Efficient study design for next generation sequencing

Abstract

Next Generation Sequencing represents a powerful tool for detecting genetic variation associated with human disease. Because of the high cost of this technology, it is critical that we develop efficient study designs that consider the trade-off between the number of subjects (n) and the coverage depth (µ). How we divide our resources between the two can greatly impact study success, particularly in pilot studies. We propose a strategy for selecting the optimal combination of n and µ for studies aimed at detecting rare variants and for studies aimed at detecting associations between rare or uncommon variants and disease. For detecting rare variants, we find the optimal coverage depth to be between 2 and 8 reads when using the likelihood ratio test. For association studies, we find the strategy of sequencing all available subjects to be preferable. In deriving these combinations, we provide a detailed analysis describing the distribution of depth across a genome and the depth needed to identify a minor allele in an individual. The optimal coverage depth depends on the aims of the study, and the chosen depth can have a large impact on study success.

Fig. 1

Comparing the observed distribution of coverage depth to the distribution estimated by the model. Graphs are for the three subjects, with the lowest, median, and highest, coverage depth, among the 116 individuals genotyped at the Sanger Center. The black dots show the true proportion of SNPs with the specified read depth (x-axis), the red/unbroken line shows the distribution of a negative binomial with that subject’s mean depth and ζ = 4, and the blue/dashed line shows the poisson distribution.

Fig. 2

(A) The power to detect a heterozygote individual as a function of average depth and α-level when r = 0.01, p_MA = 0.5, and assuming K_ij follows a negative binomial distribution. (B) The power to detect a heterozygote individual for different values of ζ (α = 10⁻⁵). Note the change in x-axis. (C) The power to detect a heterozygote individual for different values of p_MA, or different read biases (α = 10⁻⁵, ζ = 4).

Fig. 3

Main Figure: (A) The power to detect a rare variant for all possible combinations of n (number of subjects) and μ (depth of sequencing) using a likelihood ratio test, with λ_j~Poisson, r = 0.001, a≡ α = 0.0001, and MAF = 0.005. (B) The black/unbroken line is power as a function of μ when n × μ = 500 with the above parameters. The red/dashed and green/dotted lines show the μ/power relationships when r = 0.01 and a≡α = 0.01, respectively. (C) The black/unbroken line shows the optimal λ as a function of read error rate, with α = 0.0001, MAF = 0.005, n × μ = 500, and λ~Poisson. The red/dashed line shows the optimal λ when α is raised to 0.01.

Fig. 4

For fixed cost, the power to detect an association decreases with μ. The sharpness of the decline depends on the relative risk attributable to the SNP. The relationship between power and μ is illustrated for four different values of the relative risk when MAF = 0.05, r = 0.01, α = 0.0001, and n × μ = 50,000.