Simulated replicates from real data

[Chuang1118]

Dear seqgendiff team,

Thanks for your excellent work and build such useful package!

DE analysis tools generally recommend a minimum of 3 biological replicates per group for robust analysis.
In my case, 2 replicates per group. I would like use seqgendiff to generated 1 simulated replicates from real data.

I have a question for how to use the function thin_2group() or thin_diff() in practices.

Regarding to Applying Different Thinning Functions and Simulate RNA-seq Data from Real Data section in the tutorial, I'm not sure, I understood the parameters.

It's possible to generated 1 simulated replicate by permuting the expression levels e.g. 20%, adding a standard deviation of 0.25 ?

Which function can do the task ? thin_2group() or thin_diff()

thout <- thin_2group(mat = assay(airway), 
                     prop_null = 0.9,   # I mis-understand this parameter.
                     group_prop = 0.5, # mis-understand
                     signal_fun = stats::rnorm,
                     signal_params = list(mean = 0, sd = 0.8))

I have processed by thin_diff, but I am not sure the parameters.
input: 18789*4

                mock_1 mock_2 znf_1 znf_2
ENSG00000000419    355    263   586   284
ENSG00000000457     39     34    88    44
ENSG00000000460    118     60   198   108
ENSG00000000938      1      2     7     1
ENSG00000000971      7      2    40    12
ENSG00000001036      3      6     8     7

my code

n <- ncol(assay(dds)) 
p <- nrow(assay(dds))

## Choose simulation parameters
design_perm <- cbind(0:(n-1),rep(0.2, n)) # 20% permuting the expression levels
coef_perm <- matrix(rnorm(p * ncol(design_perm), sd = 0.25), nrow = p) # st == 0.25

## Thin
thout <- thin_diff(mat = assay(dds),
                   design_perm = design_perm,
                   coef_perm = coef_perm)

output: df

                mock_1 mock_2 znf_1 znf_2 sim_mock_1 sim_mock_2 sim_znf_1 sim_znf_2
ENSG00000000419    355    263   586   284        355        212       549       215
ENSG00000000457     39     34    88    44         33         31        72        44
ENSG00000000460    118     60   198   108        101         56       168       108
ENSG00000000938      1      2     7     1          1          1         7         1
ENSG00000000971      7      2    40    12          7          1        35        11
ENSG00000001036      3      6     8     7          2          6         7         7

I have checked there are 33.75379 % values motified by sim_mock_1.

sum(df[[1]] != df[[5]])/length(df[[1]])

I have used set.seed(1234), but I got the non-reproducible results.

Thanks for your help!
Chuang