Statistical analysis of the GAMES studies

Abstract

The statistical analysis of the GAMES collaborative studies presented a number of novel challenges. Estimates of allele frequencies from pooled DNA samples are subject to additional sources of error over and above the multinomial sampling error which is assumed in classical chi-squared tests for association ( Barratt et al., 2002 Barratt B.J Payne F Rance H.E Nutland S Todd J.A Clayton D.G Identification of the sources of error in allele frequency estimations from pooled DNA indicated an optimal experimental design. Annals of Human Genetics. 2002; 66: 393-405 Crossref PubMed Scopus (120) Google Scholar , Sham et al., 2002 Sham P Bader J.S Craig I O'Donovan M Owen M DNA pooling: a tool for large-scale association studies. Nature Reviews. 2002; 3: 862-871 Crossref Scopus (462) Google Scholar ). Firstly, the assumption that a pool contains an equal amount of DNA from each subject will only be approximately true since quantification is subject to error. Secondly, amplification by PCR will introduce further random error. Finally, reading of concentrations of PCR products will also introduce additional errors. A fundamental difficulty is that, while reading PCR products was replicated in the design, allowing estimation of this source of error, neither pool formation nor PCR replication, was part of the standard GAMES design. It is, therefore, impossible to formally estimate the error variance attributable to each of these additional sources. For the same reason, it is not possible to estimate how the additional components of variance (and covariance) depend upon allele frequencies. As a result, formal significance tests are impossible.