pathogen_detection_example.md

Pathogen Detection Example

A patient with respiratory symptoms seeks to find out the pathogen that is causing them these symptoms.

Make sure all bacterial reads needed to create your reference dataset also known as a training dataset are available.

bash 1_before_starting.sh

bash 2_before_starting.sh

Sketch your training dataset and sample to your preference.

Using k=31

Note: training and sample datasets are required to have the same ksize. Please note that since we are sketching from a list of genomes. We can use the following sourmash sketch command:

sourmash sketch fromfile genome_list.csv -p dna,k=31,scaled=1000,abund -o training_database.k31.sig.zip

Sketch your sample fasta file

yacht sketch sample --infile ./lung_sample.fasta --kmer 31 --scaled 1000 --outfile lung_sample.k31.sig.zip

Make training data for k=31

yacht train --ref_file training_database.k31.sig.zip --ksize 31 --num_threads 64 --ani_thresh 0.95 --prefix 'training_database.k31' --outdir ./ --force

Identify whether the patient has a infection and what pathogen is causing the disease.

yacht run --json training_database.k31_config.json --sample_file lung_sample.k31.sig.zip --significance 0.99 --num_threads 64 --min_coverage_list 1 0.6 0.2 0.1 --out ./k31_result.xlsx

Results

Using a ksize of 31, YACHT finds that M. pneumoniae is present in the lung sample.

What if we decrease ksize to 15?

If we use small ksizes like 15, we would expect to not find that the patient is infected by M. pneumoniae. Let's set up the experiment. Note that a ksize below 7 may not produce results and is not recommend.

Sketch Lung Sample using a k=15

sourmash sketch fromfile genome_list.csv -p dna,k=15,scaled=1000,abund -o training_database.k15.sig.zip

Sketch your sample fasta file

yacht sketch sample --infile ./lung_sample.fasta --kmer 15 --scaled 1000 --outfile lung_sample.k15.sig.zip

Make training data for k=15

yacht train --ref_file training_database.k15.sig.zip --ksize 15 --num_threads 64 --ani_thresh 0.95 --prefix 'training_database.k15' --outdir ./ --force

Pathogen Detection using YACHT

Identify whether the patient has a infectin and what pathogen is causing the disease.

yacht run --json training_database.k15_config.json --sample_file lung_sample.k15.sig.zip --significance 0.99 --num_threads 64 --min_coverage_list 1 0.6 0.2 0.1 --out ./k15_result.xlsx

Results

Using a ksize of 15, YACHT finds/does not fine that M. pneumoniae

Let's decrease ANI to 0.85

Make training data for k=15

yacht train --ref_file training_database.k15.sig.zip --ksize 15 --num_threads 64 --ani_thresh 0.85 --prefix 'training_database.k15_ani0.85' --outdir ./ --force

Pathogen Detection using YACHT

Identify whether the patient has a infectin and what pathogen is causing the disease.

yacht run --json training_database.k15_ani0.85_config.json --sample_file lung_sample.k15.sig.zip --significance 0.99 --num_threads 64 --min_coverage_list 1 0.6 0.2 0.1 --out ./k15_ani0.85_result.xlsx

Results

Using a ksize of 15, YACHT finds/does not fine that M. pneumoniae