Gene co-expression analysis

Question

[mjoyraj]

Hi, I have a FPKM matrix of 9644 columns and 15 rows from a RNA-seq expression experiment. The columns are the gene and the rows represents expression of the gene at different time points. My basic idea is to identify transcription factor binding site (TFBS) upstream of target gene. I want to do de-novo motif discovery based on over-represented sequence search in regulatory regions of target gene along with a group of co-expressed gene. Therefore, my first idea is to search genes co-expressed with my target gene. How can I do the co-expression analysis in R. How can I visualize the results? Please help me with R scripts. Let us consider the data name is 'Rdata.csv'

[Asaf]

Is it a time-series experiment?

[mjoyraj]

Yes, its a time-series experiment.

[mjoyraj]

Thank you.

[mjoyraj]

[e-mail]

Answer 1

[biogirl]

If it is time-series data, I would suggest using the TCAP method (temporal clustering by affinity propagation): http://bioinformatics.oxfordjournals.org/content/26/3/355.short

[mjoyraj]

Thank you. I will go through the paper.

Answer 2

[EagleEye]

You can do time series gene expression using your target genes along with all other genes and look for your target gene in the time series profiles. Your target gene along with other genes will be clustered in particular profile or pattern of expression, which confirms the co-expressing genes with same pattern of expression as your target.

Use this program: STEM

Hope that I understood your question properly.

[mjoyraj]

Thank you. I will see the link and paper. Does any option is available to do the analysis in in R? I thought of finding the correlation between gene expression and filter top 1% correlations. Let us consider the top 1% correlations as set. The set will next be segregated to subsets based on mutual correlation. The genes within a subsets are co-expressed genes and gene ontology analysis will be done to predict if they are co-regulated or not. A set of co-regulated genes are believed to be regulated by same TF and their co-regulated genes are likely to have same motifs. Since my approach is de-novo motif discovery I would like to see whether the co-expressed/co-regulated genes have over represented sequence motif in their regulatory region. Since the motif discovery is all based on finding correct set of co-expressed/co-regulated gene, I want to go with the best available approaches. Please suggest, which approach out performs others. If any analysis is available in R-platform please help me with the pipeline and script.

[EagleEye]

I have no idea whether there is an R package to do it. But my suggestion:

1. First you can select the target gene along with co-expressed genes from STEM analysis (which also provides you with significant values). If the profile or pattern is similar with significant value means the genes in the profile are well correlated.
2. You can use DAVID for your protein coding genes to get the GO terms or if you want command line tool with more extensive results you can use GeneSCF (only for Human): Gene Set Clustering based on Functional annotation (GeneSCF)
3. Use those Co-expressed genes sequence (there are many tools available to get sequence from chromosome coordinates, like blast).
4. Get all sequence and find the motif using the tools like MEME.

[mjoyraj]

Thanks for your valuable suggestion. Can you please elaborate the step 2 and 3. My genes are from bird genome, so can I use the the said packages (as you mentioned its only for human). What I got from your answer is that:

1. Find co-expressed gene by STEM
2. Find ontology between the co-expressed gene
3. Filter co-expressed gene having ontology
4. Retrieve regulatory region of co-expressed gene having ontology
5. Analyse regulatory regions for motif discovery in MEME

If I am right, and got your points correctly, please let me know.

Many thanks again.