I’ve been spending some time recently experimenting with forensic loci / STR matching from consumer whole genome sequencing data and ended up putting together a couple of small pipelines:
- https://github.com/SpikeTreeLab/forensic_loci
- https://github.com/SpikeTreeLab/hisat_genotype_forensic_loci
This started mostly as a learning project while exploring bioinformatics and trying to better understand sequencing limitations around forensic markers.
One thing I tested was comparing:
- WGS data from Sequencing.com
- against a CE profile from AlphaBiolabs
and I was able to recover 17/22 matching markers, which I thought was pretty interesting considering the differences between sequencing and CE approaches.
I’m definitely not claiming this is production-ready or validated forensic software — more an experimental workflow and learning exercise — but I’d genuinely appreciate feedback from people with experience in:
- STR analysis
- forensic genomics
- marker calling from short reads
- validation/QC approaches
- or existing tools/workflows I should look into
Happy to hear criticism as well if there are obvious methodological issues or pitfalls I’m missing.
•
written
by
Spike Tree •
1 answer
This is cool. I published a tool called STRprofiler a few years ago for comparing standardized STR profiles from PowerPlex and similar platforms, but didn't try to rip the loci from WGS directly. Well, I wanted to, but it was out of scope, and I didn't have the time anyhow.
It'd be interesting to see a more formalized benchmark against a handful of datasets, with some real example outputs. Tons of well-sequenced cell lines have published STR profiles that can be retrieved via Cellosaurus. The sensitivity of this approach would be important to assess prior to any real use.
A clearer list of the loci that it searches for would also be helpful. I'd also recommend trying to get it into a more robust workflow language, e.g. Nextflow. These would be a great standalone addition to nf-core.
Log in to answer this question.