Reproducibility with BWA-MEM2

[jarbet]

This is a notice that BWA-MEM2 alignment results may not be reproducible if the order of the reads within a fastq file is changed ¹ or if a different number of cpus are used when parallel computing².

All reads affected by the first issue are given quality scores of 0, and can be filtered out of downstream analyses if desired¹. For the 2nd issue, make sure to use the same # of CPUs for reproducibility.

Note the default # CPUs changes based on the type of node requested: so all else being equal, make sure to use the same node type for reproducibility. See the config files here for the default # CPUs by node type, for align_DNA_BWA_MEM2.

Footnotes

1. Firtina, Can, and Can Alkan. "On genomic repeats and reproducibility." Bioinformatics 32.15 (2016): 2243-2247. ↩ ↩²

2. Biostars ↩