Assessment of validity of DNA measurement in tissue sections by model simulation

Computer modelling is used to simulate nuclear segments obtained by random sectioning through tissue. A few more computations lead to DNA measurement simulation. Two methods of DNA measurement (a direct one and a variant of the "plug" method) are tested on simulated diploid, tetraploid and octaploid cell populations. The two methods result in negatively skewed DNA frequency distributions. Both the right skewness and the coefficient of variation of measurements are increasing with the ploidy level because nuclear DNA content is assumed to be related to nuclear size in the chosen model. Observed mean values are biased underestimates of expected values but are strongly correlated to the degree of ploidy. The variant of the "plug" method gives rise to smaller coefficients of variation. Finally, the bias introduced by measuring nuclear segments instead of whole nuclei increases the variance of measurements but contributes to less than half the experimentally observed variance. Our conclusion is that microspectrophotometry on tissue sections is a valuable method for DNA content evaluation of small clusters of pathological cells as one may find in endoscopic biopsies.