Question: Coverage profile comparison
gravatar for faramer86
13 months ago by
faramer860 wrote:


I have results of targeted sequencing. It is several csv files  with information about amplicons (unique id of amplicon and number of reads). It looks like that (*image below). And I have to compare two coverage profiles of two samples. There are several methods to do that (Pearson correlation, Euclidean distance, Chi-square, t-test, clustering analysis and so on...), but I don't quite sure what will be statistically correct?

The goal is to validate the result. I have two samples (control and new) and I want to know that the coverage profile of new sample is the same with control sample.

*important moment: inside sample there is kind of competition for reagents, so if one amplicon will be "covered" high-covered then others will be "covered" less.



ADD COMMENTlink modified 13 months ago • written 13 months ago by faramer860

Not an answer but a suggestion: it sounds like you could fit a localised regression to the data and then compare coverage profiles that way. By doing this, you could also introduce covariates into the model. I have never done this previously; so, remains a suggestion.

ADD REPLYlink written 13 months ago by Kevin Blighe350

Thanks, Kevin! I will try this! And what do you think about correlation or chi-square? Is it statistically correct to use them in this case?

ADD REPLYlink written 13 months ago by faramer860

I think that the localised regression idea will stand better in a publication or report. You could fit the model and then check the coefficient to show that there are no differences between the 2 samples.

A correlation with a highly statistically significant p-value could assist, too. I am not sure about the use of a Chi Square test.

ADD REPLYlink written 13 months ago by Kevin Blighe350
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 189 users visited in the last hour