Building software for probabilistic genotyping is technically demanding. Building it so it holds up in court systems across multiple countries is something different. That distinction drove much of how STRmix™ was designed.
Dr. Jo Bright is a Senior Science Leader at the New Zealand Institute for Public Health and Forensic Science, formerly known as ESR. She leads the STRmix™ team, which is responsible for the development and training of the probabilistic genotyping software used in forensic laboratories internationally. STRmix™ combines probabilistic techniques and biological models to assess all possible genotype combinations that might describe a DNA profile, and to calculate a match statistic in the form of a likelihood ratio when a person of interest is identified.
Bright presented at ISHI 36, where she sat down with Kiersten Fultz, a 2025 ISHI Student Ambassador, to talk about how STRmix™ came to be, what it means to design a tool with courtrooms in mind from the start, and what the field looks like from here.
Can you start by describing what STRmix™ actually does?
Bright’s explanation is methodical: STRmix™ is software designed to help forensic biologists interpret DNA profiles. It applies probabilistic techniques and biological models to assess genotype combinations and, if a person of interest is identified, to calculate a likelihood ratio.
“It’s a probabilistic genotyping method, and it’s just a combination of probabilistic techniques and biological models. And we can use that software to assess all the possible genotype combinations that might describe a DNA profile. And then if we have a person of interest, we can calculate a match statistic in the form of a likelihood ratio.”
The processing speed has also improved substantially since early development. What once took days for some interpretations now takes seconds.
How did the team approach building something that would be used in court?
The challenge wasn’t just technical. From early on, the team understood that STRmix™ wasn’t going to stay within one laboratory or one country.
“Something that we’ve really been mindful of is that STRmix™ isn’t just like an academic exercise, that we knew that we would be using it for our case work in New Zealand. But then we realized really quickly it was going to be an international tool for many different laboratories. So we were really mindful that we were really transparent about the models that went into it.”
That transparency wasn’t optional. Because the software would be used in court proceedings across multiple jurisdictions, the underlying science had to be defensible to a level that purely academic tools don’t require.
“We had to put a lot of really good science in there because it was going to be used in that court system.”
How do you approach training analysts who need to understand the statistics behind the software?
One of the ongoing challenges in probabilistic genotyping is that the statistical basis can feel remote to biologists who weren’t trained in that area. Bright describes the STRmix™ training as deliberately designed around that gap.
“That’s something that we try to do with our training — to make the training really focused for forensic biologists. And whilst we recognise that there is a lot of statistics, we try to make sure that we cover that in a way that a biologist can understand.”
Fultz noted she had used STRmix™ in a course that compared probabilistic genotyping output against manual interpretation. What stood out to her was how quickly the software processed data relative to traditional methods.
Where do you see DNA mixture interpretation heading in the next five to ten years?
Bright names three directions she expects to see develop: the application of probabilistic genotyping methods to Y-STR profiles, the extension of probabilistic methods to more upstream processes in the analysis pipeline, and a larger role for AI.
“I think what we’ll see moving forward is probably the application of these probabilistic genotyping methods to Y-STR profiles. And I think we’ll also see some of the probabilistic methods being applied to more upstream processes like analysis. And I also think we’ll see that AI has a big, big part to play.”
What advice do you have for students entering the field?
Bright is practical. She points to two skills she sees as differentiators: statistics and programming.
“Take every opportunity that is offered to you that you can. But also, more practically, look at taking a statistics course at school. And then you’ll be really quite appealing to a forensic laboratory. And if you can learn a language like Python or even R, some sort of programming language, that will help you with your career, particularly if you want to get into a little bit more research.”
Fultz said that working with RStudio in her own research had been exactly that kind of experience: intimidating at first, rewarding once she got into it.
For Bright, the practical skills matter. But so does showing up — to conferences, to conversations, to the parts of the field that are still being figured out.