對F3族群定位數量性狀基因座的統計方法研究

Abstract

用來研究數量性狀基因座(Quantitative trait loci, QTL) 的定位之族群大多為回交與自交族群, 而本文提出一針對互交F3族群的QTL 定位統計模式, 並且討論其定位之效率與估計。由於數量性狀基因座的基因型態未知, 故統計模式為一個混合常態的模型。將混合常態模型視為不完整資料問題, 可以採用EM 演算法求出QTL 的作用與位置的估計值。由於F3提供更多的重組訊息, 所以此種族群的定位有可能會比F2來的精準。因此, 本文藉由所提之統計模式研究F3族群在QTL 定位上的性質, 並在控制樣本個數、遺傳率與兩基因座之間的距離來做模擬研究, 比較F2和F3族群在QTL 定位上的效率, 我們發現在針對兩連鎖QTL 的偵測和估計方面, 一般而言F3族群都比F2族群容易同時偵查出兩個連鎖的QTL, 並且估計值較為準確, 尤其是在小樣本, 基因座距離很靠近或是遺傳率很低的時候, 其偵測的效果差異更為顯著。本研究可做為針對利用Advanced intercross populations 來進行QTL 定位研究的基礎同時對於改善QTL 的定位有莫大的幫助。

Most of the current statistical methods of QTL mapping are developed for the backcross and F2 populations. The inter-crossed F3 population is also a popular experimental population for QTL study. As the F3 population can provide more recombinants than the backcross and F2 populations, it could be more efficient in mapping for closely linked QTL. We propose a statistical model to estimate the effects and positions of two closely linked QTL for the F3 population. As the genotypes of QTL are usually unknown, the tatistical model is a finite mixture model. By treating the mixture model as an incomplete data problem, the EM lgorithm is implemented to obtain the maximum likelihood estimates of the parameters. Simulations were used to illustrate the performance of the proposed statistical QTL mapping model in the F3 population and evaluate the relative efficiency in the F3 population (as comparing to that of the F2 population) in QTL mapping under several factors, such as sample size, genetic distance between QTL (and markers), the proportion of trait variance contributed by QTL (i.e. heritability). It is found that the QTL mapping by using the F3 population is more powerful and precise in separating two closely linked QTL and estimating the parameters than the use of the F2 population under those controlled factors, especially for small sample size, close genetic distance and low heritability. The paper can help outlining a strategy of detecting tightly linked QTL for the advanced intercross populations to improve the resolution of genetic architecture of quantitative traits.