[BioC] How to get started with bioconductor ?

Gordon Smyth smyth at wehi.edu.au
Thu Sep 11 12:29:45 MEST 2003

At 07:25 PM 10/09/2003, Luke Whitaker wrote:

>I need to do some straightforward analysis on two-dye microarray
>data. I want to be able to estimate fold-regulation for each gene
>between two conditions (eg treated/untreated etc).
>The first chips I had to analyse had 200 genes and I "rolled
>my own" analysis using lm() etc. which seemed to work OK, but I
>now have chips with aprox 1,700 genes and the method I was using
>(essentially ANOVA with one level of explanatory factor per gene)
>now runs out of memory.
>I would like to use Bioconductor, but looking at the documentation,
>I cannot see the forest for the trees. I have had a look at limma,
>and think it may do what I want, but it seems to use quite a
>sophisticated bayesian analysis which I do not completely understand,
>and I would prefer to start at least with something fairly simple. As
>usual, I have much less time than I would like to do this.
>Is there one basic bioconductor function for estimating fold
>regulation in a two dye experiment ?

Well, I would have thought that limma is the natural package for you to use.

When you fit a model in limma, using any of the linear model functions 
lmFit, lm.series. rlm.series or gls.series, you get a fit object. The 
estimated log-fold changes are in the component fit$coefficients. Limma 
doesn't calculate ordinary t-statistics for you, because these are 
demonstrably bad. But if you want them you can get them from

ordinary.t <- fit$coef / fit$stdev / fit$sigma

The corresponding degrees of freedom are in fit$df. I am not quite clear 
what else it might be that you would like limma to provide that it doesn't. 
If you can explain further we would be interested to hear and possibly 
respond. It doesn't quite make sense to me to be asking for a package which 
is *less* capable. Since you are capable of rolling your own analysis, I 
suspect that ease of use is not the problem.


>Luke Whitaker
>Bioconductor mailing list
>Bioconductor at stat.math.ethz.ch

More information about the Bioconductor mailing list