This vignette describes how to use the RAREsim R package to estimate the expected distribution of rare variants when simulating rare variant genetic data. The R functions within the package are described in more detail.
A complete simulation example with RAREsim can be found online at the RAREsim Example Code GitHub page.
The example below simulates a one cM block on chromosome 19 to match target data from the African ancestry group from gnomAD v2.1 (Karczewski, et al., 2020).
RAREsim has three main steps. Within the RAREsim R package here, we address Step #2.
Simulate genetic data with an abundance of rare variants using HAPGEN2 (Su, 2011). By simulating with HAPGEN2 default parameters and input haplotypes with information at all sequencing bases, including monomorphic sites, HAPGEN2 simulates an abundance of rare variants.
An example simulation with HAPGEN2 can be found on the RAREsim Example Code GitHub page. The simulation is done using HAPGEN2 default parameters and haplotype files with information at every sequencing base within the region of interest.
After haplotypes are simulated with HAPGEN2, all the bases that are monomorphic in the simulated data will be removed in the haplotype and legend files (see the example code).
A MAC file will be created - a file with a single column counting the number of alternate alleles for that variant. The MAC file will be compared with the expected number of variants per MAC bin in the pruning step.
Estimate the expected number of variants in minor allele count (MAC) bins. Users may fit target data, manually enter parameters, or use default parameters to estimate the number of variants per MAC bin.
The number of variants is estimated by combining the Number of Variants function and the Allele Frequency Spectrum (AFS) function.
Probabilistically prune the rare variants to match the estimated number of variants in each MAC bin. RAREsim prunes the simulated variants by returning all or a subset of alternate alleles back to reference.
To install RAREsim execute the following code or visit https://bioconductor.org/install/ for instructions.
The R package here is used for step (2), to estimate the number of rare variants per MAC bin.
For a given region, the Number of Variants function estimates the number of variants per Kb, fnvariant(n), for a sample size n. Estimating the number of variants can be achieved by 1) fitting target data to estimate parameters, 2) using default parameters, or 3) directly inputting parameters to the function. Additionally, a user may skip this function if they already have an estimate for the total number of variants expected in the region (e.g. 1000 variants).
Target data is used to estimate ϕ and ω to optimize the function fnvariant(n) = ϕnω to fit the target data.
The Number of Variants target data consists of various sample sizes (n) and the observed number of variants per Kb in the region of interest. Ancestry specific data is advised. Data should be formatted with the first column as the number of individuals (n) and the second column as the observed number of variants per Kb in the region of interest (perkb).
Here we will fit the example target data for the African ancestry population.
## n per_kb
## 10 0.2627568
## 20 0.6831678
## 50 1.5239897
## 100 2.7326712
## 200 4.3092123
## 500 7.6199485
## 1000 12.1919176
## 2000 19.3914551
## 3070 25.2246571
## 5000 33.4226707
## 5040 33.7905302
## 8128 45.1941773
The target data is used to estimate ϕ and ω within a least squares loss function, optimizing using sequential quadratic programming (SQP). This optimization is implemented via the fit_nvariant function.
## Warning in slsqp(tune, fn = leastsquares, hin = hin.tune, control =
## list(xtol_rel <- 1e-12)): The old behavior for hin >= 0 has been deprecated.
## Please restate the inequality to be <=0. The ability to use the old behavior
## will be removed in a future release.
## $phi
## [1] 0.1638108
##
## $omega
## [1] 0.6248848
The output of the fit_nvariant function are the parameters phi (ϕ) and omega (ω), respectively. The estimated parameters can then be used to determine the expected number of variants per Kb, given the number of individuals to be simulated, Nsim.
To visually check the fit of the function, we can plot the target data and the function.
plot(nvariant_afr$n, nvariant_afr$per_kb,
xlab = 'Sample Size', ylab = 'Variants per Kb')
lines(nvariant_afr$n, (nvar$phi*(nvariant_afr$n^nvar$omega)))
To simulate the sample size observed in the target data used here, (Nsim = 8128), we calculate $\hat{f_{nvariant}} (N_{sim}) = \hat{\phi} N_{sim}^{\hat{\omega}}$. This can be done with the nvariant function.
Parameter values estimated from the target data are used here, as well as the sample size.
## [1] 45.46027
RAREsim also provides ancestry specific default parameters for phi (ϕ) and omega (ω). To use the default parameters, the ancestry must be specified: African (AFR), East Asian (EAS), Non-Finnish European (NFE), or South Asian (SAS).
## [1] 43.66395
Finally, parameters can be directly input into the Number of Variants function.
## [1] 45.46026
The AFS function inputs a MAC (z) and outputs the proportion of variants at MAC = z, (fafs(z)). This is done by estimating α and β to optimize the function $f_{afs}(z) = \frac{b}{(\beta+z)^{\alpha}}$. Here b ensures that the sum of the individual rare allele count proportions equals the total proportion of rare variants, prv.
The AFS function inputs a data frame with the upper and lower boundaries for each bin and proportion of variants within each respective bin. The default bins used here and within the evaluation of RAREsim are:
MAC = 1
MAC = 2
MAC = 3 - 5
MAC = 6 - 10
MAC = 11 - 20
MAC = 21 - MAF = 0.5%
MAF = 0.5% - MAF = 1%
These are the recommended bins when simulated sample sizes above 3,500.
When simulating sample sizes between 2000 and 3500, the recommended MAC bins are:
MAC = 1
MAC = 2
MAC = 3 - 5
MAC = 6 - MAF = 0.25%
MAF = 0.25% - MAF = 0.5%
MAF = 0.5% - MAF = 1%
If a sample size below 2000 is desired, we recommend simulating 2000 individuals and taking a random sample to reach the desired sample size.
Estimating the AFS can be achieved by 1) fitting target data to estimate parameters, 2) using default parameters, or 3) directly inputting parameters to the function. Additionally, a user may skip this function if they already have an estimate for the proportion of variants in each rare MAC bin.
Here we will fit the example target data for the African ancestry population. The target data was obtained from the gnomAD v2.1 vcf files.
The first two columns in the target data identify the lower and upper boundaries of each MAC bin. The third column specifies the observed proportion of variants within each MAC bin in the target data.
## Lower Upper Prop
## 1 1 1 0.50257998
## 2 2 2 0.16305470
## 3 3 5 0.08255934
## 4 6 10 0.05882353
## 5 11 20 0.03715170
## 6 21 81 0.05675955
## 7 82 162 0.01754386
The fit_afs function estimates the parameters alpha (α), beta (β), and b.
## Warning in slsqp(tune, fn = calc_prob_LS, hin = hin.tune): The old behavior for
## hin >= 0 has been deprecated. Please restate the inequality to be <=0. The
## ability to use the old behavior will be removed in a future release.
## $alpha
## [1] 1.594622
##
## $beta
## [1] -0.2846474
##
## $b
## [1] 0.297495
##
## $Fitted_results
## Lower Upper Prop
## 1 1 1 0.50753380
## 2 2 2 0.12582725
## 3 3 5 0.12226962
## 4 6 10 0.06152310
## 5 11 20 0.04187244
## 6 21 81 0.04709594
## 7 82 162 0.01235050
The output includes the estimated parameters alpha (α), beta (β), and b, as well as the estimated proportion of variants in each MAC bin (Fitted_results).
To check how well the function is fitting, we can plot the target data with the fitted function.
# Label the target and fitted data
afs_afr$Data <- 'Target Data'
af$Fitted_results$Data <- 'Fitted Function'
# Combine into one data frame
af_all<-rbind(afs_afr, af$Fitted_results)
af_all$Data <- as.factor(af_all$Data)
#plot
ggplot(data = af_all)+
geom_point( mapping = aes(x= Lower, y= Prop, col = Data))+
labs(x = 'Lower Limit of MAC Bin',
y = 'Propoortion of Variants')
As with the Number of Variants function, default parameters can be used to estimate the parameters for the AFS function with the afs R function. As the default parameters are ancestry specific, the ancestry needs to be specified as pop = ‘AFR’, ‘EAS’, ‘NFE’, or ‘SAS’. The function requires a MAC bin dataframe with the bins specified.
The MAC data frame is the first two columns of the AFS target data.
Using the MAC bins as input and specifying an African ancestry, the default parameters are used to estimate the proportion of variants within each bin.
## Lower Upper Prop
## 1 1 1 0.51575451
## 2 2 2 0.12460586
## 3 3 5 0.12032463
## 4 6 10 0.06046027
## 5 11 20 0.04121670
## 6 21 81 0.04656603
## 7 82 162 0.01228838
The AFS function can also directly input the parameters alpha, beta, and b.
## Lower Upper Prop
## 1 1 1 0.50753386
## 2 2 2 0.12582721
## 3 3 5 0.12226954
## 4 6 10 0.06152304
## 5 11 20 0.04187238
## 6 21 81 0.04709586
## 7 82 162 0.01235047
Once the total number of variants (nvariant) and proportion of variants per bin (afs) have been estimated, the expected number of variants per MAC bin can be estimated. An example using the total number of varants and estimated proportion of variants per MAC bin is shown below.
The MAC bin boundaries should be defined based on the sample size that will be simulated.
Continuing our example, we will input the total number of variants we expect in the region (Total_num_var = 865.0634) and the estimated proportion of variants per MAC bin (af$Fitted_results).
bin_estimates <- expected_variants(Total_num_var = 865.0634, mac_bin_prop = af$Fitted_results)
print(bin_estimates)
## Lower Upper Data
## 1 1 1 Fitted Function
## 2 2 2 Fitted Function
## 3 3 5 Fitted Function
## 4 6 10 Fitted Function
## 5 11 20 Fitted Function
## 6 21 81 Fitted Function
## 7 82 162 Fitted Function
The output of the expected_variants function is the exected number of variants in each MAC bin within the simulation region. This output is input for the pruning function.
The Number of Variants and AFS function can also be calculated within the expected_variants function.
bin_estimates <- expected_variants(Total_num_var =
19.029*nvariant(phi = 0.1638108,
omega = 0.6248848, N = 8128),
mac_bin_prop = afs(mac_bins = mac, pop = 'AFR'))
print(bin_estimates)
## Lower Upper Expected_var
## 1 1 1 446.16029
## 2 2 2 107.79195
## 3 3 5 104.08842
## 4 6 10 52.30196
## 5 11 20 35.65505
## 6 21 81 40.28256
## 7 82 162 10.63023
The output from the expected_variants function will be used as input for simulations. See https://github.com/ryanlayer/raresim for more details.
## R version 4.4.1 (2024-06-14)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 24.04.1 LTS
##
## Matrix products: default
## BLAS: /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
## LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so; LAPACK version 3.12.0
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_US.UTF-8 LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## time zone: Etc/UTC
## tzcode source: system (glibc)
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] ggplot2_3.5.1 RAREsim_1.11.0 BiocStyle_2.35.0
##
## loaded via a namespace (and not attached):
## [1] gtable_0.3.6 jsonlite_1.8.9 compiler_4.4.1
## [4] BiocManager_1.30.25 highr_0.11 jquerylib_0.1.4
## [7] scales_1.3.0 yaml_2.3.10 fastmap_1.2.0
## [10] R6_2.5.1 labeling_0.4.3 knitr_1.48
## [13] tibble_3.2.1 nloptr_2.1.1 maketools_1.3.1
## [16] munsell_0.5.1 bslib_0.8.0 pillar_1.9.0
## [19] rlang_1.1.4 utf8_1.2.4 cachem_1.1.0
## [22] xfun_0.48 sass_0.4.9 sys_3.4.3
## [25] cli_3.6.3 withr_3.0.2 magrittr_2.0.3
## [28] digest_0.6.37 grid_4.4.1 lifecycle_1.0.4
## [31] vctrs_0.6.5 evaluate_1.0.1 glue_1.8.0
## [34] farver_2.1.2 buildtools_1.0.0 fansi_1.0.6
## [37] colorspace_2.1-1 rmarkdown_2.28 tools_4.4.1
## [40] pkgconfig_2.0.3 htmltools_0.5.8.1