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


















Principal Investigator/Program Director Williams, Robert W.


Graphical display of interval mapping analysis

Like Map Manager QT, the NTB will use graphical display of analytical results. An example of an interval mapping of simulated F2 data by Map Manager QT appears in Figure 1. The NTB will display these figures in color in an HTML page; such figures can be copied and pasted into a graphics program or saved as a file.

Although these figures have become a standard method for presenting QTL mapping results, they are largely window-dressing. The only useful information they convey is the maxima for the LRS, points which indicate possible QTL positions. Although the shape of the LRS curve has been interpreted as giving a confidence interval for QTL position (Lander and Botstein 1989) , this method is not reliable (Ooijen 1992; Mangin, Goffinet et al. 1994; Darvasi and Soller 1997)

For some types of crosses, the graphical display is unnecessary. As explained below under Fast Interval Mapping, Whittaker, Thompson, and coworkers (1996) have published an elegant method for additive QTLs which determines LRS maxima directly, This method is implemented only in Map Manager QTX. The NTB will use this method for mapping in recombinant inbred lines, which detect only additive QTL components, reserving the graphical display for intercrosses, which can detect dominance components.

Precision of QTL location

The NTB will provide two estimates of confidence intervals for QTL position, both empirically derived. Lander and Botstein (1989) proposed a simple theoretical rule for constructing confidence intervals for QTL position: the position of the maximum LOD score is taken as the position of the QTL, and the region in which the LOD score is within one LOD unit of the maximum is taken as a 96.8% confidence interval. However, this method is not accurate except for strong

QTLs. Better methods for confidence intervals have been derived by theoretical analysis (Mangin, Goffinet et al. 1994) and simulation (Ooijen 1992; Darvasi and Soller 1997) . For moderate QTLs, Darvasi provides the expression 530/Nv for a 95% confidence interval, where N is the population size and v is the proportion of variance explained (Darvasi and Soller 1997; Darvasi 1998) . However, a bootstrap method seems to be the most accurate current method for estimating a confidence interval (Visscher, Thompson et al. 1996) . The NTB will provide both of these estimates.

Software design

The NTB will be written largely in Python and C++. Python <> will be used for interface routines that present online forms and/or interpret the information returned in those forms. We anticipate that these interface routines will change more frequently, in response to user suggestions, than the analysis routines, and the flexibility of an interpreted language will be useful in implementing those changes. Python was recommended for this purpose by K. Jacobs, one of the designers of the graphic user interface for Statistical Analysis for Genetic Epidemiology <>. Speed will be important for the analysis routines, so these will be written in C++. Most of these will be simplified versions of code written for Map Manager QTX. The analysis routines can be integrated with the Python code by compiling them as new Python modules. We may also use Java applets for some functions in the data entry forms, but in the interest of making the forms as responsive as possible, these will be kept to a minimum and/or made optional.

Software produced by this project will be copyrighted to control modification and distribution, but it will be made available without charge, either as executables or as source code, to nonprofit organizations.



Next Topic