SNP calling

[hollygene]

Using GATK Best Practices

[hollygene]

https://github.com/hollygene/CornellPostdoc/blob/9dfb02780e0c640dcc6758e847f5b5265adaea64/variantGATK.sh

[hollygene]

Step 1: Produce file of accession numbers from all of the sequences we have phenotypic data for

https://github.com/hollygene/CornellPostdoc/blob/9dfb02780e0c640dcc6758e847f5b5265adaea64/misc_scripts.sh#L11

[hollygene]

Step 2: Download the sequences from NCBI

Using: https://github.com/hollygene/CornellPostdoc/blob/9dfb02780e0c640dcc6758e847f5b5265adaea64/get_SRR_data.sh#L4-L9

[hollygene]

Step 3: Create unmapped bams from fastqs

Using: https://github.com/hollygene/CornellPostdoc/blob/9dfb02780e0c640dcc6758e847f5b5265adaea64/variantGATK.sh#L12-L28

[hollygene]

Step 4: Mark Illumina adapters

https://github.com/hollygene/CornellPostdoc/blob/41c55b7b4f4e5378a4465d70d1dd5ba0bf2e836f/variantGATK.sh#L36-L53

[hollygene]

Step 5: Validate Sam File

https://github.com/hollygene/CornellPostdoc/blob/41c55b7b4f4e5378a4465d70d1dd5ba0bf2e836f/variantGATK.sh#L61-L72

[hollygene]

Step 6: Convert from Sam to Fastq

https://github.com/hollygene/CornellPostdoc/blob/41c55b7b4f4e5378a4465d70d1dd5ba0bf2e836f/variantGATK.sh#L78-L94

[hollygene]

Need a reference genome for next step - asking Stanhope/lab slack for recommendations on which version/strain to use

[hollygene]

Stanhope recommends using the PANTHER database reference genome

Escherichia coli | E. coli | ECOLI | EnsemblGenome | Reference Proteome 2020_04

https://www.ebi.ac.uk/reference_proteomes/

the E coli reference:

ftp://ftp.ebi.ac.uk/pub/databases/reference_proteomes/QfO/Bacteria/UP000000625_83333.fasta.gz

[hollygene]

^ that is actually a proteome so gatk didn't work

Need a GENOME: ftp://ftp.ebi.ac.uk/pub/databases/reference_proteomes/QfO/

• this opens a folder in Finder
• click through Bacteria to get to UDP 83333 (E coli)
• choose the "DNA" fasta to download

[hollygene]

How to pick the best reference genome?

We are wanting to find SNPs that are associated with particular resistance phenotypes So ideally the reference would not be resistant to any abx A true "wild type" genome

However, we could also use a consensus sequence and call SNPs in samples from that Pros: no need for a possibly very diverged reference sequence Cons: Our dataset is biased because a lot of them are resistant

• could maybe use one of the dog isolates that ISN'T resistant?

[hollygene]

WDL notes WDL: script that describes the workflow Cromwell: Java-based job scheduler that can use various backend environments Run mode & server mode

[hollygene]

Dockstore info page Descriptor file: script in wdl that tells the program what to do, basically tools: more info on each task + Docker container it is using for that task test parameters file: file with the input files (can make this manually) Launch tab: actual commands for running the file

• Run locally with Dockstore CLI (can run on your local machine command line)

[hollygene]

Decided to choose a reference genome that is from a canine

Genome chosen: https://www.ncbi.nlm.nih.gov/assembly/GCA_002310695.1#/def From: https://www.ncbi.nlm.nih.gov/genome/browse#!/prokaryotes/167/ (filtered by host organism + complete genome) Strain: 1428 1 chromosome, 4 plasmids

AMR Genotypes: complete: acrF, blaCMY-2, blaEC,mdtM,tet(B) point: cyA_S352T

[hollygene]

Creating unmapped bams from fastq files

for file in ${raw_data}/*_1.fastq

do

FBASE=$(basename $file _1.fastq)
BASE=${FBASE%_1.fastq}
java -jar /programs/picard-tools-2.19.2/picard.jar FastqToSam \
    FASTQ=${raw_data}/${BASE}_1.fastq \
    FASTQ2=${raw_data}/${BASE}_2.fastq  \
    OUTPUT=${unmapped_bams}/${BASE}_fastqtosam.bam \
    READ_GROUP_NAME=${BASE} \
    SAMPLE_NAME=${BASE}

done