Skip to content

Latest commit

 

History

History
186 lines (103 loc) · 5.39 KB

README.md

File metadata and controls

186 lines (103 loc) · 5.39 KB
Raw

pathoFind

Author: Priya Lakra

Identifying non-human reads in human NGS data

This is the pipeline for identyfing non-human reads in human NGS data. The workflow contains three steps including substratcion of human reads, identification of non-human reads, and analysis.

OS and softwares used:

System details: HPC cluster; OS= Centos 7

Softwares Version
fastqc 0.11.9
multiqc 1.10
trimmomatic 0.39
bowtie2 2.3.5.1
STAR 2.7.8a
NCBI-Blast+ 2.11.0
Python 3.7.10
samtools 1.9

All scripts are in the script folder!! 🤓

Create your working enviornment

conda create --name env

Activate your working enviornment

conda activate env

1. Substratcion of human reads

1.1. Quality check using fastqc, multiqc

bash scripts/preprocess/pl_qc.sh [options]

1.2. Trim adapters

Change trimmomatic parameters depending on your fastq files, and perform fastqc also on trimmed output files

bash scripts/preprocess/pl_trim.sh [options]

1.3a. For RNAseq data, STAR aligner is used https://physiology.med.cornell.edu/faculty/skrabanek/lab/angsd/lecture_notes/STARmanual.pdf

1.3b. For WGS data, BWA/bowtie2 aligner is used

Mapped reads from step1.3a,b can be used for their respective analysis. Unmapped reads from step1.3a,b will be the input for step2 of the workflow.

2. Identification of non-human reads

It can be achieved via two ways, one involving Bowtie2 aligner and one involving NCBI-BLAST+

2.1. Creating bowtie2 index

2.1a. For bacteria, fungi, and viruses, sequences were taken from [NCBI-RefSeq] (ftp://ftp.ncbi.nlm.nih.gov/genomes/refseq/) genomic sequences database

The script makes a text file (genomic_file) containing ftp links of all bacterial RefSeq sequences and download all sequences

bash scripts/pathoDatabse/pl_bacgendown.sh [path_name]

Similarly, download fungal and viral sequences

bash scripts/pathoDatabse/pl_fungendown.sh [path_name]

mkdir viral
wget ftp://ftp.ncbi.nlm.nih.gov/refseq/release/viral/*.fna.gz 
gzip *.fna (optional)

Concatenate files to create a single fasta file containing all sequences

cat *.fna > in_file.fna

Downloading pathogenic bacterial sequences from PATRIC [optional]

mkdir patric_pathogens

for i in `cat patric_genomes2.txt`; do wget -qN "ftp://ftp.patricbrc.org/genomes/$i/$i.fna"; done

2.1b. Formatting fasta header name (optional):

for i in *.fna; do awk -F '[/^> ]' 'NF>1 {print ">"$2"."$3"."$4"."$5"."$6"."$7"."$8;next}{print$1}' ${i} > ${i}_header_edited.fna; done`

2.2. Making reference genome index

qsub scripts/pathoDatabse/pl_indexgen.sh

qsub command is used here for submitting the job script to HPC.

Note: Here I generated index for ~100Gb fasta file using bowtie2-build. The process took ~20 hours on HPC cluster. Following similar steps make bowtie2 human genome index.

2.3. Align with bowtie2 and processing output sam files

Take output files from 1.3a,b and align against human genome with bowtie2. Extract the unaligned reads and then align with pathogen database.

bash scripts/bowtie/pl_NGS_process.sh [options]

3. Output analysis

Note: Rename files for the ease of processing. Example script in scripts/bowtie/preprocessFiles

Output analysis include calculating genome coverage for each organism.

Genome coverage = (read length * number of reads)/genome size

Note: For read lengths, I took median read length following these calculations:

count read lengths in sam or bam files

for i in *.out.sorted.bam; do samtools view -F 4 ${i} | awk '{print length($10)}' > /nfs_master/priya/nonhuman_out/processed_files/read_len/${i}.len.txt; done
a <- read.csv('len.txt', header=FALSE)
colnames(a) <- lengths # set col name as lengths 
mean(a$lengths)
median(a$lengths)
shapiro.test(a$lengths[1:5000])

Output analysis is performed using pandas package of python!

python scripts/bowtie/pl_bowtie_output.py

Visualization is done using R!

install.packages("pheatmap")
library(pheatmap)

fil <- read.csv("filename.csv", sep=",")

# replace NA with zeros (optional)
fil[is.na(fil)] = 0

# select columns with numbers only to make heatmap
filn <- as.matrix(fil[ ,3:33])

# change row names to the organim names (optional)
rownames(filn) <- fil$Organism 

# scale values in the matrix (optional)
filn_s <- scale(filn)

# plotting heatmap
pheatmap(filn_s)
pheatmap(filn)

Scripts for aligning reads from NGS data using BLAST.

4: Aligning against pathogen database using NCBI-BLAST+

4.1. Downloading standalone NCBI-BLAST+

wget https://ftp.ncbi.nlm.nih.gov/blast/executables/LATEST/ncbi-blast-2.11.0+-x64-linux.tar.gz
tar -zxvpf ncbi-blast-2.11.0+-x64-linux.tar.gz

4.2. Making BLAST database

bash scripts/pathoDatabse/pl_blastdb.sh [options]

blastdbcheck -db <database_name> -dbtype nucl -must_have_taxids

One can also mask low complexity regions before making database. Check.

4.3. BLASTn search

qsub -v dbdir=/nfs_node1/priya/blast_database,dir=/nfs_node1/priya/blast_trial pl_blastn.sh

5. Output analysis

For outfmt 7:

5.1. Extract hits with a threshold percentage identity, specific organism, and add genome size.

bash scripts/blast/pl_blastout_analysis.sh [options]

5.2. Calculate genome coverage for each organism. Similar to step 3.

python scripts/blast/pl_blast_output.py

Visualization is done using R!