Multivariate simple interval mapping (SIM) is one of the most popular approaches for multiple quantitative trait locus (QTL) analysis. Both maximum likelihood (ML) and least squares (LS) multivariate regression (MVR) are widely used methods for multi-trait SIM. ML-based MVR (MVR-ML) is an expectation maximization (EM) algorithm based iterative and complex time-consuming approach. Although the LS-based MVR (MVR-LS) approach is not an iterative process, the calculation of likelihood ratio (LR) statistic in MVR-LS is also a time-consuming complex process. We have introduced a new approach (called FastMtQTL) for multi-trait QTL analysis based on the assumption of multivariate normal distribution of phenotypic observations. Our proposed method can identify almost the same QTL positions as those identified by the existing methods. Moreover, the proposed method takes comparatively less computation time because of the simplicity in the calculation of LR statistic by this method. In the proposed method, LR statistic is calculated only using the sample variance–covariance matrix of phenotypes and the conditional probability of QTL genotype given the marker genotypes. This improvement in computation time is advantageous when the numbers of phenotypes and individuals are larger, and the markers are very dense resulting in a QTL mapping with a bigger dataset.

中文翻译：

---

基于回归的快速多性状全基因组 QTL 分析

多变量简单区间作图 (SIM) 是多数量性状基因座 (QTL) 分析中最流行的方法之一。最大似然 (ML) 和最小二乘 (LS) 多元回归 (MVR) 都是广泛用于多特征 SIM 的方法。基于 ML 的 MVR (MVR-ML) 是一种基于期望最大化 (EM) 算法的迭代且复杂的耗时方法。虽然基于 LS 的 MVR (MVR-LS) 方法不是一个迭代过程，但 MVR-LS 中似然比 (LR) 统计量的计算也是一个耗时的复杂过程。基于表型观察的多元正态分布假设，我们引入了一种新方法（称为 FastMtQTL）用于多性状 QTL 分析。我们提出的方法可以识别与现有方法识别的几乎相同的 QTL 位置。而且，由于该方法计算 LR 统计量的简单性，所提出的方法所需的计算时间相对较少。在所提出的方法中，LR 统计量仅使用表型的样本方差 - 协方差矩阵和给定标记基因型的 QTL 基因型的条件概率来计算。当表型和个体的数量较大时，计算时间的这种改进是有利的，并且标记非常密集，导致 QTL 映射具有更大的数据集。