It is widely accepted that multiple genes are responsible for the common diseases that affect humanity. Furthermore, these genes interact with each other and with the environment in complex relationships that are poorly understood. Several statistical methods attempt to solve this problem by detecting and locating these genes. The relative performance of these methods is largely unknown, as close form solutions to problems do not exist. Power analyses have to be performed by simulation. However, simulation is difficult due to the problem of testing a large number of genetic models that might adequately represent the actual range of genetic architecture underlying these complex traits. Also, these methods are all family-based, but the optimal family type (sibpair, sibship, nuclear, extended) is largely unknown.
A typical trait might have as much as 30-40% of its total variation attributable to genetic variation. The assumption made here is that this variation is due to a few genes (oligogenes), i.e., between two and six. Each of those genes will have at least two forms (alleles) to produce the variation. The relationship between the alleles can be additive or non-additive (dominant). The allele frequencies can vary. The relationships between genes can also be additive or non-additive. Individual alleles can interact with environmental factors. All of these factors must be accounted for in a simulation that captures the reality of genetic systems.
A second problem is that of family structure. A common belief is that large families are more powerful than an equivalent number of individuals organized in small families or even sibpairs. This problem obviously has important implications for the design of studies of complex traits, yet little work has been done due to the computational demands.
99/27 |
"Linkage of Diastolic Blood Pressure to D2S1790 in a Random Sample of Mexican American Families," L.D. Atwood, P.B. Samollow, J.E. Hixson, M.P. Stern, and J.W. MacCluer, University of Minnesota Supercomputing Institute Research Report UMSI 99/27, February 1999. |
Currently, this project is using supercomputing resources to address these problems by considering a broad range of genetic models and family structures. For each model-family combination, the two most popular methodological approaches are used to detect and localize genes effecting a simulated trait. These two approaches fall largely into two areas-model-based analysis and model-free analysis (or parametric and non-parametric). These approaches are represented in the software packages pap (model-based) and genehunter2 (model-free). These programs are the primary tools for analysis. For each model-family combination, at least 100 random replicates are simulated. From these replications, power and error curves are generated so that the relative performance of the two methods can be compared directly.
|
|
URL: http://www.msi.umn.edu/about/publications/annualreport/ar2000/depts/PubHealth/atwood.html |
|
| This page last modified on Friday, 30-May-2008 16:14:09 CDT | ||
| Please direct questions or problems to help@msi.umn.edu | ||
|
Website related questions or problems should be directed to
webmaster@msi.umn.edu
The Supercomputing Institute does not collect personal information on visitors to our website. For the University of Minnesota policy, see www.privacy.umn.edu. © 2001 by the Regents of the University of Minnesota |
||