14.06.2013 Views

Databases and Systems

Databases and Systems

Databases and Systems

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

88<br />

Figure 2: Whitehead Radiation Hybrid Map vs. Stanford RH Map of Chr.3<br />

(The nonlinear relationship evident between the maps is somewhat<br />

surprising; this is to be expected between linkage <strong>and</strong> physical maps because<br />

recombination is inhibited near the centromeres <strong>and</strong> telomeres, resulting in a<br />

compression of genetic distances in those areas <strong>and</strong> a relative expansion in the<br />

chromosome arms, where most recombination takes place. In the figure both maps<br />

are radiation hybrid maps, which might have been expected to be more linearly<br />

correlated. The nonlinearity may be a function of the different radiation doses used to<br />

create the panels the maps were based on, which affects the density of breakpoints. If<br />

the centromere is nonr<strong>and</strong>omly retained in radiation hybrid cells, then the different<br />

panels might experience different degrees of centromeric distortion. However these<br />

are just guesses; to my knowledge this effect has not been well studied.)<br />

Although linear transformation provides a practical solution to the map alignment<br />

problem, it is by no means perfect. With any universal coordinate scheme a locus<br />

which appears on several maps may be assigned slightly different universal<br />

coordinates in each of them, <strong>and</strong> so will appear several times in several different<br />

places in the universal coordinate space. Ideally these places should be close to each<br />

other, unless the locus was incorrectly placed in one of the maps, or a locus<br />

identification error was made during entry of the map into GDB. Figure 3 shows one<br />

way to visualize this effect, which we call a dispersion plot: each x,y point represents<br />

a locus having a universal coordinate x from one map, <strong>and</strong> y from another. (The same<br />

pair of maps will also generate a point for this locus at y,x, so the plot is symmetric.)<br />

In the ideal case where every locus ended up at a unique universal coordinate all the<br />

points would fall along the line y=x. This distance of points from this line is a<br />

measure of the variability in the assignment of universal coordinates, <strong>and</strong> the overall<br />

thickness of the distribution intuitively represents how good the linear transformation<br />

approach is.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!