25-Apr-2024

How to use threshold-based DNA software on low-level mixture evidence

Our JFS Letter's threshold analysis above raises an interesting question. When can STRmix and other data-discarding threshold PG software be safely used on low-level DNA mixtures?

In both the Letter’s California case and New York v. Hillary, varying the threshold changed STRmix LR values a million-fold (see table). The STRmix results that were reported at only one threshold level were unhelpful to the defendants. Yet at lower thresholds (that used more data) other STRmix LR values were exculpatory. STRmix contradicted itself.

TrueAllele, which has no threshold parameter, simply used all the data to accurately exclude the defendants from the DNA evidence. No threshold, no contradiction. Just science.

Clearly one preset threshold is not enough. STRmix needs to have a threshold sensitivity study conducted whenever it is used on low-level DNA mixtures. A court needs to know how changing STRmix input data affects its LR information output.

Disclosing only one of many self-contradictory results is unfairly prejudicial, potentially confusing, a waste of time, and can mislead a jury. A STRmix prosecutor or defender should reveal the full spectrum of LR results over a range of threshold parameter values. Or contact Cybergenetics for an accurate TrueAllele answer.