[BioC] Help on log issue

Simon Anders anders at ebi.ac.uk
Sun Jan 25 15:52:50 CET 2009


Prashantha Hebbar wrote:
> We make use Agilent scanner and feature extraction software for TIFF (16
> bit) image extraction. But I have some doubts on the Feature extraction
> output. In output, I find all intensity values (example: log ratio,
> rProcessedSignal, gProcessedSignal, gMeanSignal, rMeanSignal etc) are given
> in log base 10. Now my question is,
> 1.       Whether it is common to measure microarray binary data (TIFF) in
> log base 10 or log base 2? 

I suppose you are right that in the microarray community, everybody
seems to use binary logarithms (log2). Bear in mind, however, that this
is rather uncommon in other fields: most physicists and electrical
engineers prefer to work with decimal logarithms (log10). Maybe the
read-out software has been written by someone working outside of
molecular biology.

> 2.       In case, measuring microarray intensities in log base 10 is right,
> how come LIMMA will come to know that, intensity values are in log base 10?
> How does it calculates  M and A value in log base 2?

My guess would be that LIMMA won't notice. But as you know, log10 and
log2 values are proportional, and as LIMMA works with t values, which
are ratios of log-scales quantities, the conversion factor between log10
and log2 will cancel out and won't affect your p values.

Still, you could simply divide everything by log10( 2 ) to convert from
log10 to log2.


| Dr. Simon Anders, Dipl. Phys.
| European Bioinformatics Institute, Hinxton, Cambridgeshire, UK

More information about the Bioconductor mailing list