Software Development Of Mapping Quantitative Trait Loci By Using Penalized Maximum Likelihood Method

Quantitative trait loci (QTL) mapping is to locate the position of QTL on the chromosome and estimate its genetic effects by using the relationship between phenotypic values of quantitative and marker information of the entire genome. At present, QTL mapping is widely applied in genetic analysis of complex traits, plant genomics, genetic improvement of germplasm and molecular marker assisted selection breeding.Until now, there have been proposed many approaches for QTL mapping, including single marker analysis, interval mapping, composite interval mapping, multiple interval mapping and Bayesian methods. However, epistasis plays an important role in genetics and evolution of complex traits. An epistatic genetic model should include potential pair-wise interaction effects of all QTL in genome. In this case, the model is saturated quickly as the number of major QTL increases. If only considering the sub-model of the whole model, there may be the risk to miss some important interaction effects. Therefore, Zhang & Xu (2005) proposed a penalized maximum likelihood (PML) method to estimate the parameters in the oversaturated genetic model. In the PML, penalized likelihood function to be maximized is the production of likelihood function and the penalty factor, which is composed of the joint prior distribution of all the parameters. In the software developed in this article, all parameters for all situations were estimated by the PML approach.As the widely application of mapping QTL for quantitative traits, many QTL mapping packages have been developed, such as, Windows QTL Cartographer, MapQTL and QTLNetwork. However, the operation of these packages seems to be difficult in the term of parameter setting and data transformation. Therefore, a package for mapping QTL, named as PMLqtl (Penalized Maximum Likelihood package for QTL mapping), was developed in windows platform by using the PML approach. The operation of this new package is just to import the excel data sheet wihout complex parameters setting and with strong feasibility.The package PMLqtl is a data analysis platform that maps the position of QTL for quantitative trait on the chromosome and estimate all effects of the QTL by using both the PML and the relationship between phenotypic observations of quantitative trait and molecular marker information on the genome. The results were as follows. 1) The development of package PMLqtll in the germplasm resource of inbred lines. In the PMLqtll, three approaches, including multi-QTL Haseman-Elston regression, multi-QTL in-silico mapping and multi-factor ANOVA, wre used to detect QTL in the germplasm resource of inbred lines. The genetic models under consideration were multi-QTL genetic model for the first two methods and main-effects+environmental effects+ environmental interaction for the last one. The package PMLqtll, along with excel data format, the easy operation and the visible results, makes users to work quickly and easily.2) The development of package PMLqtl2 in experimental population derived from two inbred lines. In the PMLqtl2, segregating population includes F2, F2.3, backcross (BC), double haploid (DH) and recombinant inbred lines (RIL); the methods to be selected includes mulit-marker joint analysis and multi-marker interaction analysis; and the effects to be considered in the genetic model are additive and additive-by-additive for BC, DH and RIL, and additive, dominant, additive-by-additive, additive-by-domianat, dominant-by-additive, dominant-by-domianat for F2 and F2:3. The operation of the PMLqtl2 is the same as the PMLqtl1.3) The validation of the PMLqtll and PMLqtl2. First, two BC datasets and two inbred-line datasets, including phenotypic values of quantitative trait and molecular marker information, were simulated by using Monte Carlo simulation with SAS 9.13. Then, the two packages were used to analyze the simulated datasets above. The last step was to validate the results by the true values previously set in SAS. The results showed that the two packages are feasible and reliable. In addition, permutation experiments could be carried out in the software of PMLqtl, and the results from permutation ecperiments could improve the accuracy of QTL detection and reduce false positive rate in the detection of QTL.