Digital gene expression in CLC without reference genome

[RyNkA]

Hi all,

I'm currently thinking to use CLC genomics workbench for analysis of gene expression between 2 sample (fish in 2 different conditions). However, where i got stuck is that my fish does not have a reference genome. So i think that

- I'll use one condition to do a de novo assembly first. This to create a "reference genome" without annotation. Then use the read of 2 condition and map it back to the ref => calculate RPKM => digital gene expression

Is there any problem in my methods ?

[Blahah404]

You should pool the reads from both conditions to do the de-novo assembly. By assembling only one condition, you could be missing out on transcripts that are only present at high enough coverage for assembly in the other condition. Reads from those transcripts would then be missed from your expression analysis.

You also don't want to use RPKM for differential expression calculations - RPKM is useful for plotting different genes against one another, because it normalises by transcript length. For differential expression you want the raw expression estimates.

RSEM and eXpress both perform very well for the quantification step with de-novo assemblies. If you use eXpress for quantification, you should use the rounded data from the 'estimated counts' column as input to your differential expression analysis. If you use RSEM, the 'expected counts' column is the correct one to use for differential expression.

[RyNkA]

You should pool the reads from both conditions to do the de-novo assembly. By assembling only one condition, you could be missing out on transcripts that are only present at high enough coverage for assembly in the other condition. Reads from those transcripts would then be missed from your expression analysis.

You also don't want to use RPKM for differential expression calculations - RPKM is useful for plotting different genes against one another, because it normalises by transcript length. For differential expression you want the raw expression estimates.

RSEM and eXpress both perform very well for the quantification step with de-novo assemblies. If you use eXpress for quantification, you should use the rounded data from the 'estimated counts' column as input to your differential expression analysis. If you use RSEM, the 'expected counts' column is the correct one to use for differential expression.

Sorry because i'm just enter the field in the last 2 months...I dont know much about RSEM and eXpress quantification step, is there anywhere i can look for tutorial or manual how to running them ?

[Blahah404]

I'd recommend using eXpress - it's a bit easier than RSEM (and the results are extremely similar): http://bio.math.berkeley.edu/eXpress/tutorial.html

[RyNkA]

I'd recommend using eXpress - it's a bit easier than RSEM (and the results are extremely similar): http://bio.math.berkeley.edu/eXpress/tutorial.html

Hi there, I already maintain to create a text file from eXpress that contain all the counts and stuff. What to do next ? You stated earlier above about use those number as input in DGE analysis - did you mean by using edgeR, DEseq or bayseq ?