Re: Bug#155071: ITP: loki -- [Biology] MCMC multipoint linkage and segregation analysis of multiple QTL
On Thu, 1 Aug 2002, Carlos Enriqe Carleos Artime wrote:
> > Fine. Could you perhaps make some comments for Debian-Med or should we just
> > copy this description to
> > http://www.debian.org/devel/debian-med/microbio
> Ups, I'm not a Debian developer (yet...), so I would need first a
> sponsor for the package.
> Anyway, I'm not fluent in English, so I quote here some paragraphs
> from Loki's documentation just in case they are useful:
> The programs described here are an implementation of the methods
> described in S.C. Heath (1997) "Markov chain Monte Carlo segregation
> and linkage analysis for oligogenic models",
> Am. J. Hum. Genet. 61:748-760, for MCMC analysis of quantitative
> traits. The programs are intended for performing multipoint linkage
> analyses on large, possibly complex, pedigrees. In contrast to
> `normal' linkage packages, the exact penetrance model and even the
> number of QTL affecting the trait do not have to be specified "a
> priori". Joint estimation of QTL number, position, and effects is
> possible, avoiding the problems that can arise from mis--specification
> of these parameters. It should be noted that although the aims are
> the same as with traditional linkage packages, the operation of the
> programs and (more importantly) interpretation of the output are very
> different. This program does *not* produce LOD scores - if you only
> want LOD scores then you should look elsewhere. One reason for this
> is that the LOD score approach is less flexible than the one used
> here, particularly when models with multiple trait loci are
> considered. The other reason is that LOD scores are computationally
> costly (and tricky) to reliably estimate using MCMC. The approach
> used here is to estimate the posterior probability for any given
> chromosome region of at least 1 trait locus being in that region (and
> possibly over a given size threshold).
> It is also possible to perform affected only IBD sharing analyses.
> Although the same programs are used as for the parametric analyses,
> the way the output is interpreted is quite different. In addition, it
> is also possible for Loki to estimate the pairwise realized kinship
> coefficients (i.e. the kinship coefficients for all pairs in the
> pedigree at particular genomic positions conditional on linked
> markers). These are currently not used by Loki, though in the future
> it is planned to use this mechanism for implementing variance
> component approaches for linkage mapping. Loki has been used to
> produce kinship coefficient matrices for variance component analyses
> using external programs (Visscher PM, Haley Cs, Heath SC, Muir WJ,
> Blackwood DHR, 1999, "Detecting QTLs for uni- and bipolar disorder
> using a variance component method", Psych. Genet. 9:75-84).
> There are two main programs, Prep and Loki, which respectively
> handle data input and the MCMC analysis. The data preparation
> program, Prep, reads in data from a wide variety of input formats,
> tests for consistency of pedigree and marker data, recodes the data,
> and outputs standardized binary files for the analysis program, Loki.
> Also included in the package are several small programs and scripts
> which help in the interpretation of the output from Loki.
> Also included with the loki_2.3 distribution is a small collection of
> short programs and perl scripts which I use to help analyze the output
> from Loki. These are included in the hope that they might be of use
> to other people. Feedback on the scripts (and on Loki as a whole) is
> welcome. If the scripts prove useful then they might be expanded to
> fit with people's requirements.
> Simon C. Heath
> > I think includion into med-bio meta package would make sense.
> > Kind regards
> > Abdreas.
> The package includes at present a routine with a non-profit license,
> that makes it non-free (med-bio-contrib?). Upstream author (Simon
> Heath) told me that he has already replaced that routine by one of his
> own. So the next release of the program, coming soon, will be
> completely free according to the DFSG.