What Is STRmix and How Is It Used in Minnesota Courts?

A defendant’s guide to probabilistic DNA evidence — and why a high likelihood ratio is not the same as proof of guilt.

You’ve just been told the DNA doesn’t lie. Here’s what they didn’t tell you about the number.

STRmix is probabilistic genotyping (PG) software used by forensic laboratories, including the Minnesota Bureau of Criminal Apprehension (BCA), to interpret complex DNA mixtures (samples containing genetic material from two or more people).

In Minnesota criminal cases, prosecutors rely on STRmix to generate Likelihood Ratios (LRs) that can sound staggeringly definitive: “The probability of observing this DNA mixture is greater than 100 billion times more likely if Defendant and “x” unknown unrelated individuals are the source than….”

But here is the critical reality most defendants are never clearly told:

STRmix does not determine guilt. It produces a statistical opinion built on human assumptions,  and those assumptions can be wrong.

As a criminal defense attorney with more than 10 years of DNA litigation experience, STRmix training equivalent to, or greater than, that of state lab analysts, and a Fellowship in the American Academy of Forensic Sciences (AAFS), I have seen firsthand how STRmix evidence is presented in Minnesota courtrooms — and how rarely the limitations behind the numbers are disclosed to juries.

This article explains what STRmix actually is, how it is used in Minnesota courts, and why a rigorous challenge to its underlying assumptions can be the difference between conviction and acquittal.

This is not legal advice. Every case is fact-specific, and you need a case-specific consultation.

What Is STRmix? The Plain-English Explanation

STRmix is a software that was designed to solve a problem that stumped forensic scientists for decades: when a crime scene sample contains DNA from three, four, or even five contributors, traditional analysis cannot separate each contributor’s genetic markers (alleles). Historically, labs declared such samples “uninterpretable.”

STRmix attempts to resolve this by running millions of computer simulations — using a statistical engine called Markov Chain Monte Carlo (MCMC) — to determine which combination of contributors most plausibly explains the raw DNA data (the electropherogram). The result is not a match or a non-match. It is a Likelihood Ratio.

What Is a Likelihood Ratio?

A Likelihood Ratio (LR) compares two competing hypotheses:

1. Prosecution hypothesis: The defendant is a contributor to the mixture.
2. Defense hypothesis: The DNA came from unknown, unrelated individuals.

It is important to note that it is the analyst who picks the defense hypothesis, not the defense attorney or the defendant.

An LR of greater than 100 billion means the DNA profile is one hundred billion times more likely under the prosecution’s assumption than the defense’s. But critically:

A Likelihood Ratio measures the strength of evidence under specific assumptions. It does not measure the probability that the defendant is guilty.

Confusing these two things — and they are very different things — is the heart of what is called the transposed conditional or the Prosecutor’s Fallacy. More on that below.

How Probabilistic Genotyping Works: The Five Mechanisms

Understanding how STRmix builds its statistical model is essential to understanding where it can fail. Each stage below represents a point where human judgment, or data quality, can compromise the result.

1. Analyzing the Raw Data

STRmix begins by looking at the raw DNA data which includes looking at peaks that are appearing in locations known to be consistent with human genetic markers (alleles) across multiple chromosomal locations. The height and balance of those peaks reflect how much DNA was present at each marker. The software evaluates peak heights, peak height balance ratios, and potential artifacts like stutter (a known byproduct of the PCR amplification process).

Key challenge: Stutter and other artifacts can mimic real alleles. The software’s ability to correctly classify them depends on the lab’s internal validation parameters, which vary by each lab.

2. Estimating the Number of Contributors (NOC)

Before STRmix runs its simulations, the analyst must tell it how many people contributed to the mixture. This is not something the software determines independently. It is a human judgment call.

If an analyst assumes three contributors when there were actually four, every calculation that follows is built on a flawed foundation. This is garbage-in, garbage-out forensic science.

The Number of Contributors (NOC) determination is one of the most contested and consequential decisions in any STRmix case, and it is one of the first things I examine in the Barron Forensic Integrity Protocol. If the NOC is wrong, the LR is not just “off”—it can be mathematically meaningless.

3. Running MCMC Simulations

With the NOC set, STRmix uses Markov Chain Monte Carlo (MCMC) algorithms to explore billions of possible contributor combinations, accepting or rejecting configurations based on how well they fit the observed peaks. Think of it as a very complex, computerized process of elimination.

The critical issue is convergence: MCMC simulations must “settle” into a stable state before the result is reliable. If the chain does not converge, or if the “burn-in” period (the initial steps discarded to stabilize the model) is insufficient, multiple runs of the same data can produce different Likelihood Ratios. This instability is a measurable form of scientific uncertainty that prosecutors and lab analysts rarely explain to juries.

4. Accounting for Stochastic Effects in Low-Template DNA

Many Minnesota cases involve touch DNA — tiny amounts of skin cells left behind on a steering wheel, a firearm grip, or a door handle. When DNA quantities are extremely low (low-template DNA), the PCR amplification process becomes subject to stochastic effects: random statistical fluctuations that cause:

1. Allele dropout — a real allele fails to amplify and disappears from the profile.
2. Allele drop-in — a false, artifactual allele appears in the profile.

When a lab applies STRmix to samples below their validated threshold, the reported LR may be mathematically precise but factually unstable. For more on that problem, see Low Copy Number DNA Evidence.

5. Generating the Likelihood Ratio

The final output is the LR. This number is what prosecutors present in court. But the LR is not a biological fact, it really isn’t even a fact at all. It is the end product of every decision, assumption, and data-quality issue that preceded it. It is a statistical opinion, and like any opinion, it can be wrong.

Why STRmix Matters in Minnesota Criminal Cases

Minnesota courts have generally found STRmix to be admissible under the Frye-Mack standard, which evaluates whether a scientific technique is generally accepted in the relevant expert community. The BCA and other Minnesota labs have conducted internal validation studies that courts have used to support admissibility rulings.

But admissible is not the same as reliable. And reliable under ideal lab conditions is not the same as reliable in your specific case, with your specific sample, analyzed by your specific analyst.

In Minnesota DNA prosecutions, STRmix is frequently used in cases involving:

1. Touch DNA on weapons, vehicles, or clothing
2. Mixed samples with three or more contributors
3. Low-level, degraded, or environmentally compromised samples
4. “Resurrected” evidence — older samples that were previously declared uninterpretable

These are not always clean, high-quality reference samples. They are interpretation-driven cases where the assumptions fed into STRmix can have an enormous effect on the output.

Real-World Case Illustration: When the Number Changes

A Case Example from Practice

A firearm is recovered in a Minnesota felony case. The BCA analyst reports a Likelihood Ratio of greater than 100 billion. The verbal equivalent scale on the report lists this as “very strong support.” “Likelihood Ratios” are often translated into these subjective “Verbal Qualifiers” by the lab, which adds another layer of human interpretation to the “objective” math. The prosecution frames this as near-certainty that the defendant held the gun.

A forensic audit of the raw data revealed four critical issues:

1. The analyst assumed three contributors, but the allelic data was consistent with four.
2. The DNA quantity was below the lab’s own validated threshold, making interpretation questionable as to foundational reliability.
3. The MCMC simulations showed convergence instability.
4. When reviewing the contributor position that provided the highest LR for defendant approximately a quarter of the profile did not list Defendant’s genotype as the best explanation for the evidence at those locations (loci).

Same evidence. The same DNA. The same gun. A completely different narrative unfolds because the assumptions and conclusions are challenged. That is not a flaw in the science. That is the science working exactly as it should, someone just has to be looking at it. This is why I say the science can be accurate and still legally misleading.

What This Means Legally

When a forensic audit reveals the issues described above, an unsupported NOC determination, DNA quantity below the lab’s own validated threshold, MCMC convergence instability, and a contributor position where the defendant’s genotype is not even the best explanation for a quarter of the profile, the evidence does not just become questionable. It becomes a pretrial admissibility argument. Under Minnesota Rules of Evidence 702 and 403, the defense has a legitimate basis to challenge whether this evidence meets the foundational reliability requirement at all, and whether its prejudicial impact outweighs its probative value. That argument does not come from the courtroom. It comes from the data.

For Defendants: What You Need to Know

If you are facing charges in Minnesota where DNA evidence is central to the prosecution’s case, here is what matters most:

1. A high Likelihood Ratio does not mean you are guilty. It means the data fits one mathematical model better than another, under conditions chosen by a lab analyst.
2. DNA presence does not equal DNA guilt. Think of DNA like glitter — it spreads easily, transfers unexpectedly, and once it’s somewhere, it is extraordinarily difficult to determine how it got there. Secondary transfer (DNA moving through indirect contact) is biologically real and legally important.
3. The prosecution is presenting a statistical narrative, not an objective fact. Every number you hear in that courtroom was produced by a model built on human decisions. Those decisions can be challenged.
4. You need a case-specific, forensic-level audit. A deep dive into the raw electropherogram data, the analyst’s NOC decision, the lab’s validation studies for your specific sample type, and the convergence behavior of the MCMC simulations in your case.

Most defendants accept the number as presented in the lab report. That decision, more than what the evidence can reliably say, can be the difference between a conviction or an acquittal.

For Defense Attorneys: What to Look For

STRmix cases look like DNA cases. They are more than that. They are probabilistic modeling cases that happen to involve DNA and that distinction determines whether the evidence gets meaningfully challenged or just cross-examined around the edges.

The questions that matter in these cases are not legal questions. They are questions like: Did the MCMC simulations converge? Was the NOC determination documented contemporaneously, or reconstructed after the fact? Does the lab’s internal validation study actually cover this sample type at this DNA quantity? Would different, but equally plausible subjective decisions or assumptions change the output in the case?

Most defense attorneys — even excellent ones — have not spent years challenging forensic DNA cases, lived inside forensic lab validation studies or sat through the same STRmix training that BCA analysts receive from the developers of STRmix themselves. That’s not a criticism. It’s a recognition that this is a distinct discipline, and treating it like standard expert-witness cross-examination is one of the most common ways these cases are under-defended.

Here is what I have seen happen when STRmix evidence goes unchallenged at the data level: the defense cross-examines the analyst competently, the analyst answers competently, and the jury walks away with a greater than 100 billion number they were never given a real reason to doubt. The assumptions behind that number, the decisions that determined it, were never exposed, because exposing them requires access to the underlying lab file and the forensic fluency to know what you’re looking at.

I work with defense attorneys across Minnesota in several ways depending on what the case needs: full forensic audit and expert coordination, consulting on discovery strategy and cross-examination preparation, or co-counsel arrangements for cases heading toward a 702 admissibility hearing. The goal is simple: make the state prove the number, not just present it.

If you have a case where STRmix is central to the prosecution’s theory, the earlier we talk, the more options you have.

507-822-5735 | vbarronlawoffice.com

The Prosecutor’s Fallacy: The Most Dangerous Misconception in Minnesota DNA Trials

The most common, and most damaging, misunderstanding of STRmix evidence is called the Transposed Conditional, or the Prosecutor’s Fallacy. It is when the probability of the evidence given a hypothesis is wrongly confused with the probability of hypothesis given the evidence. It works like this:

1. What STRmix actually says: “The probability of observing this DNA mixture is greater than 100 billion times more likely if the Defendant and “x” unknown unrelated individuals are the source than if x+1 unknown, unrelated individuals are the source.”
2. What juries often hear or what prosecutors say: “The probability that the defendant is in this mixture is greater than 100 billion”

These are not the same statement. They are not even close to the same statement. The transposed conditional happens when someone flips the statement to talk about the truth of the hypothesis rather than the strength of the evidence. The distinction is subtle, but extremely important. The numbers are large, and in a courtroom, juries do not understand this difference unless it gets explained to them.

The science can be accurate and still be legally misleading. A Likelihood Ratio is a statement about evidence — not a statement about guilt.

I use plain-language analogies to make this clear to juries. One example: Imagine finding a set of tire tracks at a crime scene that match the defendant’s vehicle. The probability of seeing those tracks given that the defendant was there might be very high. But that tells you nothing about all the other vehicles in the county that share the same tire model. The likelihood ratio tells you the tracks fit — it does not tell you the defendant was the only person who could have made them.

A piece of evidence could have an extremely high LR for a Defendant and an even higher LR for other people in the population.

The Barron Forensic Integrity Protocol™: A Structured Defense

Challenging STRmix requires more than skepticism — it requires a disciplined, stage-by-stage audit of every point where the evidence can be questioned. My approach structures that audit across seven stages:

1. Scene Integrity — Was the sample collected before contamination could occur?
2. Collection and Handling — Were proper protocols followed during evidence collection?
3. Foundational Reliability — Does the lab’s validation study cover this situation?
4. Lab Data Auditing — Do the electropherogram files support the analyst’s reported findings? This is where most problems live. The printed report is a summary of decisions already made. The raw file is where you see whether those decisions were justified — or whether the data told a different story that never made it into the report.
5. Interpretation Challenge — Were the NOC and other inputs scientifically justified?
6. SOP and Disclosure Review — Did the lab follow its own protocols, and was everything properly documented?
7. Validation Study Review — Does published science support the lab’s methodology at this DNA quantity and mixture complexity?

This is not a checklist exercise. It is a forensic investigation — and it frequently uncovers issues that the lab analyst did not disclose, did not document in their report, or did not consider. For a broader forensic review model, see the Barron Forensic Integrity Protocol.

Key Takeaway

STRmix is a statistical model, not a biological truth. It is a powerful interpretive tool and precisely because of that power, it demands rigorous scrutiny.

A high Likelihood Ratio is not a conviction. It is the starting point for a forensic defense that demands access to the underlying data, the lab’s protocols, the analyst’s decision records, and an independent review that can stress-test every assumption.

In cases where STRmix evidence is the centerpiece of the prosecution, outcomes often turn on a single question: was the science accepted as presented or was it challenged?

If you are a defendant in Minnesota facing DNA evidence you don’t fully understand, or a defense attorney who needs forensic support for a complex DNA case, don’t wait. The earlier a forensic audit begins, the more options you have.

Call or text Ginny Barron directly: 507-822-5735 | Visit vbarronlawoffice.com to submit a confidential case inquiry.

Frequently Asked Questions

1. What is STRmix in plain terms?

STRmix is probabilistic genotyping software used in Minnesota criminal courts to analyze mixed DNA samples; samples containing DNA from more than one person. Instead of a simple match or non-match, it produces a Likelihood Ratio representing how much the evidence favors one hypothesis over another. That number depends entirely on the assumptions fed into the model by a human analyst.

2. Does a high Likelihood Ratio mean the defendant is guilty?

No. A Likelihood Ratio measures the relative strength of DNA evidence under specific statistical assumptions — it is not a probability of guilt. Courts and juries frequently conflate these two concepts, which is the foundation of the Prosecutor’s Fallacy. Every case is fact-specific, and you need a case-specific consultation.

3. Is STRmix admissible in Minnesota courts?

Yes, Minnesota courts have generally accepted STRmix under the Frye-Mack standard for scientific reliability. However, admissibility does not equal accuracy. The evidence can still be challenged under Minnesota Rule of Evidence 702 on the reliability of specific inputs, the analyst’s NOC decision, data quality, and whether the lab’s validation study covers your specific situation.

4. Can STRmix give a wrong result?

Yes. STRmix results are sensitive to human assumptions — particularly the number of contributors, interpretation thresholds, and how artifacts are classified. Changing one assumption can shift the Likelihood Ratio by orders of magnitude. For low-template DNA samples, stochastic effects create additional instability that may not be disclosed in the final report.

5. What is secondary transfer, and why does it matter in DNA cases?

Secondary transfer means DNA can move from one person or surface to another through indirect contact such as a handshake, a shared surface, or a piece of clothing. DNA is like glitter: it spreads, it sticks, and its presence does not tell you when it arrived or who put it there. STRmix can help identify whose DNA is present but it cannot tell you how or why it got there.

6. What is the Prosecutor’s Fallacy in DNA cases?

The Prosecutor’s Fallacy (also called the Transposed Conditional) is when someone flips the statement about the strength of the evidence to talk about the truth of the hypothesis. A Likelihood Ratio of greater than 100 billion does not mean the defendant is one billion times more likely to be in the mixture or 100 billion times more likely to be guilty. It means the data fits one statistical model greater than 100 billion times better than another, under specific assumptions. These are fundamentally different things.

7. How can a defense attorney challenge STRmix evidence in Minnesota?

Effective challenges require access to the raw electropherogram data, the analyst’s documented NOC determination, the lab’s internal validation studies, and MCMC convergence data. An admissibility hearing under Minnesota Rule of Evidence 702 can also be used to challenge whether the lab’s stated conclusions and methodology applied in the case meet the foundational reliability requirement. These are not routine discovery requests. They require knowing what you’re looking for and why it matters, which is why early forensic consultation changes what’s possible in these cases.

This article is for general educational purposes only and does not constitute legal advice. Every case is fact-specific, and results depend on the particular facts and evidence involved. Contact Barron Law Office for a case-specific consultation. | Barron Law Office | Westbrook, Minnesota | 507-822-5735 | vbarronlawoffice.com