You've just about answered your own questions. You can't have that
many changing variables. Groups like the CDC, USDA, and many hospital
microbiologists for example standardize all of their procedures. For
their food-borne and nosocomial pathogen tracking they use the...
- same universal standard on every gel
- similar run conditions
- same software (Bio-Rad Molecular Analyst in their case)
- same interpolation (cubic spline in their case)
- same band-finding parameters
- same correlation methods
- same clustering algorithms
In most applications, I think reproducibiity is more desirable than
absolute precision, so the small differences between methods isn't as
important as getting matches via the same method. Does this sound
right to you?
On Fri, 22 May 1998 21:49:54 GMT, vargaa at unbc.ca wrote:
>The software I use for RFLP-analysis, RFLPScan, by Scanalytics, uses four
>functions to calculate fragment sizes: eg: log linear piecewise. I
>compared all four methods and of course got 4 different sizes for the same
>>If comparisons between laboratory RFLP databases are necessary, especially
>banding topologies, how can datasets be compared if different formulae were
>used to estimate fragment sizes? How can 'fingerprints' be compared if
>different methodologies (curve-fittings) were used to estimate band sizes?