摘要 : The ability to sequence mitochondrial genomes quickly and cheaply has led to an explosion in available mtDNA data. As a result, an expanding literature is exploring links between mtDNA features and susceptibility to, or prevalence... 展开 The ability to sequence mitochondrial genomes quickly and cheaply has led to an explosion in available mtDNA data. As a result, an expanding literature is exploring links between mtDNA features and susceptibility to, or prevalence of, a range of diseases. Unfortunately, this great technological power has not always been accompanied by great statistical responsibility. I will focus on one aspect of statistical analysis, multiple hypothesis correction, that is absolutely squired, yet often absolutely ignored, for responsible interpretation of this literature. Many existing studies perform comparisons between incidences of a large number (A/) of different mtDNA features and a given disease, reporting all those yielding p values under 0.05 as significant links. But when many comparisons are performed, it is highly likely that several pvalues under 0.05 will emerge, by chance, in the absence of any underlying link. A suitable correction (for example, Bonferroni correction, requiring p<0.05/N) must, therefore, be employed to avoid reporting false positive results. The absence of such corrections means that there is good reason to believe that many links reported between mtDNA features and variou tfceases are false; a state of affairs that is profoundly negative both for fundamental biology and for public health. I will show that statistics matching those claimed to illustrate significant links can arise, with a high probability, when no such link exists, and that these claims should thus be discarded until results of suitable statistical reliability are provided. I also discuss some strategies for responsible analysis and interpretation of this literature. 收起