How Will The Choice Of The Read Counting Method Affect The Resulting Differential Expression?

Question

[camelbbs]

Hi all,

I see currently there have several ways to calculate gene read counts for input files to DE detection.

I compared htseq-count, cufflinks, bedtools multicov, partek and these method generated different read counts for one gene.

I am curious if that will affect the Differential gene expression results. Which one is the common use.

Thanks,

Ch

Answer 1

[Istvan Albert]

Yes it will affect the differential gene expression results most likely in the following way:

• the strong observations (top quarter) with lots of data behind them will produce the same results no matter which method you use.
• the weak observations (bottom quarter) will vary wildly depending on what you pick. But not just at this step but even before, how you choose to filter the data and all other implicit or explicit assumptions etc.

As with any method: use what you understand well. Recognize and accept the fact that there is no true answer, each one is just an approximation. The performance of a technique almost always depends on the properties of the data.

[camelbbs]

enter code hereThanks a lot !!!