1k
theobroma221.1k wrote:

I don't follow, miRNAs disappeared?...isn't the grn object the matrix needed to plot the network?

ADD COMMENTlink written 2.1 years ago by theobroma221.1k

I mean grn created by MI dose not contain any miRNAs and there only mRNA in rows and columns while I was going to plot network by bith of them. whathappened for miRNA?

ADD REPLYlink written 2.1 years ago by F3.4k
1

Ok, but your tables do not show the error you are asking for help on in your post. So, it is a bit disconnected. Can you please show how the miRNAs are not in the co-expression or correlation matrix? Before you do this, first I think it is at least due to the eps value you used (0.05) for the arcane.a algorithm, a positive numeric value used to remove the weakest edge of each triple of nodes. Second, are you certain this is the algorithm you should use?? Can you tell me why you didn't choose the knnmi.cross algo in the parmigene package? I only looked at the reference manual and didn't search for any workflow/vignette, so sorry for my ignorance if I am wrong. Also, I think there may be other (perhaps better) ways to demonstrate miRNA-mRNA interactions such as by the sequence nucleotides, or using the http://mircomb.sourceforge.net/docs/miRComb-vignette.pdf package. Skimming this package vignette workflow there are two separate datasets, one for the mrna and the other for the mirna, then a correlation matrix is made this way. As such, as stated above this made me think you are using the incorrect algo for starters.

ADD REPLYlink written 2.1 years ago by theobroma221.1k

thank you I built MI matrix by knnmi.cross(mat1, mat2) function

ADD REPLYlink written 2.1 years ago by F3.4k
1

Ok, thanks. I read https://academic.oup.com/bioinformatics/article/27/13/1876/184634/parmigene-a-parallel-R-package-for-mutual

Your miRNA may be removed because they may be consider as indirect edges. I would recommend to use a package that is specific to miRNA-mRNA interactions. This parmigene method or aracane.a() function doesn't seem to do what you want to do, per se. This being said, what happens when you use the multiplicative algo, aracane.m()? Are the miRNAs removed?

ADD REPLYlink written 2.1 years ago by theobroma221.1k
1

Also, what happen when you use eps = 0.15 in the additive method, and what happens when you use a tau = 0.05 and 0.15 in the multiplicative method? Thanks.

ADD REPLYlink written 2.1 years ago by theobroma221.1k
1

Sorry...also try the mrnet function. This may be what you need to use. Thanks.

ADD REPLYlink written 2.1 years ago by theobroma221.1k
1

I think I should try miRComb might work. miRComb needs many packages as dependencies I am still downloading them.

ADD REPLYlink written 2.1 years ago by F3.4k

thank you for your kindly helps. actually grn inferred by clr, ARACNE or whatever in this package give a grn depends on miRNAs or mRNAs placed in row the same as the result. for example if in MI matrix miRNAs are in rows whole of grn is only miRNA.. vice versa

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by F3.4k
1

Ok, then the aracane.a function removes the miRNA due to how it computes the adjacency matrix. What happens when you do grn =mrnet(MI) ?

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by theobroma221.1k

I shared screen shots of MI and grn in R

as you consider MI contains both miRNAs and mRNAs but grn only mRNAs :(

MI https://y10i.imgup.net/MIde38.jpg

grn

https://u14i.imgup.net/grnc539.jpg

ADD REPLYlink written 2.1 years ago by F3.4k
1

What value did you use for the k argument in knnmi.cross(mat1, mat2) ? This is important.

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by theobroma221.1k

knnmi.cross(mat1, mat2, 5) as default

ADD REPLYlink written 2.1 years ago by F3.4k

sorry miRComb needs miRData R package as dependency but whatever I am trying to install that telling

install.packages("miRData")

Installing package into ‘C:/Users/Lenovo/Documents/R/win-library/3.2’

as ‘lib’ is unspecified)

Warning in install.packages :

package ‘miRComb’ is not available (for R version 3.2.5)

how I can download that please?

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by F3.4k
1

Are you using Linux, Mac or windows? It states that you need to use install.packages("miRComb_x.x.zip") for windows and "miRComb_x.x.tar.gz" for Linux and Mac. And x.x is the package version number. See https://sourceforge.net/projects/mircomb/files/?source=navbar

When you post, please put as much info as possible to help you in the least amount of back and forth. Thanks.

ADD REPLYlink modified 2.1 years ago • written 2.1 years ago by theobroma221.1k
1

First you need to download the file to the local folder, then in R use the above command.

ADD REPLYlink written 2.1 years ago by theobroma221.1k

yes you all right I am in windows

ADD REPLYlink written 2.1 years ago by F3.4k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 909 users visited in the last hour