The readiator is a script to simulate RNA-Seq data. The basic idea is to take reference sequences and extract a number sub-sequence, with a requested read length, out of it. The tool needs a fastq-file from a real experiment to simulate sequencing errors. The script can simulate single end as well as paired end data sets with replicates. Additionally, it is possible to generate data sets that contain differentially expressed genes.

-f references.fa -fq fastq-file -l read length -r library size (default 1.000.000) -sz number of replicates -s seed (default random number between 1 and 1000) -t filter for min length of reference sequences that are considered (default 100) -o output directory (default simulated_fastq) -p to run a paired-end simulation -d simulate a data set that contains differentially expressed genes

Example:

single end data:

python3.7 readiator.py -f test_data_set/referenceSequences.fa -fq test_data_set/example.fastq -l 50 -r 10000 -sz 3 -s 5 -t 100

paired end data:

python3.7 readiator.py -f test_data_set/referenceSequences.fa -fq test_data_set/example.fastq -l 50 -r 10000 -sz 3 -s 5 -t 100 -p

Important is that the fastq file contains quality strings with a length at least as long as the requested read length

Let n_g be the number of kilo bases per gene g. such that

$$ \sum_i^{g} n_{g} = N $$

Thus, N is the number of kilo bases of all genes to be simulated, denoted as components in the following. We assume, that the first k components belong to the first q genes, i.e. the components are sorted with respect to the genes. Now, get a random dirichlet decomposition $D$ of all $N$ components with size 1. In the following, we sum up over the decomposition for each gene. Such that,

$$ d_g = \sum_{j=0}^{n_g} D[o+j] $$

and $d_g$ is the sum of the elements of the decompositions for all components of gene $g$. Since the components are ordered, the offset is defined as

$$ o = \sum_{k=0}^{g-1} n_g $$

such that $\sum D = 1$. Now, we have a $d_g$ for each gene and only need to multiply the value $d_g$ with the desired number of reads in the formulate, i.e. $10^6$ (-r).