This function creates a synthetic limited proteolysis proteomics dataset that can be used to test functions while knowing the ground truth.

create_synthetic_data(
  n_proteins,
  frac_change,
  n_replicates,
  n_conditions,
  method = "effect_random",
  concentrations = NULL,
  median_offset_sd = 0.05,
  mean_protein_intensity = 16.88,
  sd_protein_intensity = 1.4,
  mean_n_peptides = 12.75,
  size_n_peptides = 0.9,
  mean_sd_peptides = 1.7,
  sd_sd_peptides = 0.75,
  mean_log_replicates = -2.2,
  sd_log_replicates = 1.05,
  effect_sd = 2,
  dropout_curve_inflection = 14,
  dropout_curve_sd = -1.2,
  additional_metadata = TRUE
)

Arguments

n_proteins

a numeric value that specifies the number of proteins in the synthetic dataset.

frac_change

a numeric value that specifies the fraction of proteins that has a peptide changing in abundance. So far only one peptide per protein is changing.

n_replicates

a numeric value that specifies the number of replicates per condition.

n_conditions

a numeric value that specifies the number of conditions.

method

a character value that specifies the method type for the random sampling of significantly changing peptides. If method = "effect_random", the effect for each condition is randomly sampled and conditions do not depend on each other. If method = "dose_response", the effect is sampled based on a dose response curve and conditions are related to each other depending on the curve shape. In this case the concentrations argument needs to be specified.

concentrations

a numeric vector of length equal to the number of conditions, only needs to be specified if method = "dose_response". This allows equal sampling of peptide intensities. It ensures that the same positions of dose response curves are sampled for each peptide based on the provided concentrations.

median_offset_sd

a numeric value that specifies the standard deviation of normal distribution that is used for sampling of inter-sample-differences. Default is 0.05.

mean_protein_intensity

a numeric value that specifies the mean of the protein intensity distribution. Default: 16.8.

sd_protein_intensity

a numeric value that specifies the standard deviation of the protein intensity distribution. Default: 1.4.

mean_n_peptides

a numeric value that specifies the mean number of peptides per protein. Default: 12.75.

size_n_peptides

a numeric value that specifies the dispersion parameter (the shape parameter of the gamma mixing distribution). Can be theoretically calculated as mean + mean^2/variance, however, it should be rather obtained by fitting the negative binomial distribution to real data. This can be done by using the optim function (see Example section). Default: 0.9.

mean_sd_peptides

a numeric value that specifies the mean of peptide intensity standard deviations within a protein. Default: 1.7.

sd_sd_peptides

a numeric value that specifies the standard deviation of peptide intensity standard deviation within a protein. Default: 0.75.

mean_log_replicates, sd_log_replicates

a numeric value that specifies the meanlog and sdlog of the log normal distribution of replicate standard deviations. Can be obtained by fitting a log normal distribution to the distribution of replicate standard deviations from a real dataset. This can be done using the optim function (see Example section). Default: -2.2 and 1.05.

effect_sd

a numeric value that specifies the standard deviation of a normal distribution around mean = 0 that is used to sample the effect of significantly changeing peptides. Default: 2.

dropout_curve_inflection

a numeric value that specifies the intensity inflection point of a probabilistic dropout curve that is used to sample intensity dependent missing values. This argument determines how many missing values there are in the dataset. Default: 14.

dropout_curve_sd

a numeric value that specifies the standard deviation of the probabilistic dropout curve. Needs to be negative to sample a droupout towards low intensities. Default: -1.2.

additional_metadata

a logical value that determines if metadata such as protein coverage, missed cleavages and charge state should be sampled and added to the list.

Value

A data frame that contains complete peptide intensities and peptide intensities with values that were created based on a probabilistic dropout curve.

Examples

create_synthetic_data(
  n_proteins = 10,
  frac_change = 0.1,
  n_replicates = 3,
  n_conditions = 2
)
#> # A tibble: 1,632 × 14
#>    protein   peptide    condition sample peptide_intensity change change_peptide
#>    <chr>     <chr>      <chr>     <chr>              <dbl> <lgl>  <lgl>         
#>  1 protein_1 peptide_1… conditio… sampl…              17.5 TRUE   TRUE          
#>  2 protein_1 peptide_1… conditio… sampl…              17.6 TRUE   TRUE          
#>  3 protein_1 peptide_1… conditio… sampl…              17.7 TRUE   TRUE          
#>  4 protein_1 peptide_1… conditio… sampl…              15.9 TRUE   TRUE          
#>  5 protein_1 peptide_1… conditio… sampl…              15.9 TRUE   TRUE          
#>  6 protein_1 peptide_1… conditio… sampl…              15.9 TRUE   TRUE          
#>  7 protein_1 peptide_1… conditio… sampl…              21.2 TRUE   TRUE          
#>  8 protein_1 peptide_1… conditio… sampl…              21.3 TRUE   TRUE          
#>  9 protein_1 peptide_1… conditio… sampl…              21.3 TRUE   TRUE          
#> 10 protein_1 peptide_1… conditio… sampl…              17.7 TRUE   TRUE          
#> # ℹ 1,622 more rows
#> # ℹ 7 more variables: peptide_intensity_missing <dbl>, coverage <dbl>,
#> #   n_missed_cleavage <int>, charge <dbl>, pep_type <chr>, peak_width <dbl>,
#> #   retention_time <dbl>

# determination of mean_n_peptides and size_n_peptides parameters based on real data (count)
# example peptide count per protein
count <- c(6, 3, 2, 0, 1, 0, 1, 2, 2, 0)
theta <- c(mu = 1, k = 1)
negbinom <- function(theta) {
  -sum(stats::dnbinom(count, mu = theta[1], size = theta[2], log = TRUE))
}
fit <- stats::optim(theta, negbinom)
fit
#> $par
#>       mu        k 
#> 1.699882 2.124010 
#> 
#> $value
#> [1] 17.50891
#> 
#> $counts
#> function gradient 
#>       57       NA 
#> 
#> $convergence
#> [1] 0
#> 
#> $message
#> NULL
#> 

# determination of mean_log_replicates and sd_log_replicates parameters
# based on real data (standard_deviations)

# example standard deviations of replicates
standard_deviations <- c(0.61, 0.54, 0.2, 1.2, 0.8, 0.3, 0.2, 0.6)
theta2 <- c(meanlog = 1, sdlog = 1)
lognorm <- function(theta2) {
  -sum(stats::dlnorm(standard_deviations, meanlog = theta2[1], sdlog = theta2[2], log = TRUE))
}
fit2 <- stats::optim(theta2, lognorm)
fit2
#> $par
#>    meanlog      sdlog 
#> -0.7606984  0.6093069 
#> 
#> $value
#> [1] 1.302677
#> 
#> $counts
#> function gradient 
#>       75       NA 
#> 
#> $convergence
#> [1] 0
#> 
#> $message
#> NULL
#>