David L. Duewer and Dennis J. Reeder
Chemical Science and Technology Laboratory, National Institute of Standards and Technology, Gaithersburg, MD, 20899

× Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø

In collaboration with the College of American Pathologists and member laboratories of the Technical Working Group for DNA Analysis Methods, we have developed graphical analysis techniques for use with interlaboratory RFLP proficiency studies. These graphical techniques provide intuitive, yet quantitative access to a great deal of useful information often lost in data tabulations and univariate statistical summaries.

Given that most samples give two-banded RFLP patterns at most loci, a single bivariate scattergram captures in a very natural manner all the reported information for a given sample at a given locus. Each multilaboratory scattergram plotting "high band" versus "low band" results for all participants enables quantitative assessment of "how close" the results are to one another as well as qualitative identification of any "outlier" data.

Detailed evaluation of the (number of samples) x (number of loci reported) multilaboratory scattergrams can, however, become tedious. Proper control of the origin and scale of the multilaboratory scattergram axes allows construction of summary plots for individual laboratories. If the axes are all centered on the multilaboratory median and scaled to constant expected measurement standard deviation (Stolorow et al. 1996), all data submitted by a given laboratory for all samples at all loci can be combined as one graph. These standardized single laboratory scattergrams are powerful tools for demonstrating RFLP measurement competence and for diagnosing potential measurement problems.

Thirty-four laboratories reported a total of 46 sets of results, with some laboratories reporting results for more than one analytical method. The level of experience differed widely among the laboratories, ranging from novice to very experienced. The basic analytical methods included traditional static gel electrophoresis (post-electrophoresis analysis, with fragment size related to migration distance after a fixed time) and newer dynamic gel electrophoresis (real-time analysis, with fragment size related to migration time at a fixed distance). A variety of amplification reagents and protocols, gel matrices, and electrophoretic conditions were employed by the participating laboratories.

All results that were reported as nominal alleles ("calling to the allele", i.e., assigned an allelic name rather than a quantitative size) were in agreement. Results reported as quantitative fragment size systematically varied by as much as four base pairs. While these quantitative sizings were each successfully calibrated to provide correct allelic calls, they could not be successfully compared without calibration. Some of these differences are directly attributable to differing amplification reagents, calibration protocol, instrumentation, and gel matrix composition.

REFERENCES