Comparison of the likelihood ratio and identity-by-state scoring methods for analyzing sib-pair test cases: A study using computer simulation

Abstract

To assess the power and significance of the likelihood ratio (LR) and the identity-by-state scoring (IBS) methods for a pair of siblings, we performed computer simulations by use of 10 DNA markers (HLA-DQα, D1S80, and 8 short tandem repeat loci) that were frequently analyzed in paternity tests in Japan. The combined power of discrimination of these 10 markers in the Japanese population is 0.999 999 999 98. Pedigrees each consisting of 10 000 pair of full-siblings, half-siblings and unrelated individuals were generated and typed on all markers as random samples. Both the summation of log10LR and IBS of each group had approximate standard normal distribution with significant differences between the means. Statistical studies showed that the LR method has 91% power to detect unrelated individuals and 38% power to detect half-siblings as not full-siblings with a 5% false-positive rate, whereas the IBS method does 87% and 42% powers, respectively. In 62% of full-siblings, in contrast with only 0.2% of unrelated individuals, the values of LR exceeded 100 which was equivalent to 0.99 of probability of full-sibship at 50% prior probability. The advantage of the LR method over IBS method was convincing especially for the detection of unrelated individuals as not half-siblings, however, the latter would be also informative for sib-pair tests if sufficient number of polymorphic markers are available. © 2001 Tohoku University Medical Press.