decOM: Similarity-based microbial source tracking of ancient oral samples using k-mer-based methods

decOM is a high-accuracy microbial source tracking method that is suitable for contamination quantification in paleogenomics, namely the analysis of collections of possibly contaminated ancient oral metagenomic data sets. In few words, if you want to know how contaminated your ancient oral metagenomic sample is, this tool will help 🧹🦷🧹

In longer words decOM comes with a k-mer matrix of ancient Oral(aOral) metagenomic samples and its possible contaminants: Sediment/Soil, Skin and modern oral (mOral). It represents every sink the user inputs as a presence/absence vector of k-mers and then estimates and outputs the proportions of each source environment in the sink. Ex: K-mer AAACG is present in the input sink S and in source S1 labelled as Skin, S5 labelled as aOral and S7 labelled as mOral, hence one ball is added to the bin of Skin, aOral and mOral respectively. After every entry in the the sink vector is compared against every entry of every vector in the sources, decOM outputs the estimated environment proportions and the hard label assigned to the sink $ss$ is that of the environment with the highest contribution.

• System requirements
• Installation
• Before running decOM
• Test
• Output files
• Usage
• Additional features
  • aOralOut
  • MST
  • LOO
• Command-line options

System requirements

decOM has been developed and tested under a Linux environment and it only works in Linux-like systems. It requires certain packages/tools in order to be installed/used:

• miniconda3

Installation

Install decOM through conda:

git clone https://github.com/CamilaDuitama/decOM.git  
cd decOM  
conda env create -n decOM --file environment.yml  
conda deactivate
conda activate decOM  

To make the decOM command available, it is advised to include the absolute path of decOM in your PATH environment variable by adding the following line to your ~/.bashrc file:

export PATH=/absolute/path/to/decOM:${PATH}  

Before running decOM

Please make sure you read the ⚠️NOTE reminders on this README, as they might prevent you from running into errors when trying out our method

⚠️NOTE: BEFORE running decOM you must first download the folder decOM_sources.tar.gz and decompress it. You can either follow the link or use wget (it has to be installed in your computer first):

wget https://zenodo.org/record/6513520/files/decOM_sources.tar.gz  
tar -xf decOM_sources.tar.gz

⚠️NOTE: Make sure you have reading and writing permissions in the folder where decOM is running and where the input matrix of sources is located!

The path of the extracted directory decOM_sources will be requested through the input parameter -p_sources each time decOM is run (see examples below).

Test

Single sink

You can test if decOM is working by using the aOral sample present in the test/sample/ folder, ex: SRR13355807.

decOM -s SRR13355807 -p_sources decOM_sources/ -k tests/sample/SRR13355807.fof -mem 10GB -t 5 

⚠️NOTE: The final memory allocated for each run of decOM will be your input in -mem times the number of cores (-t). In the previous run we used 10GB * 5 = 50 GB. It is recommended to run regular decOM with at least 10GB of memory and 1 core. If you are testing your own k-mer matrix of sources, please try to allocate at least as much memory as the size (in GB) of your matrix.

Several sinks

You can test if decOM with several sinks by using the files inside test/several_samples/ as follows:

decOM -p_sources decOM_sources/ -p_sinks tests/several_samples/sinks.txt -p_keys tests/several_samples/ -mem 10GB -t 5 

decOM relies on DASK for parallelization. Once you start running decOM and the client is set, you can see the diagnostic dashboard to follow the process and better tune parameters such as -mem and -t, make sure you can connect to your local host and visualise it here: http://127.0.0.1:8787/status

Output files

decOM will output one .csv file with the k-mer counts and proportions, a folder with the vector representing the sample of interest, from now on called sink (s), and a barplot if indicated by the user.

decOM_output/  
├──decOM_output.csv 
├──result_plot_sinks.pdf 
├──result_plot_sinks.html 
├──{s}_vector/  

The decOM_output.csv file is a dataframe that contains one row per sink. The columns correspond to the raw number of k-mers per source environment, the running time per sink, the sink name and the proportions. The result for the one sample explained before should look like this:

Sediment/Soil	Skin	aOral	mOral	Unknown	Running time (s)	Sink	p_Sediment/Soil	p_Skin	p_aOral	p_mOral	p_Unknown	decOM_max
182	281	197859	37023	334	196.7268	SRR13355807	0.0772	0.1192	83.9527	15.7091	0.1417	aOral

The result_plot_sinks.pdf and result_plot_sinks.html are static and interactive plots (respectively) for the proportions of source environments per sink. The {s}_vector/ folder is the output of kmtricks filter + kmtricks aggregate.

UNKOWNS IS A FEATURE IN DEVELOPMENT!: The contribution of an unknown source is a feature in development. For a more accurate assessment of the contamination proportions in a sample, please take into account the raw counts for Sediment/Soil, Skin, aOral, mOral.

Usage

You can use as input your fastq/fasta file from your own experiment, you can download an ancient oral sample of interest from the AncientMetagenomeDir or from the SRA.
The users of decOM can represent their own metagenomic sample as a presence/absence vector of k-mers using kmtricks. This sink can be compared against the collection of sources we have put together.

Once you have downloaded the folder with the matrix of sources and the fastq file(s) of your sink(s), you have to create a key.fof file per sink.
The key.fof has one line of text depending on your type of data:

-Single-end:
s : path/to/file/s_1.fastq.gz

-Paired-end :
s : path/to/file/s_1.fastq.gz; path/to/file/s_2.fastq.gz

⚠️NOTE: As decOM relies on kmtricks, you might use a FASTA or FASTQ format, gzipped or not, which means you have to change the key.fof file accordingly.

Since you now have the fasta/fastq file of your sink, the folder with the matrix of sources and the key file, simply run decOM as follows:

Single sink

decOM -s {SINK} -p_sources decOM_sources/ -k {KEY.FOF} -mem {MEMORY} -t {THREADS}

The parameter -s refers to the name of the sink, whereas the parameter -k is the path to the file key.fof. Make sure that the name you use for your sink is the same as the name you use in the key.fof. So if your sink is called s, -s s_1, your key.fof should be s_1 : s_1.fastq

Several sinks

If you want to assess the contamination of several sinks, you need one key.fof file per sink, and they must be inside the folder p_sources

decOM -p_sinks {PATH_SINKS} -p_sources decOM_sources/ -p_keys {PATH_KEYS} -mem {MEMORY} -t {THREADS}

The parameter -p_sinks refers to a .txt file with a list of sinks limited by a newline (\n). Each sink in this file must have a corresponding key.fof file in the folder p_keys/ . While p_sinks is the path to a file, p_keys is a path to a folder that should contain as many key.fof files as sinks.

If you have n samples, your the content of the file p_sinks should be:

s_1
s_2
s_3
.
.
.
s_n 

And the content of p_keys should be:

p_keys/  
├──s_1.fof
├──s_2.fof
├──s_3.fof
.
.
.
├──s_n.fof 

Additional features

aOralOut:

This feature was thought for the users who prefered having mOral, Skin and Sediment/Soil samples as sources but no aOral samples. Once you download and unzip the new matrix of sources:

wget https://zenodo.org/record/6772124/files/aOralOut_sources.tar.gz
tar -xf aOralOut_sources.tar.gz

Simply run decOM-aOralOut as follows:

decOM-aOralOut -s SRR13355807 -p_sources aOralOut_sources/ -k tests/sample/SRR13355807.fof -mem 10GB -t 5 

MST:

This feature was thought for the users who prefer building their own k-mer matrix of sources. To run decOM-MST you need to create your own p_sources folder and additionally you need a -m or map file with one label per source.

To create the p_sources you can run kmtricks (already in your conda environment for decOM) as follows:

kmtricks pipeline --file kmtricks.fof --run-dir p_sources --mode kmer:pa:bin --restrict-to-list 1

kmtricks aggregate --run-dir p_sources --pa-matrix kmer --output p_sources/matrices/matrix.pa --format bin

kmtricks dump --run-dir p_sources --input p_sources/matrices/matrix.pa.lz4 -o p_sources/matrices/matrix.pa.txt

⚠️NOTE: The parameter –restrict-to-list can be changed to add more partitions to your final matrix, kmtricks will tell you how many partitions are expected in total. Ex: –restrict-to-list 1,2,3

⚠️NOTE: For a 10GB k-mer matrix, the expected resources of memory (–mem 10GB) and number of threads (-t 5) are enough to run in a couple of minutes. If your matrix is larger, consider enforcing also the parameter –nb_partitions (Ex: –nb_partitions 1000) or allocating more memory/nthreads to your job

You additionally need a -m file which is a .csv file of two columns: Env and SampleID. This is a x by 2 table, where x is the number of sources in your input k-mer matrix (number of columns in the kmtricks.fof used to run kmtricks). SampleID refers to the unique identifier of each source sample, and Env is the corresponding label for the source environment from where each sample was taken.

You can run decOM-MST with test data as follows:

kmtricks pipeline --file tests/MST/kmtricks.fof --run-dir p_sources --mode kmer:pa:bin --restrict-to-list 1

kmtricks aggregate --run-dir p_sources --pa-matrix kmer --output p_sources/matrices/matrix.pa --format bin 

kmtricks dump --run-dir p_sources/ --input p_sources/matrices/matrix.pa -o p_sources/matrices/matrix.pa.txt

decOM-MST -s SRR13355807 -p_sources p_sources/ -m tests/MST/map.csv -k tests/sample/SRR13355807.fof --mem 10GB -t 5

LOO:

This feature was thought for the users who would like to perform a leave-one-out experiment on their own matrix of sources. This experiment consist on creating a k-mer matrix of sources, extracting one sample to be used as sink and leave the reas out as sources. Once this iteration is finished, decOM-LOO would take out a different sink from the matrix of sources, and compare it against the rest of samples.decOM-LOO stops once every sample has been compared against the rest. To run decOM-LOO you need to create your own p_sources folder and additionally you need a -m or map file with one label per source.

⚠️NOTE:This feature will only work if there are at least 2 samples representative of each source environment

To create the p_sources you can run kmtricks (already in your conda environment for decOM) as follows:

kmtricks pipeline --file kmtricks.fof --run-dir p_sources --mode kmer:pa:bin 

kmtricks aggregate --run-dir p_sources --pa-matrix kmer --output p_sources/matrices/matrix.pa --format bin

kmtricks dump --run-dir p_sources --input p_sources/matrices/matrix.pa.lz4 -o p_sources/matrices/matrix.pa.txt

⚠️NOTE: Building a k-mer matrix with any other parameters of kmtricks and using it as input for decOM-LOO has not been tested.

You additionally need a -m file which is a .csv file of two columns: Env and SampleID. This is a x by 2 table, where x is the number of sources in your input k-mer matrix (number of columns in the kmtricks.fof used to run kmtricks). SampleID refers to the unique identifier of each source sample, and Env is the corresponding label for the source environment from where each sample was taken.

You can run decOM-LOO with test data as follows:

kmtricks pipeline --file tests/LOO/kmtricks.fof --run-dir p_sources --mode kmer:pa:bin

kmtricks aggregate --run-dir p_sources --pa-matrix kmer --output p_sources/matrices/matrix.pa --format bin 

kmtricks dump --run-dir p_sources/ --input p_sources/matrices/matrix.pa -o p_sources/matrices/matrix.pa.txt

decOM-LOO -p_sources p_sources/ -m tests/LOO/map.csv --mem 10GB -t 5

Command line options

usage: decOM [-h] (-s SINK | -p_sinks PATH_SINKS) -p_sources PATH_SOURCES (-k KEY | -p_keys PATH_KEYS) -mem MEMORY -t THREADS [-o OUTPUT] [-p {True,False}] [-V] [-v]

Microbial source tracking for contamination assessment of ancient oral samples using k-mer-based methods

Argurguments:

  -h, --help            show this help message and exit
  
  -s SINK, --sink SINK  Write down the name of your sink. It must be the same as the first element of key.fof. When this argument is set, -k/--key must be defined too

  -p_sinks PATH_SINKS, --path_sinks PATH_SINKS
                        .txt file with a list of sinks limited by a newline (\n). When this argument is set, -p_keys/--path_keys must be defined too.

  -p_sources PATH_SOURCES, --path_sources PATH_SOURCES
                        path to folder downloaded from https://zenodo.org/record/6513520/files/decOM_sources.tar.gz

  -k KEY, --key KEY     filtering key (a kmtricks fof with only one sample). When this argument is set, -s/--sink must be defined too.

  -p_keys PATH_KEYS, --path_keys PATH_KEYS
                        Path to folder with filtering keys (a kmtricks fof with only one sample).You should have as many .fof files as sinks.When this argument is set,
                        -p_sinks/--path_sinks must be defined too.

  -mem MEMORY, --memory MEMORY
                        Write down how much memory you want to use for this process. Ex: 10GB

  -t THREADS, --threads THREADS
                        Number of threads to use. Ex: 5

  -o OUTPUT, --output OUTPUT
                        Path to output folder, where you want decOM to write the results. Folder must not exist, it won't be overwritten.

  -p {True,False}, --plot {True,False}
                        True if you want a plot (in pdf and html format) with the source proportions of the sink, else False

  -V, --version         Show version number and exit

  -v, --verbose         Verbose output