Community Research and Development Information Service - CORDIS

Statistical aspects of evaluating quantitative effects at QTL, and incorporating these in the estimates of breeding value of candidate bulls

When applying MABLUP methodologies in which QTL are included as random effects the variance explained by the polygenic component is estimated simultaneously with the variance explained by QTL and thus the problems addressed in WP12 are overcome. Other issues found necessary replaced the work months allocated to this issue.

We negotiated the scientific use of the MABLUP software developed at VIT Verden (Vereinigte Informationssysteme Tierhaltung w.V., Heideweg 1, 27283 Verden/Aller, Germany). All partners in collaboration with their national breeding organisations are currently producing data needed for MABLUP testruns. Results of the testruns will allow breeding organisations to evaluate implementation of MAS in their cattle populations, based on some of the QTL found in the BovMAS project.

An improved mapping methodology for selective DNA pooling data:
Approximate interval mapping (AIM) was developed: Single marker across sire test statistics (TS) are more or less strongly affected by the number and QTL status of the specific sires that are heterozygous for a given marker. Given the single marker TS an approximate multiple marker method was developed to predict TS for markers for which a sire was homozygous or at any other location on the chromosome, to extract maximum information on linkage. A simple selection index analogy was used to make multipoint predictions, exploiting the fact that the prediction of a TS at location l, given an observed TS on location i, is only a function of the genetic distance and hence the recombination rate. Power and map resolution of the proposed method were assessed by simulation. Power of AIM is higher and average bias of QTL location estimates smaller compared to single marker mapping. The advantage of AIM increases with decreasing marker informativity of the sires.

Power of different pooling strategies -index versus single trait mapping- were evaluated by stochastic simulation:. When mapping is focused on multiple traits the relative propitiousness of selective DNA pooling is reduced, even though genotyping costs are not crucial in this methodology. One approach of dealing with this limitation is to use a selection index for pool formation. Traits considered were: milk yield (MY), protein percent (PP), maternal fertility (FF) and somatic cell count (SCC) versus using a selection index, composed of all four traits, for pool assignment. Results indicate that the significant loss in power for single traits when mapping is on an index does not justify pool formation according to a selection index in selective DNA pooling, although power of QTL detection with favourable pleiotropic effects on the selection index was increased and cost savings were obtained. It seems to be advisable for most multi-trait studies to prepare separate pools for each trait instead of following the strategy of selection index pool formation. However, in situations with few, highly correlated traits and a shortage of resources, pooling on a specific mapping index might be an option.

To study different selection criteria for pool formation, real data of 21.616 Austrian RP2 &RP3 daughters were analysed for Protein Yield PY (h2 -0,30) and maternal fertility (non return rate after 90 days) (h2-0,02). High and low pools were created by means of 4 different selection criteria: yield deviation (YD), corrected yield deviation (CYD), which were corrected for half of the dam EBV, BLUP EBV and corrected EBV (CEBV), which were calculated by subtracting half of the dam EBV from the EBV of the daughter. Results showed that EBV are not appropriate selection criteria, because especially for low heritability traits there is a high selection pressure on the maternal side. Correction for the maternal influence is achieved using CYD and CEBV. Optimal selection criteria for pool formation are CYD since they represent unregressed values leading to unbiased estimations of QTL effects. Thereupon wherever possible CEBV were used as selection criteria in the genome scans of all RP. Following these analysis we evaluated the power of different selection criteria via stochastic simulation. Each sire was randomly to generate 2,000 female progeny per sire and selective DNA pooling was simulated with the 10 % best and worst daughters according to their EBV or CEBV for MY, PP, FF and SCC.

Furthermore MPFP (Marginal Proportion of False Positives) a criterion for setting significance levels was developed as a modification of the PFP (Mosig et al. 2001; Fernando et al. 2003). While the PFP for a particular test represents the PFP among all tests up to the (marginal) test in question; the MPFP represents the PFP of the last group of tests added to reach the marginal test. MPFP for the marginal test is inferred from the change in overall PFP produced by adding the last group of tests. MPFP will generally be greater than PFP, and provide more flexibility in deciding which tests are to be declared significant.


Johann SOELKNER, (Head of Unit)
Tel.: +43-1-476543271
Fax: +43-1-476543254
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top