Mixtures and Why More Contributors Means Less Certainty
A mixed sample is not simply a harder version of a single-source profile. The analyst must decide how many people contributed before anything else follows, and that decision is an estimate whose error travels into every figure reported afterward.
The rule in short
A profile is a mixture when the peaks at one or more loci cannot come from a single person. Interpretation begins by assigning a number of contributors, an assignment that is uncertain and that constrains every later step. Low template amounts produce dropout, drop-in, stutter and unbalanced peaks, which is why analytical and stochastic thresholds exist. Inclusion statistics discard information, and the NIST foundation review presses laboratories to work inside a validated range.
A sample is a mixture when more than one person contributed the DNA in it. That happens on shared objects, on clothing, under fingernails, and wherever a swab collects material from a surface many hands have touched. Interpretation of a mixture is a different activity from interpretation of a single-source profile, and almost every disputed DNA result in a criminal case is disputed because the sample was mixed.
What marks a profile as mixed
One person contributes at most two variants at a locus. When a trace shows three or more peaks at any locus, at least two people contributed. Uneven peak heights across a locus point the same way, since a single person's two variants ordinarily amplify to comparable heights. A mixture can also hide: two people who happen to share variants at a locus produce a pattern indistinguishable from one person at that locus.
Where the contributors are present in clearly different amounts, the peaks separate into a set of tall peaks and a set of short ones. Deconvolution is the attempt to sort those peaks into individual genotypes, and it works best when the ratio between contributors is wide and the total amount of DNA is generous. As the ratio narrows toward equal contribution, the tall and short sets stop being distinguishable and the mixture resists resolution into components at all.
Nothing in that process identifies who the components are. It produces candidate genotypes and weights, and the comparison with a reference and the reporting of a figure follow separately, in the manner described in how a DNA statistic is expressed.
Assigning the number of contributors
The first substantive decision is how many people to assume. The usual approach counts the greatest number of peaks seen at any locus and divides, then adjusts for peak height patterns and for what is known about the item. The NIST scientific foundation review on mixture interpretation identifies this assignment as one of the harder problems in the discipline, and treats it as a source of error rather than a settled preliminary.
The bias runs one direction. Because unrelated people frequently share variants, additional contributors can be concealed behind peaks already attributed to others, and the count is understated more often than overstated. The likelihood of concealment grows with each additional person, so the assignment is least reliable exactly where it matters most. Studies in which analysts are given mixtures of known composition report disagreement about the count, with the spread widening as contributors increase.
The assignment then propagates. It sets which genotype combinations are considered, whether a component is treated as resolvable, and which statistical model is applied. A figure calculated under one assumed count is not comparable to a figure calculated under another, and a report that gives the figure without the assumption omits half the result.
An assumed number of contributors is a condition the analyst supplies before the calculation runs, in the same way a reference profile is supplied. It is not a conclusion drawn from the calculation, and the software or worksheet that returns a figure has no way to test whether the assumption was right. A laboratory that runs the interpretation under more than one assumed count generates a range, and the range is often the more informative output.
What happens at low template
Small amounts of starting material make the copying step erratic. If a variant is present in only a few copies, the reaction may amplify one variant of a pair far more than the other, or may miss one entirely. Dropout is the loss of a real variant from the trace. Its opposite, drop-in, is the appearance of a variant that came from sporadic contamination rather than from a contributor, and it typically shows as a single low peak that fits no one.
Stutter compounds the problem. The copying step regularly produces a small product one repeat unit shorter than the true variant, and that artifact sits at the position where a genuine minor variant would appear. Distinguishing an elevated stutter peak from a real low-level allele is a judgment call, and it is the call most often revisited on review. Heterozygote imbalance, where a person's two variants amplify to very different heights, blurs the same line from the other direction.
Two thresholds structure the response. The analytical threshold marks the height below which a signal is not called a peak at all, separating data from baseline noise. The stochastic threshold marks the height below which a peak may be present without its partner, so a single peak cannot be read as evidence that the contributor carried the same variant twice. Data below the stochastic threshold are not unusable, but they cannot be read as if amounts were plentiful.
| Interpretation question | Single-source profile | Mixture with a clear major | Complex or low-template mixture |
|---|---|---|---|
| Number of contributors | Established by the peak pattern itself | Usually apparent, occasionally understated | Assumed, and frequently disputed |
| Use of peak heights | A quality check | The tool that separates the components | Unreliable, because amounts are too small |
| Effect of shared variants | None | Minor components may be masked | Masking is expected throughout |
| Risk of dropout | Low where the sample is intact | Confined to the minor component | Present across the profile |
| Statistic ordinarily available | A frequency estimate for one genotype | A ratio for the resolved major component | A weak ratio, or no figure at all |
Inclusion statistics and what they lose
Older approaches sidestep deconvolution. Random man not excluded, expressed as a combined probability of inclusion, asks what proportion of the population would be included as a possible contributor to the observed set of peaks. It requires no assumption about the number of contributors and no assignment of peaks to individuals, which is why it stayed in use for so long.
The cost is information. The calculation treats every observed peak as equally available to every contributor and ignores peak heights entirely, so it cannot distinguish a reference that fits the major component from one that fits scattered minor peaks. It also breaks down where dropout is possible: if a variant may be missing from the trace, the set of people the data exclude is unknown, and an inclusion figure calculated as though nothing dropped out overstates the case. Guidance has narrowed the method to samples where every contributor's variants are plainly visible. Where they are not, the modeling approach set out in probabilistic genotyping and its validation is the alternative.
Working inside a validated range
A laboratory establishes what its interpretation method can handle by running mixtures of known composition across contributor numbers, ratios and template amounts. That internal study defines a range: the conditions under which the method was demonstrated to perform. The foundation review's practical instruction is that laboratories should interpret within that range and should say so when a case sample falls outside it.
Case samples do not respect ranges. A swab from a shared surface may hold contributors in numbers no validation study covered, at amounts below anything tested. The written procedure should state what happens then, and the honest options are to report the result as inconclusive or to report it with the limitation attached. Whether the reported figure came from inside the demonstrated range is one of the first questions in the reliability showing for a forensic method, and it is separate from the question of how the material arrived, which belongs to transfer and the limits of touch DNA.
Points to carry away
- A profile is treated as a mixture when the number of peaks at a locus exceeds what one person can contribute.
- The assigned number of contributors is an estimate, and allele sharing between people causes that number to be understated more often than overstated.
- Peak height ratios separate a major from a minor contributor only while enough template is present for the heights to be informative.
- Below the stochastic threshold a missing peak may mean absence or may mean dropout, and the analyst cannot tell which from the trace alone.
- Combined probability of inclusion discards peak height information and cannot be applied where dropout is possible.
- The NIST scientific foundation review states that mixture interpretation grows harder as contributors increase and that laboratories should work within a validated range.
Questions readers ask
Does a larger contributor count always make the reported figure weaker?
Usually, though not by a fixed rule. Adding contributors multiplies the genotype combinations that could explain the same set of peaks, so the data become consistent with more people and the resulting figure moves toward neutrality. A well-resolved major contributor in a sample with several people can still yield a strong figure, because the major component behaves much like a single-source profile. The weakening is sharpest for minor components, where the peaks are low, sharing is common and dropout cannot be ruled out.
What does a laboratory do when a sample falls outside what it validated?
Practice varies. Some laboratories report the result as inconclusive and say so. Others report a comparison without a statistic, and others still apply the interpretation method beyond the conditions their internal study covered. The written procedure normally states which of these the laboratory permits, and the internal validation summary states which conditions were actually tested. Comparing the two documents against the case sample is the direct way to establish whether the reported result sits inside or outside the range the laboratory demonstrated.
Why can two analysts reach different conclusions about the same trace?
Most interpretation decisions involve judgment applied to a threshold. Whether a low peak is a real allele or elevated stutter, whether an unbalanced pair reflects one person or two, and how many contributors best explain a locus are all calls made by a person reading a trace. Laboratories reduce that spread with written rules, training and technical review, and some restrict the analyst's knowledge of the reference profile until the interpretation is fixed. The record of who decided what, and when, sits in the case file.
Sources
- NIST — DNA Mixture Interpretation: A NIST Scientific Foundation ReviewReviews the empirical basis for mixture interpretation, the difficulty of determining the number of contributors, and the need to work within a validated range.
- NIST — Scientific Foundation ReviewsDescribes what a foundation review is: an assessment of the scientific basis of a method and of the published evidence for its reliability.
- National Institute of Justice — DNA Evidence: Basics of AnalyzingDescribes extraction, quantitation, amplification and separation, and the marker systems available when a sample is degraded or male-limited.
- NIST — Organization of Scientific Area Committees for Forensic SciencePublishes the registry of standards covering interpretation, validation and reporting that a laboratory may adopt.
- 34 U.S.C. § 12592 — DNA identification indexRequires participating laboratories to follow published quality assurance standards, hold accreditation, and undergo periodic external audits.
- National Academies — Strengthening Forensic Science in the United States: A Path ForwardCalls for enforceable standards and for methods whose limits and error sources are stated rather than assumed.
Premier Defense Law is a publication, not a law firm. This article states general rules and cites its sources; it is not advice about any particular case, and the law differs by state and changes over time.
More in Forensic Evidence
Fingerprint Comparison and the ACE-V Method
A latent print comparison runs through analysis, comparison, evaluation and verification. Sufficiency at the analysis stage is the examiner's judgment and is not fixed by any national minimum point count. An automated search returns ranked candidates rather than conclusions. Verification may or may not be blind. Black box testing measures the accuracy of conclusions without examining how they were reached, and a reported error rate describes study participants, not a single comparison.
Retaining a Defense Expert and Paying for One
Section 3006A(e) authorizes investigative, expert and other services necessary for adequate representation where the person is financially unable to obtain them. The application may be made ex parte and heard ex parte, so the request does not disclose the theory of the defense. Compensation is capped at an amount the court may exceed on certification approved by the chief judge of the circuit. Rule 706 supplies a court-appointed route, and Rule 16 governs disclosure once the expert testifies.
Firearms and Toolmark Identification
Firearms examination compares class, subclass and individual characteristics on fired components under a comparison microscope. The identification threshold is agreement judged sufficient by the examiner, not a fixed count of matching striae. Subclass carryover from consecutively produced tooling can imitate individual agreement. Correlation databases return ranked candidates, and conclusion wording has moved away from claims of identification to the exclusion of every other firearm.