Using microarray data to model epigenetic changes...

3 posts / 0 new
Last post
ryan_m
ryan_m's picture
Using microarray data to model epigenetic changes...

Can co-expression of genes that are physically close on the chromosome be used to model epigenetic changes at the chromosomal level?

I would like to suggest an approach for inferring large-scale epigenetic changes without the use of any of these new fancy methods of directly detecting epigenetic modifications. Rather, I suggest using microarray data to infer epigenetic modifications by clustering genes based not on only their expression patterns but also their chromosomal positions. Any clusters containing a bunch of genes that are physically close together on the chromosome should indicate expression differences that were caused by chromosomal structure rather than standard promoter/enhancer/silencer type interactions.
The principle that such clusters exist and can be detected has been proven previously (see Ref), but nobody to my knowledge has yet done a time-course type assay.
I suggest taking a period of development where we expect large epigenetic modifications to occur (my guess would be ES cell to embryoid body or further developmental stage, though these sorts of samples are not the easiest to come by) and run a series of transcriptome microarray chips (a couple of replicates for each time point). I expect/hope that we would be able observe trends of certain clusters appearing or disappearing, reflecting spatially regulated gene expression. This would allow us to indirectly visualize (infer, actually) large-scale epigenetic modifications during development and should greatly widen our current perspectives on gene expression.

Ref:
Visualizing chromosomes as transcriptome correlation maps: evidence of chromosomal domains containing co-expressed genes--a study of 130 invasive ductal breast carcinomas. Cancer Res. 2005 Feb 15;65(4):1376-83.

bgood
bgood's picture
First, for those not familiar

First, for those not familiar with epigenetics (like myself),

check it out on Wikipedia

Next, so your main hypothesis is that:
"Any [co-expression] clusters containing a bunch of genes that are physically close together on the chromosome should indicate expression differences that were caused by chromosomal structure rather than standard promoter/enhancer/silencer type interactions. "

To prove this, I think that you would have to add another step to your analysis that checked to see if the clustered genes actually did have different promoters etc. Otherwise, it seems likely that you would detect co-located gene families, resulting from duplication events, that are co-regulated for the simpler reason of sharing promoters etc.

With that additional evidence (which shouldn't be too hard to drum up), I think you could make a pretty strong case with this idea.

ryan_m
ryan_m's picture
I think you're absolutely

I think you're absolutely right. Thanks for that suggestion. This kind of experiment would obviously require a lot of follow-up validation. There have been some good examples of the use of DNAse to check for "open" chromatin regions as well as some methods of methylation detection (bisulfite sequencing etc). I would likely want to follow up with both of these methods to conclude any inference of epigenetic (rather than genetic) influences on gene expression.