Project 1: The Mouse Brain Library Project 2: Internet Microscopy (iScope) Project 3: Neurocartographer and Segmentation of the MBL Project 4: The Neurogenetics Tool Box



























 

 

 

 

 

 

 

 

 

 

 

 

  

 

  
 

RESEARCH PLAN

  
 

Principal Investigator/Program Director Williams, Robert W.

 
 

Staffing and oversight

The programmer requested for this project will be responsible for coding the functions of the NTB under Dr. Manlys direction. This programmer will also perform some initial testing of the system but will not be responsible for exhaustive testing. The intern to be hired in Year 02 will be responsible for exhaustive testing, under Dr. Manlys direction, using simulated and real data. The Genotyping and Mouse Colony Core, under the direction of Drs Williams and Gu, will generate the marker genotypes, in particular the genotypes of the G10 advanced intercross, needed for operation of the NTB. Genotypes needed for QTL mapping with recombinant inbred lines will be selected from public data by Dr. Williams or Dr. Manly. Documentation and tutorials will be written by Dr. Manly with advice from Dr. Williams.

Once the NTB is operational, the production servers will be maintained at the University of Tennessee, Memphis, by Dr. Williams. Both Drs. Manly and Williams will have access to these servers, and they will be jointly responsible for maintaining the software and associated data files.

Communication

Development of the NTB will not require large-scale or high-speed data transfer between Buffalo and Memphis. The code developed in Buffalo will be tested locally with local copies of the genotype data files, then transferred periodically to the production server in Memphis. The marker genotype data needed by the NTB, although extensive, will be relatively static, and copies in Buffalo will need updating only infrequently. Although we may establish a shared file system between Buffalo and Memphis, ftp would be sufficient.

Interval mapping

The heart of the interval mapping method is a set of expressions that give the expected effect for a QTL in the target interval, dependent on the analysis point (the hypothetical position of the QTL in the interval), the genotype of the flanking markers, the population type, and the dominance and interference models. The options for dominance models and interference models described above are implemented by choosing the appropriate set of expressions. Sets of these expressions for different situations have been published by many authors (Lander and Botstein 1989; Knapp, Bridges et al. 1990; Carbonell, Gerig et al. 1992; Haley and Knott 1992; Martinez and Curnow 1992; Moreno-Gonzalez 1992) . Similar expressions for dominant markers have apparently not been published, but, assuming normal segregation ratios, these can be created as appropriately weighted averages of the terms for codominant markers.

Establishing QTL significance

After estimating regression coefficients under the assumption of a normal error distribution, the NTB, like Map Manager QT, will calculate a likelihood ratio statistic or LRS (Haley and Knott 1992) as a measure of the significance of a hypothetical QTL. For interval mapping, it will calculate this statistic for every analysis point in the intervals considered. The point at which the LRS is a maximum is interpreted as a possible location of a QTL, and the value of the LRS at that point is interpreted as a measure of the statistical significance of that QTL. The LRS can be interpreted as a c2 statistic. Strictly speaking, it is an approximate c2 statistic; but, by comparing this LRS with the likelihood ratio calculated by the maximum-likelihood method in Mapmaker, Haley and Knott (1992) showed that the approximation can be quite good. In cases for which the LRS is based on one degree of freedom, the LRS can also be converted to a traditional base-10 LOD score. The significance of a QTL can be estimated by comparing the maximum LRS value with significance thresholds for c2 statistics or LODs. However, permutation tests provide significance thresholds that are not affected by the mixture problem or by the distribution of the environmental effect. In addition, a new empirical method has been proposed that promises to provide empirical thresholds with greatly reduced computational time (Piepho, personal communication). This new method is a specific application, validated by simulation, of an approximate threshold formula proposed by Davies (1977, 1987) . The NTB will provide one or both of these functions for generating empirical significance thresholds, and neuroscientists will be encouraged to use these methods routinely.

 

 
   
   
   
 

EXPERIMENTAL PLAN