How to identify significant differentially expressed genes and gene regulatory networks from microarray data.

Question

[morteza.mahmoudisaber]

I have two sets of gene expression data from agilent microarray (SurePrint G3 Human Gene Expression v3 8x60K Microarray), one for a gene perturbation sample and the other as control. And now I need to identify the significantly differentiated expressed genes and the regulatory networks affected by the gene perturbation. I have good knowledge on Python and intermediate on R.

Online search I found lots of suggestions. But I wanted to know If anyone has experience and suggestions for a good point to start.

Thank you

[zizigolu]

I have recently done such a job so.

1- by http://mapman.gabipd.org/web/guest/robin I extracted DE genes and by R implementation of GENIE3 algorithm inferred a robust GRN with the most true edges specially if you normalize arrays by RMA method (robin software will do that for you). on the other hand when you are ready with DE genes list you can use ARACNE algorithm embedded in Cytoscape to infer a GRN.

or

2- R minet package has these simple functions to infer a GRN

**data(your normalized gene expression list in which genes are in columns and samples are in row)

mim <- build.mim(syn.data,estimator="spearman")

net <- aracne(mim)**

however might be you want to evaluate the predicted edges in your GRN, I don't know if there is gold standard for your desired organism to be used as a reference network for evaluating your inferred network. in brief you can find modules in your GRN for example by Glay app in Cytoscape and finally classify genes in each modules in GO terms.

3- http://dream.broadinstitute.org/gp/pages/index.jsf has some modules to infer GRNs the input file should be expression file in which genes are in columns and no need any sample name in rows and rows should leave empty

4- finally this is very good and simple step wise tutorial to construct a GRN without need to any bioinformatics skills

http://virtualplant.bio.puc.cl/Lab/doc/Moyano.etal-2014.pdf

[morteza.mahmoudisaber]

Thank you @Angel for the information. The gene expression analysis was done using Agilent microarray: "SurePrint G3 Human Gene Expression v3 8x60K Microarray " which is not supported by "RobiNA" for DEG analysis. Do you have suggestions for agilent array DGE analysis?

[zizigolu]

actually I never did Agilent data analysis but I did Robin supports that

http://mapman.gabipd.org/web/guest/forum/-/message_boards/message/21978

[Farbod]

Dear Angel hi,

Is this link about Agilent data analysis ?

Take care

[zizigolu]

hi Farbod,

I checked Robin and supports Agilent the link is a forum about Agilent in Robin software.

[rajasekargutha]

use limma or rankprod bioconductor packages for identifying DGEs

[Farbod]

Dear Morteza, Hi

maybe this post and this and DESeq2 will help you.

~ Best

[WouterDeCoster]

DESeq2 is for count data as in RNA-seq fragment counts, there will be more appropriate tools for microarray data (intensities), such as limma. The two (counts and intensities) are not equivalent since the distribution is different.

[Farbod]

Hi my friend, WouterDeCoster

you are right,

I just mentioned that PDF as a post to observing some graphs not using exactly that package.

~ Take care

Answer 1

[WouterDeCoster]

This workflow will definitely put you in the right direction: https://www.bioconductor.org/help/workflows/arrays/

[morteza.mahmoudisaber]

Thank you @Wouter. This workflow is exactly what I need but the only problem is that it is just for Affymetrix array. I have used agilent microarray: "SurePrint G3 Human Gene Expression v3 8x60K Microarray " I found the package agilp for agilent but it has only the options of:

AALoess Normalises a set of gene expression data files using LOESS
AAProcess   Extracts raw expression data from Agilent expression array scanner files.
Baseline    Constructs a file with the mean of each probe from a set of raw expression array data files
Equaliser   Trims a set of gene expression data files to include only the set of identifiers common to all files
filenamex   A file name listing utility
IDswop  Mapping expression data across bioinformatic identifiers
Loader         A file choser utility file

is there any package conducting similar analysis for agilent microarray results as that of affymetrix?

[WouterDeCoster]

It would have been useful if your initial question also included that it's an Agilent array, since that's not the most common type. I assume (but have never done it myself) that if extract raw expression data from the Agilent array you can continue with the limma package for normalisation and differential expression analysis. In addition, using google I found this: http://matticklab.com/index.php?title=Single_channel_analysis_of_Agilent_microarray_data_with_Limma

Answer 2

[Marge]

One option I would recommend to investigate co-expression networks changing between the two conditions is the R package WGCNA (Weighted Gene Co-expression Network Analysis). Several tutorials and published application examples are available and there is also very active and informative support.

Answer 3

[Farbod]

Hi,

Analyze your own microarray data in R/Bioconductor (R-code to identify DE genes based on Affymetrix microarray)

Using R/Bioconductor for Microarray Analysis

Take care

[morteza.mahmoudisaber]

Hi Farbod, Thank you for the editing and answers. Actually your answer work perfectly for Affymetrix arrays. But I have used agilent microarray that is more comprehensive and accurate than affymetrix but seemingly less popular.

[Farbod]

Is there any chance for you to use GeneSpring in your Lab, Morteza?

[morteza.mahmoudisaber]

Hi Farbod San, Unfortunately we do not have access to GeneSpring at the moment. I will try the solutions suggested here. Thank you all.

[Farbod]

Hi Morteza Jan,

Is the workflow of DEG analysis really different from "Affymetrix arrays" to "agilent microarray" ?

Take care

[zizigolu]

Farbod I think the first step in both Agilent and microarray is normalization. output of robin is a file exactly the same output of limma R package

[Farbod]

Hi there,

it seems that you are right and I think that the micro-array DEG analysis has it's standard pipeline and i think the @Wouter suggestion about "if extract raw expression data from the Agilent array you can continue with the limma package", would be a good place to start.

~ Best