[BioC] interaction term contrast in Weaver mutant example of LIMMA manual

James W. MacDonald jmacdon at med.umich.edu
Wed Sep 9 17:08:55 CEST 2009



Erika Melissari wrote:
> Dear Jim,
> 
> thank you for your always usefull explanations.
> If we observe a gene up-regulated in mutant "class" and also in the wt
> "class" but the extent of this up-regulation is different and
> statistically significant, we then should detect this gene as
> differentially expressed in this contrast...is correct?

If I understand your question correctly, then yes. You can have genes 
up-regulated in both wt and mutant, but less so in one than the other. 
If the difference in up-regulation is large enough and the variability 
within groups is small enough then you should get a significant interaction.

Best,

Jim



> 
> 
>> Not four arrays, but four treatment/sample combinations. You will need
>> replication to compute this contrast.
>>
> 
> Yes, I know that if I increase the number of contrasts I need more degrees
> of freedom to compute the statistics. I referred only to direct
> comparison. If I have more samples and perform a large reference design
> can I compute this interaction term the same?
> thank you
> 
> Erika
> 
>> Hi Erika,
>>
>> Erika Melissari wrote:
>>> Dear bioconductor list,
>>>
>>> I am studing the Weaver experiment example of LIMMA manual and I am in a
>>> maze about the biological meaning of interaction term present among the
>>> contrasts.
>>> I did not manage to understend what kind of differentially expressed
>>> gene
>>> this contrast takes-out.
>> In this experiment we can look at genes that change between the two time
>> points in the mutant samples, or that change between the two time points
>> in the wild type samples. In addition, we might be interested in those
>> genes that react differently in the two sample types. In other words, a
>> particular gene might increase expression in the wild type samples, but
>> actually decrease expression in the mutant samples.
>>
>> This third comparison is known as an interaction.
>>
>>> The interaction contrast is:
>>>
>>> (P21mt-P11mt)-(P21wt-P11wt)
>> Correct. So using our example from above, the first term will be
>> negative (since the expression level went down in mutant samples). The
>> second term will be positive, since the expression went up over time in
>> the wild type samples, but since we subtract, we end up with a
>> (possibly) large negative number.
>>
>> The same will be true of any scenario you can envision where the mutant
>> and wild type samples react differently to the incubation time. Another
>> example would be the situation where the mutant samples were
>> up-regulated at 21 hours, but the wild type samples were unaffected. In
>> this case you can see that we might have a large value for this contrast.
>>
>> However, if both wild type and mutant samples were up-regulated
>> approximately the same amount, the resulting value would be very close
>> to zero.
>>
>>> The first bracket takes-out the differentially expressed genes between
>>> the
>>> mutant subjects at two different time points(21 and 11 minutes) and the
>>> second bracket those between the wild-type subject at the same different
>>> time points.
>>> ..but what differentially expressed genes takes-out the global
>>> interaction
>>> contrast?
>>>
>>> In order to evaluate an interaction do I need absolutely the four arrays
>>> in direct comparison done in this example?
>> Not four arrays, but four treatment/sample combinations. You will need
>> replication to compute this contrast.
>>
>> Best,
>>
>> Jim
>>
>>
>>> Thank you very much for any explaination
>>>
>>> Erika
>>>
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at stat.math.ethz.ch
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>> Search the archives:
>>> http://news.gmane.org/gmane.science.biology.informatics.conductor
>> --
>> James W. MacDonald, M.S.
>> Biostatistician
>> Douglas Lab
>> University of Michigan
>> Department of Human Genetics
>> 5912 Buhl
>> 1241 E. Catherine St.
>> Ann Arbor MI 48109-5618
>> 734-615-7826
>>
> 
> 
> 
> 
> Erika Melissari
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

-- 
James W. MacDonald, M.S.
Biostatistician
Douglas Lab
University of Michigan
Department of Human Genetics
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618
734-615-7826



More information about the Bioconductor mailing list