 |
We are building a new computer program, GENINFER, to assist genetic
counselors in evaluating the risk of recurrence of genetic disorders
based on the analysis of family pedigrees. The present version of the
program integrates a convenient graphical interface that permits
counselors to draw, examine and modify family pedigrees and to enter
information relevant to risk analysis. It also includes a
general-purpose Bayesian inference mechanism that permits the rapid
calculation and display of probabilities of various genotypes, for the
consultand and all other pedigree members.
This is possible even in
the presence of complex pedigrees with multiple consanguineous
matings. The ability to support rapid calculation also enables the
user to perform sensitivity analyses. Limitations of the prototype
include a restriction to single-locus Mendelian disorders and an
inability to make use of information from RFLP markers. Planned
extensions include remedying these limitations, the incorporation of
an algorithm for automated reformatting (layout) of an existing
pedigree, improvements in the population genetic models used by the
program, and connections to external databases for acquiring data on
disease incidence, patterns of inheritance, mutation rates,
penetrance, etc.
A discussion of the original version of the program is contained in a
paper presented at MEDINFO-92,
Pedigree Analysis
for Genetic Counseling (or postscript version).
Current Effort
Ms. Rayka Yokoo has been implementing a new version of the Geninfer program
intended to run as a signed Java applet, so that it can be made available to
anyone on the Web. As a part-time effort, it is not yet finished, not
polished, and not fully debugged. Nevertheless, you may be interested
in trying it, in which case
instructions may be found here. In order to allow an applet to
write files (containing disease models and family pedigrees) to your local
hard disk, the default security sandbox of Java needs to be overridden,
hence some complexities in installation. Once the applet runs, you can
create a disease model, then a family pedigree, and draw probabilistic
inferences from the known genotypes and phenotypes of some individuals
to the unknown ones of others. |
![[GM Picture]](../graphics/mendel-sm.gif)
Gregor Mendel
![[TB Picture]](../graphics/bayes-sm.gif)
Thomas Bayes |