1.数据

首先了解要使用的example数据集。

They measured systolic blood pressure in 250 male mice from a backcross between the hypertensive C57BL/6J (B6) and normotensive A/J (A) strains. (B6xA)F1 mice were mated to B6 mice to produce a total of 250 male mice. The data are included with R/qtl, and will be referred to as the hyper data.
也就是一个高血压的回交

2.中心思想

The QTL mapping problem is best split into two distinct parts: the missing data problem and the model selection problem.
缺失数据处理和模型选择
since we observe individuals’ genotypes only at a discrete set of genetic markers and wish to consider positions between markers as the possible locations of QTL, we must use the marker genotype data to infer the genotypes at intervening locations. This is the missing data problem.

如何处理缺失数据的问题呢？
重组模型～
这其中最简单的是no crossover interference：recombination events in disjoint intervals are independent. 这样，染色体上就形成了一个马尔可夫链。也就是判断基因型的时候，只需要看它两侧的marker就可以，不需要考虑其他marker。

If riQ is the recombination fraction between marker Mi and the putative QTL, and if rij is the recombination fraction between markers Mi and Mj, then the probability that the individual has genotype AA at the putative QTL is (1−r3Q)(1−r4Q)/(1−r34); the probability that it is AB is r3Qr4Q/(1 − r34).

但是实际上，positive crossover interference是常见的。也就是不倾向在相邻位置发生重组。所以用到了hmm，这是R/qtl的核心内容。

3. R/qtl包流程

1）获得数据后，第一项任务是将它们组装成计算机文件或文件并将其导入软件。
2）其次，对数据执行各种诊断检查以识别可能的错误，例如数据输入错误，标记顺序错误和基因分型错误。
3）表型分布的方面可能导致人们考虑表型转化或使用特殊的表型模型
4）接下来，使用区间作图（或其变体之一），使用单QTL模型进行基因组扫描，以检测具有重要边际效应的基因座。
5）接下来转向基因组的二维扫描。这种二维扫描提供了识别QTL之间相互作用的机会，包括检测具有有限边际效应的QTL的可能性。
6）最后，我们将所有推定的QTL和QTL×QTL相互作用整合到一个整体的多QTL模型中

4 导入和模拟数据

#查看包的信息
library(help=qtl)
#load the dataset
data(hyper)
#see the summary
summary(hyper)
plot(hyper)
# save chr5-chr13 to a csv file
write.cross(hyper,"csv","~" ,c(5, 13))

The plot.missing function creates the plot with the pattern of missing geno- type data. It takes an argument reorder which can be used to order the individuals according to their phenotype. The genetic map is obtained with plot.map. The function plot.pheno plots a phenotype, either as a histogram (using the R function hist) or as a bar plot (using the R function barplot), depending on the nature of the phenotype.

模拟数据：只能是加性模型