---

Introduction

ADRSM (Ancient DNA Read Simulator for Metagenomics) is a tool designed to simulate the paired-end sequencing of a metagenomic community. ADRSM allows you to control precisely the amount of DNA from each organism in the community, which can be used, for example, to benchmark different metagenomics methods.

Requirements

• Conda

Installation

conda install -c maxibor adrsm

Usage

adrsm ./data/short_genome_list.csv

Output

• metagenome.{1,2}.fastq : Simulated paired end reads
• stats.csv : Statistics of simulated metagenome (organism, percentage of organism's DNA in metagenome)

Cite

You can cite ADRSM like this:

Maxime Borry (2018). ADRSM: Ancient DNA Read Simulator for Metagenomics. DOI: 10.5281/zenodo.1462743

Help

$ adrsm --help
Usage: adrsm [OPTIONS] CONFFILE

  ==================================================
  ADRSM: Ancient DNA Read Simulator for Metagenomics
  Author: Maxime Borry
  Contact: <borry[at]shh.mpg.de>
  Homepage & Documentation: github.com/maxibor/adrsm

  CONFFILE: path to ADRSM configuration file

Options:
  --version                    Show the version and exit.
  -r, --readLength INTEGER     Average read length  [default: 76]
  -n, --nbinom INTEGER         n parameter for Negative Binomial insert length
                               distribution  [default: 8]

  -fwd, --fwdAdapt TEXT        Forward adaptor sequence  [default: AGATCGGAAGA
                               GCACACGTCTGAACTCCAGTCACNNNNNNATCTCGTATGCCGTCTTC
                               TGCTTG]

  -rev, --revAdapt TEXT        Reverse adaptor sequence  [default: AGATCGGAAGA
                               GCGTCGTGTAGGGAAAGAGTGTAGATCTCGGTGGTCGCCGTATCATT
                               ]

  -p, --geom_p FLOAT RANGE     Geometric distribution parameter for
                               deamination  [default: 0.5]

  -m, --minD FLOAT RANGE       Deamination substitution base frequency
                               [default: 0.01]

  -M, --maxD FLOAT RANGE       Deamination substitution max frequency
                               [default: 0.3]

  -e, --effort INTEGER         Sequencing effort, maximum number of output reads to
                               be sampled from pool [default: 1000000]

  -s, --seed INTEGER           Seed for random generator generator  [default:
                               42]

  -t, --threads INTEGER RANGE  Number of threads for parallel processing
                               [default: 2]

  -o, --output PATH            Fastq output file basename  [default:
                               ./metagenome]

  -s, --stats PATH             Summary statistics file  [default: ./stats.csv]
  --help                       Show this message and exit.

Configuration file (confFile)

The configuration .csv file describes, one line per genome, the different simulation parameters :

• the file path to the genome fasta file - mandatory
• the mean insert size (integer) - mandatory
• the coverage (float) - mandatory
• the deamination (yes | no) - mandatory
• the mutation rate (0<float<1) - optional
• the age (integer) - optional

Example: short_genome_list.csv

genome(mandatory)	insert_size(mandatory)	coverage(mandatory)	deamination(mandatory)	mutation_rate(optional)	age(optional)
./data/genomes/Agrobacterium_tumefaciens.fa	47	0.1	yes	10e-8	10000
./data/genomes/Bacillus_anthracis.fa	48	0.2	no

Note on Coverage

Given the sequencing error, and the random choice of inserts, the target coverage might differ slightly from the real coverage (fig 1)

Figure 1: Coverage plot for simulated sequencing of Elephas maximus mitocondria. Aligned with Bowtie2 (default-parameters). Read-length = 76, insert-length = 200.

Note on Deamination simulation

The deamination is modeled using a Geometric distribution With the default parameters, the substitution frequency is depicted in fig 2:

One can try different parameters for deamination using this interactive plot: maxibor.github.io/adrsm

Figure 2: Substitution frequency.

For each nucleotide, a random number Pu is sampled from an uniform distribution (of support [0 ,1]) and compared to the corresponding value Pg of the rescaled geometric PMF at this nucleotide.
If Pg >= Pu, the base is substituted (fig 3).

Figure 3: Substitutions distribution along a DNA insert, with default parameters.

Note on Illumina base quality score

The base quality score (Qscore) is generated using a Markov chain from fastq template files.

Note on sequencing error

Until version 0.9.1, ADRSM simulated Illumina sequencing error with a uniform based model. From version 0.9.2 onwards, ADRSM simulates the sequencing error based on the QScore.

Note on sequencing effort

ADRSM initially generates reads based on the theoretical maximum from the depth coverage specified in the config file. In contrast, sequencers have a fixed number of 'slots' on a flowcell that DNA from all libraries in the sequencing pool will 'compete' for (e.g. approximate maximum of 300 million reads for a HiSeq lane). Therefore, in reality not all reads of a library are actually sequenced.

ADRSM simulates this by randomly subsampling the entire read pool generated at the maximum depth coverage, down to the number of reads provided to --effort.

Note on mutation

ADRSM can add mutations to your sequences. This allows to account for the evolutionary differences between ancient organisms and their reference genome counterparts present in today's databases.

ADRSM assumes two times more transitions than transversions.

There are two parameters for mutation simulation:

• The mutation rate (in bp/year): a good starting point is 10e-7 for bacteria
• The age (in years) of the organism