MDR,Multifactor Dimensionality R
2018-10-15 本文已影响3人
wangchuang2017
MDR,Multifactor Dimensionality Reduction (多因子降维法) to detect&nb
1.candidate SNPs. <wbr>
2. 根据SNP combination 得到不同的genotype (diplotype)。
3.1. case-control样本: 计算每种genotype中 case:control 比, 大于threshold (通常选1) 记做high-risk, 小于threshold <wbr>记做low-risk.
3.2. family 样本:计算每种genotype-PDT statistic,大于大于threshold <wbr>(通常0)记做high-risk, 小于threshold <wbr>记做low-risk。genotype-PDT statistic计算:(比较复杂,见E.R Martin,2005) 1)对于parents & affected-sibling结构的family:主要因素是 #transmitted allele pairs - #nontransmitted allele pairs. 2)对于affected & unaffected sibling结构的family:主要是 #allele pairs in affected - <wbr>#allele pairs in unaffected
4.多种genotype 降维至high-risk 与low-risk两组。
5.计算best model。将样本随机分成N份,在N-1中计算model,在剩余1份进行验证。使用 <wbr>cross-validation n次以上,计算出accuracy 和 error。 permutation得到P值。