[R] a general question

culpritNr1 ig2ar-saf1 at yahoo.co.uk
Tue Mar 10 19:31:56 CET 2009

```Hello Bogdan,

Put in those terms, option b looks more defensible. It sounds like a test of
two proportions, sometimes called z-test. The problem is that, for that test
to be used, you must be sampling from large population.

You know that under regular ChIP-seq sequencing we are lucky if we get 10
reads in a particuar region of interest. So, we get a population and that
population itself is small. So, z-test does not look applicable. Google for
z-test and you'll see the conditions.

What to do then?

Well, forget comparing signal to backgroung and, instead, compare directly
number of reads in the experiment versus number of reads in the control.
Actually, you simplify your algorithm by not having to define (and defend)
an arbitrary area to call "background".

An example?

Sure! Check out "PeakSeq enables systematic scoring of ChIP-seq experiments
relative to controls", Joel Rozowsky1, Ghia Euskirchen2, Raymond K
Auerbach3, Zhengdong D Zhang1, Theodore Gibson1, Robert Bjornson4, Nicholas
Carriero4, Michael Snyder1,2 & Mark B Gerstein1,3,4. Nature Biotechnology,
2009.

Take a look at it and let us know what you think.

PS: next time, please go for a more descriptive subject such as ChIP-seq.
That would help in the future when we need to over old messages. Don't you
think?

Bogdan Tanasa wrote:
>
> Hi everyone,
>
> Although this question is more related to ChIP and ChIP-seq, it could be
> probably anchored in a more general statistical context.
>
> The question is : what method is better  to assess the significance of the
> change  in a signal (the signal can be DNA binding, for instance) given
> the
> background and 2 conditions.
>
> <. condition1 (eg no treatment) :  background = 1;
>                                                 signal = 5;
>
> <. condition2 (eg hormonal treatment) : background = 3;
>                                            signal = 6.
>
> The methods can be :
>
> a. substract the background : i.e. (signal_treatment -
> background_treatment)
> / (signal_no_treatment - background_no_treatment)
>
> b. calculate the fold change: i.e. (signal_treatment /
> background_treatment)
> / (signal_no_treatment / background_no_treatment)
>
> c. any other method ? i.e. (signal_treatment - signal_no_treatment)  / (
> background_treatment - background_no_treatment)
>
> Thank you very much.
>
> Bogdan
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
>

--
View this message in context: http://www.nabble.com/a-general-question-tp22382289p22440722.html
Sent from the R help mailing list archive at Nabble.com.

```