Louis-MG/META-DIFF

This is the repository of the pipeline META-DIFF, which detects sequences in differential abundances between two conditions, and annotates them taxonomicaly and functionaly.

META-DIFF 🦠

This is the repository of the pipeline META-DIFF, which detects sequences in differential abundance between two conditions, and annotates them taxonomicaly and functionaly.

Motivation

Metagenomics becomes increasingly important in building our knowledge about microbes. Links between microbiome perturbations and diseases are regularily uncovered. Using kmer-based methods, this pipeline allows its users to quickly find microbial DNA sequences in differential abundances between two conditions (e.g. healthy and not healthy), and annotate them taxonomicaly and functionaly. The pipeline also isolates unclassified sequences and builds a predictive model based on the most significant unitigs' kmers abundances.

❗ Important note: the pipeline can predict genes and annotate functions for prokaryotes only. ❗

Workflow

The workflow is described by the following figure :

wiki ! 📚

Output

The output is divided in several key files:

• case/control alignement summaries:

37594334	Bacteroides ovatus strain 3725 D1 iv chromosome, complete genome
37565498	Bacteroides ovatus strain BFG-107 chromosome, complete genome
35771365	Bacteroides xylanisolvens strain CL11T00C03 chromosome, complete genome
35482957	Bacteroides xylanisolvens strain funn3 chromosome, complete genome
35106788	Bacteroides ovatus strain 2789STDY5834943, whole genome shotgun sequence
34744737	Bacteroides xylanisolvens strain H207 chromosome, complete genome
34093107	Bacteroides ovatus strain CL06T03C20 chromosome, complete genome
34079552	Bacteroides xylanisolvens strain CL11T00C41 chromosome, complete genome
33805332	Bacteroides ovatus strain FDAARGOS_733 chromosome, complete genome
33714506	Bacteroides ovatus isolate MGYG-HGUT-01378, whole genome shotgun sequence

• case/control functional annotation in the form of a barplot, table and a heatmap of pathways detected:

• performance of a classification model based on glmnet and the kmer counts of the most significant unclassified sequences:

best_model.txt:

alpha	lambda	Accuracy	Kappa	AccuracySD	KappaSD
1	0.012233193439917463	0.8071428571428572	0.6136666666666667	0.1277013475448753	0.2566471854078626

• table of unitigs to functions by condition. Each unitig is linked to the genes it contains and their funciton, KO number.

Unitig ID	Unitig seq	Gene ID	Translated Gene seq	Gene funciton	KO
Unitig1	ACGTCGCT	Gene1	ARDENE	Glucose transferase	K00001
Unitig1	ACGTCGCT	Gene2	WPH	Protease	K00004
Unitig2	GTCGATCATG	Gene1	IFPSY	Oxydase	K00761

Requirements

Check the wiki first, section Required files format. Wiki will soon contain information (like a script or something) to help building the files. Memory (RAM) needed will depend on the size of your alignment database. Disk space required mostly depends on the size of your dataset and databases. The number of kmers for 3To of CRC fasta files reached hundreds of millions, wich is about 500G of fasta files for the first step. Other steps will use less disk. The database of MicrobeAnnotator is about 690G.

Installation

Clone the repository:

git clone https://github.com/Louis-MG/META-DIFF.git

Get your functional database ready by following instructions at MicrobeAnnotator. Don't worry, it's just a few lines that take a while. Copy the path to the MicrobeAnnotator_DB in the snakemake/config.yaml file:

microbeannotator_db_path: "/path/to/MicrobeAnnotator_DB/"

Get your taxonomic database ready by looking at MMseqs2 documentation (look at the module createdb). Again, just a few lines and fasta files. Copy the path to your MMseqs2-formated taxonomic database in the the snakemake/config.yaml file:

taxonomic_db_path: "/path/to/mmseqs_DB/DB"

Usage

Build a file-of-files yourself or using the provided script:

bash kmdiff_fof_prep.sh --help

This script generates the file of file (fof.txt) for kmdiff. Its arguments are:
	--cases -c <PATH> path to the directory of cases samples.
	--controls -C <PATH> path to the directory of control samples.
	--output -o <PATH> path to where the fof should be.
	--help -h displays this help message and exits.

Output is :
	- a fof.txt named after the output parameter. The file is tab separated, format:
		control1: /path/to/control1_read1.fastq ; /path/to/control1_read2.fastq
		control2: /path/to/control2_read1.fastq ; /path/to/control2_read2.fastq
		case1: /path/to/case1_read1.fastq ; /path/to/case1_read2.fastq
		case2: /path/to/case2_read1.fastq ; /path/to/case2_read2.fastq

WARNING: depending of the denomination of your files for paired ends (_R1 and _R2, _1 and _2 ...), you will have to modify lines 55-56 and 60-61. Yeah it's annoying. Add the single-end by hand.

Add the last paths to ./snakemake/config.yaml:

# path to the file of correpsondance between seq headers and the genome name (strain etc): see wiki
seq_to_genome: "/path/to/ref.tsv"
# path to the file of correpsondance between seq headers and the species: see wiki
seq_to_species: "/path/to/detailed_ref.tsv"
# path to this repo, "META-DIFF/" inclueded:
src_path: "/path/to/META-DIFF/"
# where your results will be
project_path: "/path/to/your/project/"
# path to your file of file:
fof: "/path/to/your/fof.txt"

Finally, ssssstart the pipeline 🐍:

snakemake --cores X --use-conda

Issues

If you have any issues, let me know in the Issues space, with an informative title and description.

Citations

Coming soon ! 🎓