k-mer based base quality recalibration

This is a python prototype of kbbq; the C++ version can be found at https://www.github.com/adamjorr/kbbq

The full documentation can be found here: https://kbbq.readthedocs.io/

kbbq can be installed via pip or your favorite pip-compatible program, like Pipenv.

To install the latest release with pip::

pip install git+https://github.com/adamjorr/kbbq.git

kbbq uses a FASTQ file and a corrected FASTQ file to produce a recalibrated FASTQ file. The output file has the original read sequences but modified quality scores.

If the reads are in a file called reads.fq and reads.corrected.fq, you can recalibrate reads.fq to produce a reads.recalibrated.fq file with the recalibrate command like so:

kbbq recalibrate -f reads.fq reads.cor.fq > reads.recalibrated.fq

The FASTQ files must be interleaved. If you have read group information in the read names you can use the --infer-rg flag to infer which read group each read belongs to. You can read more about the naming conventions required to infer pairing and read group data here: FASTQ Input Help . Without --infer-rg, it is assumed all reads in the input file come from a single read group.

kbbq can benchmark quality scores given a BAM alignment, a fasta reference, a VCF, and optionally a BED file. Bases in the alignment that:

• are in a variable site or
• are outside the regions specified in the BED file or
• are soft clipped

are skipped. If a base is not skipped but does not match the reference, it is assumed to be an error.

The output file will be a tab separated file with 4 columns:

1. assigned quality scores that appear in the data set
2. the actual quality score of bases that were assigned the score
3. an optionally-provided label (modified with --label)
4. the number of bases that were assigned that quality score in the dataset

An example invokation is:

kbbq benchmark -b alignment.bam -r reference.fa -v variants.vcf -d good-sites.bed > calibration.tsv

and an example output is:

2       10      alignment.bam    6460879
3       3       alignment.bam    170777
4       3       alignment.bam    173821
5       5       alignment.bam    246709
6       5       alignment.bam    304227
7       6       alignment.bam    463211
8       7       alignment.bam    416902
9       8       alignment.bam    491355

A FASTQ file of reads can be provided with the -f flag if the reads in the file are properly named such that they can be found in the corresponding BAM file. The BAM file will be used to determine which bases are errors. For more detail on the required naming scheme, see FASTQ Input Help .

The output of the benchmark command can be plotted with the plot command. This makes a basic plot; the output type will be inferred by matplotlib based on the file extension. To plot the calibration of our file created above:

kbbq plot -o calibration.pdf calibration.tsv

You can use -t sample-size to plot the number of bases instead of the quality score calibration::

kbbq plot -t sample-size -o sample-size.pdf calibration.tsv

For further reading on program features, check out https://kbbq.readthedocs.io/en/latest/cli/cli.html.