Adam M. Stolorow, David L. Duewer and Dennis J. Reeder
Chemical Sciences and Technology Laboratory, National Institute of Standards and
Technology, Gaithersburg, MD
Given identical samples, ideal measurement systems produce identical results regardless of when, where, or by whom the analysis is performed. Real chemical measurement systems seldom if ever attain this ideal. Knowing the size and general distribution of measurement uncertainty is thus critical to the effective and confident use of any real measurement.
A number of uncertainty models have been proposed for DNA fragment size measurements made with the Hae III restriction fragment length polymorphism (RFLP) protocol commonly used in North America. For fragments of size less than 10,000 basepairs (bp), a widely used "match window" is ±2.5% about the reported size. Researchers at Minnesota Bureau of Criminal Apprehension (MBCA) have proposed a set of three "match windows" for use with fragments in different size ranges. Research conducted by the Orange County Sheriff-Coroner Department (OCSCD) resulted in a series of polynomial equations relating measurement standard deviations (SD) to fragment size.
We have previously demonstrated that the observed RFLP measurement SD for fragments of size 1,000-10,000 bp originates in a 1-4 part-per-thousand variation in the relative positions of sample and sizing ladder bands. We recently completed analysis of a study (conducted by 19 members of the Technical Working Group for DNA Analysis Methods) that was designed to efficiently characterize measurement uncertainty of fragments of size 10,000 bp. This data confirms the qualitative accuracy of our proposed functional relationship between band size and measurement SD. We can now quantitatively predict the expected interlaboratory SD for fragments of size 1,000-20,000 bp.
We here compare all of the above approaches to estimating the RFLP measurement uncertainty. A ± 2.5% window is clearly inappropriate for fragments of size 10,000 bp. Both the MBCA and OCSCD approaches empirically approximate the "true" underlying relationship.
Go to proceedings home page