danielapierro / blobtools_tutorial

Create a blobplot of a holobiont assembly without a reference genome. Example code and data are provided.

blobtools_tutorial

A practical guide to BlobTools

why use BlobTools?

BlobTools allows taxonomic partitioning of metagenome or metatranscriptome assemblies, without requiring reference genomes. Using GC content, read coverage, and taxonomy, it can visualize how much genetic information in the assembly comes from a host species versus members of its microbiome.

The example dataset here includes the metatranscriptome of an octocoral host and its associated Symbiodinium endosymbionts, during a heat stress experiment.

BlobTools was created by Laetsch DR and Blaxter ML, 2017. The current release is found here.

This guide follows "Workflow A" in the manual, providing example code and explanations at each step.

what input?

trimmed paired reads

After running Trimmomatic to trim the reads, rename the paired reads as follows.

First a base name, then an array number, then 1 (forward read) or 2 (reverse read).

This example will use two sets of trimmed paired reads:

coral_1_1.fq.gz
coral_1_2.fq.gz

coral_2_1.fq.gz
coral_2_2.fq.gz

Record which array numbers correspond to which experimental conditions. Here, coral_1 experiences control conditions, while coral_2 experiences thermal stress.

Then, run Trinity to assemble these trimmed paired reads, generating the assembly holobiont.fasta

This holobiont metatranscriptome assembly is expected to contain genetic information from the coral host and its endosymbionts.

This tutorial will use PBS scripts to allow for PBS arrays. Many of these commands can also be run as loops. Unless otherwise indicated, all PBS scripts below will have these flags:

#!/bin/bash
#PBS -V
#PBS -N job_name
#PBS -q batch
#PBS -S /bin/bash
#PBS -l nodes=1:ppn=8
#PBS -l mem=10G
#PBS -l walltime=48:00:00
#PBS -o /error_logs/job_name.out
#PBS -e /error_logs/job_name.err
#PBS -J 1-2

Burrows-Wheeler Alignment

run BWA-MEM to align trimmed paired reads to the assembly

index the assembly

bwa index holobiont.fasta

read mapping

map reads as an array in a PBS script that includes the following steps:

create an output folder

mkdir -p /alignments
cd /alignments

align the reads

bwa mem -t 8 -HM -k 19 -w 100 -d 100 -R "@RG\tID:${PBS_ARRAY_INDEX}\tSM:${PBS_ARRAY_INDEX}\tLB:${PBS_ARRAY_INDEX}\tPL:ILLUMINA"
holobiont.fasta
coral_${PBS_ARRAY_INDEX}_1.fq.gz
coral_${PBS_ARRAY_INDEX}_2.fq.gz
> coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.sam &&

convert sam to bam

samtools view -@ 8 -bS -o coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.sam_bam \
coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.sam &&

sort the bam files

samtools sort -m 1G -@ 8 coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.sam_bam 
-o coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.bam &&

index the bam files

samtools index coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.bam &&

remove all intermediate output files

rm *sam* &&

get statistics on the alignment

samtools flagstat coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.bam 
> coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.bam.flagstat

install BlobTools

Several methods for installing BlobTools can be found here

Upload blob.txt and use it to create a conda environment

conda create --name blobtools --file blob.txt

Install the dependencies

conda activate blobtools
conda install -n blobtools matplotlib 
conda install -n blobtools docopt 
conda install -n blobtools tqdm 
conda install -n blobtools wget 
conda install -n blobtools pyyaml 
conda install -n blobtools git
conda install -n blobtools pysam --update-deps 
conda install -n blobtools blobtools

you should recieve a message that all requested packages are installed.

note where the blobtools environment is located, for instance: /envs/blobtools

cd into the data directory, for instance: /envs/blobtools/lib/python3.9/site-packages/data

then get the taxdump

wget -c ftp://ftp.ncbi.nih.gov/pub/taxonomy/taxdump.tar.gz 
tar -zxvf taxdump.tar.gz

create nodesDB

blobtools nodesdb --nodes data/nodes.dmp --names data/names.dmp

the output file nodesDB.txt should be in the data/ directory

move the /data directory and all of its contents (including nodesDB) into the directory /blobtools

cp -r /home/conda/envs/blobtools/lib/python3.9/site-packages/data/ /home/conda/envs/blobtools/

coverage

generate coverage files from mapping individual bwamem alignments to the holobiont reference

cd /alignments/

for i in coral_*_toHolobiont_bwamem.bam;
do blobtools map2cov --infile holobiont.fasta
--bam $i --output $i; done

taxonomic annotation

blastn holobiont contigs to the nt database to find taxonomic hits

split the reference assembly holobiont.fasta into 10+ fasta files, using the script splitFASTA.pl

make the directory /split_holobiont

perl splitFASTA.pl -i holobiont.fasta -o /split_holobiont/ -s 10000

Run blastn as an array job.

Download the blast nt database locally (shown here) or if possible, use a remote search

Troubleshoot:
Specify #PBS -J 1-n such that n = the number of files that splitFASTA split the assembly into.
Make sure #PBS -l nodes=1:ppn=8 matches the number of threads in the blastn command -num_threads 8
Ensure that blast is updated to the most recent version with conda update blast

cd /fullpath/split_holobiont/

blastn -query holobiont-${PBS_ARRAY_INDEX}.fasta
-db /database/nt.oct.19.2021/nt -outfmt '6 qseqid staxids bitscore std'
-max_target_seqs 1 -max_hsps 1 -evalue 1e-25 -num_threads 8
-out ${PBS_ARRAY_INDEX}_to_nt.tab

In the split_holobiont directory, combine all blastn output files to one hits directory

cat *_to_nt.tab > holobiont_to_nt.tab

create a BlobTools database

run as an array make the output directory /blob_db

source ~/.bash_profile
conda activate blobtools

cd /fullpath/blob_db

blobtools create --infile holobiont.fasta
--hitsfile /fullpath/split_holobiont/holobiont_to_nt.tab
--cov /fullpath/alignments/coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.bam.cov
--bam /fullpath/alignments/coral_${PBS_ARRAY_INDEX}_toHolobiont_bwamem.bam
--out coral_${PBS_ARRAY_INDEX}_toHolobiont_blob.DB

the output databases are named coral_{array number}_toHolobiont_blob.DB.blobDB.json

create a summary table

change {1..n} depending on how many array numbers you have in total

for i in {1..4}; do blobtools view -i /fullpath/blob_db/coral_${i}_toHolobiont_blob.DB.blobDB.json 
-o coral_$i -x bestsum -r phylum; done

create a blobplot

create a blobplot of one individual file (for instance, array number 1)

blobtools plot -i coral_1_toHolobiont_blob.DB.blobDB.json 
--nohit --rank phylum --taxrule bestsum --format pdf --out coral1.phylum

create a blobplot for all files. change {1..n} depending on how many array numbers you have in total since this may take several hours, running it in a screen is highly recommended

for i in {1..4}; do blobtools plot -i /fullpath/blob_db/coral_${i}_toHolobiont_blob.DB.blobDB.json 
--nohit --rank phylum --taxrule bestsum --format pdf --out phylum_$i; done

In specifying taxonomic rank (--rank), note that BlobTools supports superkingdom, phylum, order, family, genus, and species, but does not support class, kingdom, or domain.

If working on a remote server, use rsync or scp to copy the pdf output onto your local device to see your blobplots!

good luck!