[BioC] technical vs biological replication in real-time

[Juan Pedro Steibel]

Erratum: you have 3 true reps.
Sorry about that...
JP

Juan Pedro Steibel wrote:
> Hello Niki,
> You should not treat biological and technical replicates in the same 
> way. The standard deviations provided by the termocycler software only 
> pertain to the within sample (purely technical) variability. Moreover, 
> to infer upon treatment effects in a broad inference context you should 
> refer to the sample-to-sample variability (so-called biological 
> variability). For this last purpose you have 4 reps (not 12).
>
> You can use a hierarchical model to account for different levels of 
> replication. This can be done using a linear mixed model including both 
> the test and control gene and computing the (corrected) contrasts of 
> interest.
>
> Please follow the link to a short paper with a linear model (model [3]) 
> that we have proposed for this type of analysis. We did not use R for 
> the computations, but perhaps the lme4 package could be used.
>
> http://www.wcgalp8.org.br/wcgalp8/articles/paper/23_528-1915.pdf
>
> Best,
> Juan Pedro
>
>
>
>
>
> r.athanasiadou wrote:
>   
>> Hi everyone,
>> My question is of statistical nature.
>> I am trying to finalize the analysis of real-time PCR experiments.
>> I have 4 technical replicates (pipeting errors) and three biological
>> replicates (repeat treatment).
>>
>> I use gapdh as an external reference gene.
>>
>> 1) Can I average all replication regardless of technical/biological?
>> 2) Similarily can I use the standard error from the combined standard
>> deviations and what would the population size be in that case? (would it be
>> 4 technical x 3 biological?)
>>
>>  If I can not do the above how does someone proceed with these two different
>> types of replication?
>>
>>
>> I can imagine that the technical replication reflects only on the threshold
>> cycles and the biological replication at the actual fold
>> difference[2^-(threshold cycle)] but how can I combine the two differently
>> calculated standard errors?
>>
>>
>>
>> I apologize that the e-mail is not directly linking to some BioC package but
>> there are no packages for real-time and I had nowhere else to turn!
>>
>> Thanks
>> Niki
>>
>>
>>
>>
>> Swann Building
>> The King's Buildings
>> University of Edinburgh
>> Edinburgh, EH9 3JR
>> Scotland
>> UK
>>
>> tel:(0044)0131-6507072
>> fax:(0044)0131-6505379
>> R.Athanasiadou at sms.ed.ac.uk
>>
>>
>>   
>>     
>
>   


-- 
=============================
Juan Pedro Steibel
Postdoctoral researcher
Statistical Genetics
Department of Animal Science 
Michigan State University
1205-I Anthony Hall
East Lansing, MI
48823 USA 
Phone: 1-517-353-5102
E-mail: steibelj at msu.edu