GATK中如何计算Inbreeding coefficient（近交系数）

发表于 | 分类于 | | 阅读次数

cover

关于近交系数是什么的定义，除了英文资料，中文上也给出了清晰的定义，这里引用一下：

近交系数（inbreeding coefficient）是指根据近亲交配的世代数，将基因的纯化程度用百分数来表示即为近交系数，也指个体由于近交而造成异质基因减少时，同质基因或纯合子所占的百分比也叫近交系数，普遍以F或f来表示。

GATK近交系数的计算程序在github上可以找到：AS_InbreedingCoeff.java

代码不短，但计算很简单，我主要说展示一下这个计算的核心部分并在代码中做些注释，如下：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
protected double calculateIC(final VariantContext vc, final Allele altAllele) {
    final int AN = vc.getCalledChrCount();
    final double altAF;
    final double hetCount = heterozygosityUtils.getHetCount(vc, altAllele);
    final double F;
    //shortcut to get a value closer to the non-alleleSpecific value for bialleleics
    if (vc.isBiallelic()) {
        double refAC = heterozygosityUtils.getAlleleCount(vc, vc.getReference());
        double altAC = heterozygosityUtils.getAlleleCount(vc, altAllele);
        double refAF = refAC/(altAC+refAC);
        altAF = 1 - refAF;
        F = 1.0 - (hetCount / (2.0 * refAF * altAF * (double) heterozygosityUtils.getSampleCount())); // inbreeding coefficient
    } else {
        //compare number of hets for this allele (and any other second allele) with the expectation based on AFs
        //derive the altAF from the likelihoods to account for any accumulation of fractional counts from non-primary likelihoods,
        //e.g. for a GQ10 variant, the probability of the call will be ~0.9 and the second best call will be ~0.1 so adding up those 0.1s for het counts can dramatically change the AF compared with integer counts
        altAF = heterozygosityUtils.getAlleleCount(vc, altAllele)/ (double) AN;
        // 计算inbreeding coefficient
        F = 1.0 - (hetCount / (2.0 * (1 - altAF) * altAF * (double) heterozygosityUtils.getSampleCount())); // heterozygosityUtils.getSampleCount() 获取总样本数 
    }
    return F;
}

总的来说，是利用哈迪温伯格定律来计算的。 1.0 - (hetCount / (2.0 (1 - altAF) altAF(double)N ，N是人数。这个值给出的是期望的杂合变异的个数。所以参数F说的就是”实际的hetCount”除以”期望的hetCount”再与1.0取差。当F值越接近0，就意味着实际的hetCount与理论的hetCount越接近。