[R] [R-pkgs] New package: noia
Arnaud Le Rouzic
a.p.s.l.rouzic at bio.uio.no
Fri Aug 1 17:55:03 CEST 2008
Hi the list,
A new version (0.92) of my package 'noia' will be available soon on CRAN
mirrors, and I think it might be a good opportunity to introduce it
shortly to the R community.
In summary: 'noia' will be of absolutely no interest for 99.99% of you.
The 0.01% remaining are quantitative geneticists who are interested in
measuring the effect of genes in a proper way.
Since some of you may want to know why it is worth to let this 'noia'
package take a few bytes on the CRAN server, I will try to introduce
very shortly the scientific problem. The rules that explain the
transmission of genes between generations are well known since the end
of the 19th century. One important consequence of these rules is that in
all diploid organisms (having two copies of the genomes, i.e. most of
them, including us), it is virtually impossible to produce an equal
number of offspring of each "type" from non-inbred parents: you end up,
most of the time, with frequencies such as 1/4 vs 3/4, or 7/16 vs 9/16,
or many other combinations. For quantitative traits, the link between
the genes and the characters is complex, and the effect of genes has to
be measured through models that look like cross-factor ANOVA designs.
Because of the unequal segregating ratios, these ANOVAs are highly
unbalanced, with all well-known annoying consequences: correlations
between effect estimates, non-orthogonal variance decomposition, etc.
For a long time, comparing estimates measured in differently designed
populations was not considered as a big issue. However, the recent need
of effective model selection procedures lead to the proposition of new
models aiming to a (more or less) orthogonal decomposition of genetic
effects. This package provides tools to perform linear regressions using
various models, to manipulate the genetic effects, and to compute
genotype to phenotype maps, i.e. the function explaining how the genes
expresses a given character. In addition, the 'noia' package provides a
tool to perform 'multilinear' regression, an attempt of non-linear
genotype-phenotype mapping.
In summary, 'noia' is a wrapper for lm() and nls() functions in a very
specific context: estimating the effects of genes on a given trait.
NB: 'noia' is useless if the location of the genes is not known.
Locating the genes requires a QTL detection procedure, for which
packages already exist (see Rqtl for instance).
Much more information can be found in scientific publications. Two are
particularly relevant:
* This package and its functions are described in a recent paper:
Arnaud Le Rouzic and José M. Álvarez-Castro, Estimation of Genetic
Effects and Genotype-Phenotype Maps, /Evolutionary Bioinformatics
/2008:4 225-235. http://la-press.com/article.php?article_id=887
* The corresponding statistical framework is extensively described in:
José M. Álvarez-Castro and Örjan Carlborg, A unified model for
functional and statistical epistasis and its application in QTL
analysis, Genetics, Vol. 176, 1151-1167 2007.
http://www.genetics.org/cgi/content/abstract/176/2/1151
Given the probable low impact of the package, I am not planning to
follow the R help mailing list. However, my email address is everywhere
in the manual, and I would be very happy to answer questions related to
this package; so don't hesitate to forward potential questions or to
encourage users to contact me directly.
Arnaud Le Rouzic.
_______________________________________________
R-packages mailing list
R-packages at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-packages
More information about the R-help
mailing list