P4040 Computer analysis of genetic parentage: application in equine diversity maintenance in Brazil

Genotypic STR characteristics of equine breeds in Brazil are not well described, preventing development of computational statistical tests suitable for determination of genetic parentage. With this information, the minimum number of most informative markers in Brazil could be calculated, helping decrease endogamy, maintaining the genetic variability. DNA was extracted from hair equine samples. Offsprings and both parents (LinkGen banking) were examined. For genetic equilibrium test and calculation of markers frequencies, only animals that were not parents of others were considered. This criterion was established to avoid statistical dependencies between animals. This filter resulted in 689 animals, distributed as follows: Arabian Horse (A): 178; Crioulo (C): 164; Brazilian Sport Horse (BH): 163; Quarter Horse (QH): 184. Fourteen dinucleotide ISAG markers were studied: ASB17, VHL20, HTG10, HTG4, AHT5, AHT4, HMS3, HMS6, HMS7, LEX3, LEX33, ASB2, ASB23 and HMS2. Twenty-two other tetra and pentanucleotide markers (called LINK01 to LINK22), present on Chrs 1 to 18, were also studied. A computer system for parentage examination was developed, adapting and extending the Bayesian Networks methodology. For genetic linkage analysis, markers whose descriptive level (P-value) was less than 0.05 and also markers containing only one allele were excluded. After this filtering, 16 markers remained. The question was: What is the probability of a tested individual, with certain genotype, to be the biological product of tested parents? Or P (T | A = a, B = b, i = i). P (A) P (B) P (T) P (F | A, B), P (I | M, T) are known. Thus, the probability of genetic linkage, given n markers, is given by: P (T = 1 | A = a, B = b, R = i) = P (t = 1) R/(1 + R). Of the 16 valid candidate markers, 10 were considered to make an optimum combination: ASB17, VHL20, AHT4, ASB23, LINK11, LINK12, LINK13, LINK15, LINK16 and LINK08. The following indicators were obtained: sensitivity: 0.93; specificity: 0.98; area under the ROC curve: 0.96. Regarding breeding markers, the combination of 14 was informative: ASB17, AHT4, ASB23, HMS2, LINK09, LINK10, LINK11, LINK12, LINK15, LINK16, LINK02, LINK05, LINK07 and LINK08. The estimated overall accuracy was 69%, with fluctuations (BH and QH: 63%, C: 69%; A: 80%). In conclusion, we described six new genetic markers with discrimination power for linkage calculation in Brazil. Ten new genetic markers were shown to contribute to breeding discrimination power. Financial support: FAPESP, Grant No. 2010/50258–6.