It is critical that great care is taken to prevent technical biases and confounding in sequencing to avoid distorting association results. For instance, differences in how (rare and precious) case samples are handled compared to control samples may lead to systematic false-positives that masquerade as interesting associations. Likewise, simultaneous multi-sample variant calling on only cases or only controls may lead to differential detection of variants across batches, negatively impacting the accuracy of the allele frequency estimates and association analyses. Many other, often poorly understood or hidden, technical confounders (e.g., DNA preparation, exome-capture technology, machine type, read length, depth of coverage, SNP calling algorithm, QC filters) may influence the properties of exome-sequencing data. Therefore, although the use of shared controls (e.g., from the 1000 Genomes project) has been helpful in “filtering” approaches applied to Mendelian disorders6, 9, it is not likely to be applicable to association analysis of complex diseases.
