Human chromosome analysis as computed from arm lengths measurements.

Abstract

Computer programs of both a statistical and logical nature were applied to the problem of human karyotyping. A linear discriminant function was calculated based on karyotyping and measurements made by an experienced cytogeneticist on 50 spreads, the design set, and this function was tested on a further 50 spreads, the test set. All the spreads used were from normal human females, but the procedure could easily be extended to male karyotyping, providing the appropriate ‘male’ or ‘female’ program is used in a given case. The only input measurements required and used in the design and test sets were the four arm lengths of each chromosome, measured and supplied to us as four single straightline segments from the centromere to the ends of the arms to simulate contemplated computer measurements. The analysis was done in two stages; the first assigned each chromosome into one of eight groups by its individual characteristics. The second reassigned the 46 chromosomes of each spread so that each group in each spread had its correct number corresponding to a normal human karyotype. The results of the first stage of the computer analysis disagreed with the cytogeneticist 8% of the time in the design set and 9% of the time in the test set. The second stage proved successful in decreasing this disagreement to 6% and 7% respectively. The discriminant function found and the reassignment algorithm can serve as a karyotyping system which is both quantitative and objective. It requires only arm-length measurements to be made. With computer assistance in making these and possible additional independent measurements, such a system appears promising as a basis for automating chromosome analysis.