A Sequential Test Algorithm for DNA Pooling/ Bootstrap-Based Studies

Abstract

DNA pooling is a practical way to reduce the cost of large-scale association studies to identify susceptibility loci. In contrast to individual genotyping, in DNA pooling we can combine samples from N cases and M controls into two single pooled sample tests and estimate the allele frequency of hundred of thousand SNPs using high-throughput genotyping technologies. Selection of disease-associated SNPs is performed based on their P-value after a statistical test has been run. For the subset of those SNPs determined to be significant, all individuals are then genotyped to corroborate results from the pooled samples. The strategy described above has been successfully used for years. However, there are situations in which either it is not possible to recruit the number of patients (cases) needed based on power estimations or replication of DNA pools is important. In both situations, the limitation is the availability of DNA samples. When using sequential testing, formally presented in 1945 as sequential probability ratio test (SPRT), units of interest are sequentially included as they are generated, and a statistical test is run at every stage. In the context of DNA pooling, units would be represented by the pooled samples coming from each group of cases and controls while using a bootstrapping re-sampling strategy. We have developed a SPRT algorithm for identifying diseaseassociated SNPs when comparing cases and controls via DNA pooling that is at least as powerful as the strategy previously described, but that needs less DNA samples than the strategy described above. Along with our algorithm, we also provide a way to estimate the number of stages needed, i.e., the number of SNP-chip pairs to stop the algorithm a hieving a desired significance and power levels. We illustrate how our approach works using GWAS on Oppositional Conduct Disorder (OCD) and Attention Deficit Hyperactivity Disorder (ADHD).