When the Mixture Doesn't Add Up
Introducing ISHI On-Demand
Tara Luther, Promega
Share this article
The profile looks like a four-person mixture. Or maybe three. The likelihood ratio is strong—until you adjust one assumption. The defense attorney asks you to explain it in court, and suddenly the math you trusted feels harder to defend than you expected.
If you've been there, you're not alone.
DNA mixtures remain one of the most technically demanding aspects of forensic casework, and probabilistic genotyping—despite becoming standard practice in many labs—continues to raise questions that don't have easy answers. How do you decide on number of contributors when it's ambiguous? What happens when you condition on someone who isn't actually a contributor? How do you explain a likelihood ratio to a jury using analogies that clarify rather than mislead?
These aren't hypothetical concerns. They're the questions analysts face at the bench, supervisors wrestle with during validation, and laboratory directors defend in court.
That's why ISHI is launching something new.
Introducing ISHI On-Demand: Expert Perspectives. Real-World Context. On Your Schedule.
ISHI On-Demand modules offer in-depth, expert-led conversations on some of the most complex and evolving topics in forensic DNA science. Each module brings together respected scientists, practitioners, and thought leaders to explore how forensic methods are applied, challenged, and defended in real laboratory and courtroom settings.
Rather than scripted lectures or vendor demonstrations, ISHI On-Demand focuses on professional insight, decision-making, and lived experience from the field.
What to Expect
A blended format. Carefully selected lectures paired with expert interviews that expand, contextualize, and challenge the material.
Trusted voices from the forensic community. Speakers include internationally recognized forensic scientists, statisticians, technical leaders, and practitioners with direct casework and courtroom experience.
Practical perspective, not vendor-driven content. Focused on how forensic science is actually practiced—from validation and interpretation to documentation and communication.
Flexible, on-demand access. Watch on your own time, revisit key segments, and engage with the material at the depth most relevant to your role.
Who It's Designed For
ISHI On-Demand modules are intended for a broad professional and educational audience, including:
- Forensic DNA analysts and technical staff
- Laboratory supervisors, managers, and directors
- Forensic science students and trainees
- Law enforcement professionals working with DNA evidence
- Legal professionals seeking to better understand forensic DNA interpretation and communication
For some viewers, the material may serve as an introduction or refresher. For others—particularly those already working with these methods—it may provide additional context, perspective, or insight into how peers approach shared challenges.
ISHI On-Demand is not intended to replace laboratory-specific training, validation, or policy development. It's the context, peer insight, and honest perspective that helps you use those tools more effectively.
Now Available: Probabilistic Genotyping, DNA Mixtures, and Likelihood Ratios
Our first module brings together six internationally recognized forensic scientists and statisticians who've spent years working through the challenges probabilistic genotyping presents. Now available for an introductory price of $25, the module offers something most training doesn't: honest, casework-driven conversations about what works, what doesn't, and what you need to watch out for.
What You'll Actually Learn
This isn't a software tutorial or a validation checklist. It's seven video segments designed to give you the kind of insight that only comes from experience—the perspective you'd get if you could sit down with these experts and ask them everything you've been wondering.
Foundations of Probabilistic Genotyping and DNA Mixtures gets honest about why mixtures are still hard. Richard Wivell, Jo Bright, and Dr. Michael Coble discuss how probabilistic methods support analyst judgment (not replace it), where likelihood ratios come from, and why some mixture scenarios still push the limits of what the models can handle. If you've ever second-guessed a complex profile, this segment explains why your instinct to pause was probably right.
Validation and Implementation Considerations is for anyone who's had to defend their validation plan—or sit through one that felt disconnected from actual casework. The conversation covers what works when adopting probabilistic genotyping, how to learn from labs that implemented early, and how to communicate validation results transparently when ground truth is unknowable. The discussion also addresses software updates and deciding when—and when not—to adopt new versions.
Interpretation, Assumptions, and Robustness digs into the choices that shape your results—and the risks that come with them. How do you decide on number of contributors? What happens when you condition on the wrong person? When should you trust the likelihood ratio, and when should you question it? This segment won't give you a formula, but it will give you the framework to think critically about your own casework.
Communicating Likelihood Ratios in Court tackles one of the field's biggest pain points: explaining probabilistic results to people who don't speak statistics. The speakers share what's worked for them in testimony (weather analogies, cookie boxes), discuss the ones that backfire, and explore how much transparency is enough without overwhelming a jury. The segment also introduces activity-level evaluations and differences in adoption between jurisdictions. If you've ever left the witness stand wishing you'd said something differently, this one's for you.
Lessons from the Field is where Austin Hicklin and Nicole Richetelli pull back the curtain on what they've seen across labs nationwide through an NIJ-funded project collecting implementation experiences. Implementation timelines that ran long. Training plans that didn't account for staff turnover. Labs realizing mid-validation they'd need to change analytical thresholds. This isn't about pointing fingers—it's about learning from the patterns so you can plan smarter.
Mixture Interpretation Beyond Traditional Workflows looks at where the field is heading. Cydne Holt discusses the value of sequence-based data from massively parallel sequencing, the limitations of traditional CE-based methods, and how semi-automated tools can support documentation and transparency for both STR and SNP data. It's forward-thinking without being speculative—grounded in research and real casework.
The final segment, Looking Ahead, brings everyone back together to reflect on what's next: probabilistic genotyping for Y-STRs, the role of AI in genotype assignment, continued evolution of the likelihood ratio framework, and why adaptability will matter as much as technical skill.