Beacham The use of DNA variation for stock identification of Oncorhynchus keta 



613 



IS 



Figure 1 



Observed frequency distribution of DNA band size over all chum 

 salmon, Oncorhynchus keta, populations surveyed for variation with 

 the probes pSsa-A33 (A) and pSsa-A34 (B). Frequencies of any band 

 over 200 were truncated and the limits are indicated in the figure 

 (i.e. 361 and 229). 



intervals for definition of alleles at the Ssa-A33 lo- 

 cus were defined by 15 classes as outlined in Table 

 1, and bin intervals for definition of alleles at the 

 Ssa-A34 locus were defined by 18 classes as outlined 

 in Table 2. The average bin width was 9.8 standard 

 deviations (SD) for bins containing bands <5,000 bp 

 and 8.3 SD's for bins containing bands >5,000 bp. 

 Variation previously observed with the Ssa 1 probe 

 was reanalyzed from the 44-bin characterization of 

 Taylor et al. ( 1994) to a 31-bin characterization, with 

 bin limits outlined by Beacham et al. ( 1996). 



Data analysis 



Mean allele frequencies were calculated by stock for 

 Ssa-A33 and Ssa-A34 loci with the bins considered 

 as alleles. Differences in heterozygosity among stocks 

 in different areas were evaluated by the log-likeli- 

 hood ratio or G-test statistic (Sokal and Rohlf, 1981). 

 Annual variation in the allele frequency distributions 



was evaluated for each probe for three stocks that 

 had been sampled over two years. The chi-square (^ 2 ) 

 test with 1,000 Monte Carlo simulations of the dis- 

 tribution of the x 2 statistic (Roff and Bentzen, 1989) 

 was used to evaluate significance of annual varia- 

 tion in allele frequency distributions (McElroy et al., 

 1992). Genotype frequencies within populations were 

 compared with those expected under Hardy- 

 Weinberg equilibrium by 500 Monte Carlo simula- 

 tions of the distribution of the % 2 statistic. As the 

 data set included both allele frequencies (Ssa-A33 

 and Ssa-A34) and band counts (Ssal), a generalized 

 or chord distance value (Pimentel, 1979) was calcu- 

 lated between each pair of stocks. A dendrogram was 

 obtained by clustering the distance estimates by 

 means of the neighbor-joining algorithm (Felsenstein, 

 1990). Classification of individual fish to specific 

 stocks was conducted by using the DISCRIM proce- 

 dure of SAS (SAS Institute Inc., 1989) with a jack- 

 knife sampling procedure, whereby classification 



