What DNA Actually Is, What It Actually Says, and What It Says About People Who Are Not Even in the Room

About three weeks ago, I was sitting at the clubhouse of the association I belong to, having a glass of water after training, when two of the other members came over and asked me how this business with DNA actually works in real cases, and you can probably guess what happened next, because when somebody asks me a question about a topic I have spent a substantial part of my adult life inside, I do not produce a two-minute answer, I produce a lecture. The two members were polite enough to stay through it, they asked follow-up questions, and by the time the afternoon had turned into evening, I had realized two things at the same time, the first of which was that most intelligent, educated, well-informed people have almost no idea what a DNA profile actually is, what it shows, what it does not show, and how it can be wrong, and the second of which was that I had been carrying this material around in my head for decades without ever writing it down in a form that someone outside the profession could actually read.

I have spent the past few weeks turning that afternoon conversation into this article, which is the first serious scientific piece on DNA that has ever appeared on rauscher.xyz, and it is long, detailed, and contains things that the mainstream coverage of forensic DNA has been quietly leaving out for years. It ends with a point that has kept me awake more than once in recent months, because it concerns what an entire generation of well-meaning people are doing to themselves, to their children, and to their children's children every time they send a saliva sample to a commercial laboratory in exchange for a written report about their ancestry and their Neanderthal percentage.

Before any of that, however, we need to start with the easy part, which is the science itself, and which forms the necessary foundation for everything that follows in the rest of this piece.

What a DNA Profile Actually Is

Nearly every educated person has heard that DNA is a double helix, that it encodes the genetic information of the human organism, and that each individual, with the exception of identical twins, carries a unique sequence, and all of that is correct, but almost none of it has anything to do with what a forensic laboratory actually examines when it produces what is commonly called a DNA profile. The confusion between the genome as a whole, which contains roughly three billion base pairs of sequence information, and the forensic profile, which consists of a small number of measured values at specific locations within that sequence, is the first and most persistent misunderstanding in this field, and clearing it up changes the way a person reads every DNA-related news story for the rest of their life.

The human genome is organized into twenty-three pairs of chromosomes, and within that genome there are regions that code for proteins, regions that regulate when and how those proteins are produced, regions whose function is still poorly understood, and regions that appear to have no function at all in the sense that classical genetics recognized, but whose properties make them extraordinarily useful for identification purposes. The regions forensic laboratories examine are of the last category, and they are called Short Tandem Repeats, abbreviated throughout this article as STRs, and a Short Tandem Repeat is simply a stretch of DNA in which a short sequence of bases, typically four nucleotides long, is repeated a variable number of times end to end, in much the same way as one might write the same short word over and over again on a single line of paper. The number of repetitions at any given STR locus varies considerably from person to person, and this variation is inherited according to the standard rules of Mendelian genetics, which means that each individual carries a combination of repeat counts that, taken across enough loci, becomes effectively unique. What makes STRs so valuable for forensic identification is that the number of repeats at a given locus can be measured with high precision using standard laboratory techniques, that the measured values at different loci are statistically independent of one another, and that the population distribution of those values is well characterized in reference databases, so that the combined profile of twenty or so loci produces a numerical signature whose probability of random match with an unrelated person is, under ideal conditions, smaller than one in several quadrillion.

A modern German forensic DNA profile typically measures sixteen to twenty-four STR loci along with the amelogenin marker, which serves as an indicator of chromosomal sex, and the result is a sequence of numerical values that looks something like this when written out on paper: at locus D3S1358 the measured values are fifteen and seventeen, at locus vWA they are fourteen and eighteen, at locus D16S539 they are eleven and twelve, and so on through the full panel of measured loci. The amelogenin marker should not be understood as an absolute proof of biological sex, since rare deletions on the Y chromosome and other chromosomal anomalies can produce results that diverge from the phenotypic sex of the person, and any conscientious laboratory therefore treats the amelogenin reading as a strong indication rather than as conclusive evidence in isolation. There are two numbers per locus, because each person inherits one copy of each chromosome from each parent, and because the two copies can carry different repeat counts, and that set of paired numerical values is what travels through the law enforcement database, what gets compared against reference samples, and what occupies the spreadsheet the expert witness shows to the court. It is not the person's genome, it is a numerical fingerprint derived from specific, deliberately chosen non-coding regions of the genome, designed to identify individuals while keeping medical, behavioral, and ancestral information out of the analysis to the greatest extent technically possible. The non-coding STR markers used in forensic identification do carry some residual statistical information about population background, since allele frequencies at these loci differ between human population groups, and this is precisely why reference databases must be matched to the relevant population for any meaningful random-match calculation, but no diagnostic information about the individual's ancestry, predispositions, or traits is meant to be extracted from a properly conducted forensic analysis. The distinction is critical, because it is the reason the German legal framework permits forensic DNA analysis at all. Under Section 81e paragraph 1 of the German Code of Criminal Procedure, the Strafprozessordnung, molecular-genetic analysis of material obtained through measures under Section 81a paragraph 1 or under Section 81c is permitted for the purpose of establishing the DNA identification pattern, the parentage, and the sex of the person, with comparison against reference material insofar as this is necessary to investigate the facts of the case, and the same provision expressly prohibits any other findings, declaring investigations directed at other genetic properties of the sample inadmissible in unmistakably clear statutory language.

How Little DNA Is Enough, and How Quickly You Leave It

The intuitions most people carry around about DNA evidence were formed during the first two decades of forensic genetics, roughly between 1985 and 2005, when the quantities of biological material needed to produce a reliable profile were substantial, and when a clearly visible bloodstain, a pool of saliva, or a semen sample of some volume were the canonical inputs that the public imagination came to associate with the technology, an association that has largely failed to update itself in the years since. The actual capabilities of modern laboratories have moved on by orders of magnitude, and the implications of that shift for how DNA evidence ought to be interpreted are still being absorbed by the courts, by the police, and by the general public, all of whom continue to operate on a mental model that has not been accurate for at least fifteen years.

A standard forensic DNA profile can today usually be obtained from approximately one hundred picograms of starting material, which is one hundred trillionths of a gram and corresponds to the DNA content of roughly fifteen to seventeen diploid human cells, although laboratories increasingly recognize that even at this nominal threshold the resulting profile shows stochastic effects such as elevated stutter, allele dropout, allele drop-in, and peak height imbalance, all of which need to be taken into account during interpretation. Below this range, in what the field calls low template or low copy number territory, working profiles have been reported from input quantities of fewer than thirty picograms, corresponding to the DNA content of fewer than five cells, but this is not a stable region of routine casework so much as a specialized and partly experimental zone in which the same protocol applied to identical samples can produce notably different results from one run to the next, and in which the statistical interpretation of the resulting profile becomes considerably more cautious than at higher input levels. Five cells. That is the lower bound at which the technology can sometimes be coaxed into producing a result, and the implications of this sensitivity for the behavior of biological evidence at a crime scene are precisely where the public understanding of DNA has failed to keep pace with the science, and where most of the problems this article describes have their origin.

Every human being sheds skin cells at a rate of roughly a million cells per day, most of them deposited onto clothing, bedding, door handles, keyboards, steering wheels, and every other surface that the body comes into routine contact with, and a single handshake lasting only a few seconds can transfer sufficient cellular material from one person to another to produce an interpretable DNA profile under favorable conditions. The act of speaking distributes saliva aerosols over a distance of at least a meter, with many of these aerosol droplets carrying epithelial cells from the oral mucosa, although not every droplet contains nucleated cellular material, and the proportion that does varies with the individual speaker, with the type of vocal activity, and with the ambient conditions. Touching a glass, holding a steering wheel, gripping a handrail, sitting on a chair, leaning against a wall, none of these are events that require any forensic imagination to reconstruct, they are simply the ordinary conditions under which a modern human being moves through a modern city, and they can deposit sufficient DNA on contacted surfaces to produce a profile within a relatively short time of contact, depending on the individual and on a range of biological and environmental factors. This phenomenon is known in the field as touch DNA, and the ease with which touch DNA is deposited, transferred from person to object, then transferred again from object to person or object, and subsequently recovered, is something whose full significance the legal system has still not entirely come to terms with.

One of the key pieces of research that established the scale of the transfer problem appeared in the 2016 peer-reviewed literature, and it demonstrated that DNA from one person can be deposited on objects that the person has never physically touched, through what is called secondary and even tertiary transfer, so that if person A shakes hands with person B, and person B then touches a knife, DNA from person A can subsequently be recovered from the knife despite the fact that person A has never been anywhere near the weapon in question. This is not a theoretical laboratory finding without practical relevance, it is a physical phenomenon that occurs constantly in everyday life, and its implications for the interpretation of DNA evidence at a crime scene are profound, because the presence of a person's DNA at a crime scene does not prove that the person was ever physically present at the crime scene, it proves only that the person's DNA was there, and these two statements are emphatically not the same thing, with the gap between them being exactly the place where miscarriages of justice tend to occur.

Single Source, Smear, Mixture: A Taxonomy of What Laboratories Actually Receive

In the comfortable imaginary world in which DNA evidence is simple, every sample that arrives at the laboratory comes from one person, contains a large quantity of intact genetic material, and produces a clean, unambiguous profile that either matches a reference sample perfectly or does not match at all, with no shades of gray in between, but in the real world of forensic casework, although such samples do exist, they are a minority of what laboratories actually process, and the more interesting, more difficult, and more legally consequential category consists of samples that depart from this ideal in systematic and characterizable ways. Three categories in particular deserve to be understood by anyone who wants to follow a DNA-based prosecution with any degree of critical intelligence, and they are worth describing one at a time.

The first category is the single-source trace, sometimes called a Type A trace in the German forensic literature, and this is the textbook case, a drop of blood on a doorframe, a cigarette butt with a single smoker's saliva, a semen sample from a single contributor, a visible droplet of biological material whose DNA profile shows the expected pattern of one or two allele peaks at each locus, consistent with a single human source. When such a trace is of sufficient quantity and quality, the resulting profile can be compared against a reference with essentially no ambiguity, and the random match probability for such a complete single-source profile, compared against an unrelated member of the general population, is typically on the order of one in a quadrillion or smaller, which is the kind of DNA evidence that produces the staggering numbers cited in court, and under these conditions the underlying mathematics is genuinely defensible. The problem is that most of forensic casework does not involve such clean samples, however much courtroom presentations may give the impression that it does.

The second category is the smear trace, which is what results when biological material has been physically dragged, wiped, or spread across a surface by motion, whether through the original deposition itself, through cleaning after the fact, or through the passage of other people or objects across the deposit. Smear traces still contain the DNA of the original contributor, but the spreading process distributes the available material across a larger surface area, dilutes the local cellular density, and exposes the deposited cells to environmental conditions that can accelerate degradation through enzymatic activity, microbial action, ultraviolet light, humidity, and temperature variation, all of which combine to reduce the amount of intact, amplifiable DNA that can be recovered at any given sampling point. A smear trace typically produces a profile with reduced peak heights, with allele dropout at certain loci, and with increased sensitivity to the exact sampling strategy used by the forensic technician, so that a profile generated from a smear trace can still be highly informative in many cases, but the statistical weight that can be assigned to such a profile is always lower than that of a clean single-source sample, and the interpretation must always account for the possibility that some loci have failed to amplify not because the contributor carries an unusual allele but because the sample is dilute or degraded.

The third category, and by some considerable distance the most analytically difficult of the three, is the mixture trace, which contains the DNA of two or more contributors in a single sample, and mixture traces appear routinely in casework, where they represent the normal condition rather than the exception whenever biological material is recovered from commonly shared surfaces, from clothing that has had multiple wearers or launderers, from vehicle interiors, from weapons that have been handled by more than one person, and from numerous other investigative contexts. The technical problem that a mixture trace presents is that the electropherogram, which is the graphical output of the DNA analysis showing peaks at each locus, displays more than two peaks at positions where a single contributor would show only one or two, and the laboratory's task is to disentangle those overlapping profiles and to determine which combinations of allele values are consistent with which number of contributors in which proportions. For a two-person mixture in which one contributor is dominant and the other is present at substantially lower concentration, this disentangling is sometimes manageable, for a three-person mixture with comparable contributor ratios the problem becomes mathematically severe, and for a four-person mixture with degradation artifacts, the problem approaches the boundary of what current methodology can reliably address at all.

The DNA trace recovered from the rear license plate of the burned-out getaway vehicle in the proceedings around the IKEA armored-transport robbery in Frankfurt of the ninth of November 2019 was reported in the public record as a mixture trace of precisely this kind, and on the available reporting it was not a clean single-source deposition that could simply be compared against a reference sample on the spot. According to the materials made publicly available during the proceedings, it contained material from more than one contributor, in proportions that did not appear to be those of a single dominant profile with a minor contaminant, and its interpretation required statistical techniques that had themselves evolved substantially over the years between the original analysis and the moment at which the court eventually ordered a fresh blood sample from the accused in order to recalculate the evidence under newer scientific standards. The technical work involved in such an analysis is not controversial because the underlying science is unreliable, it is challenging because the reality of mixture interpretation is genuinely difficult even when everything in the laboratory is done correctly, and because the statistical frameworks used to express the result can produce, and routinely do produce, substantially different numbers depending on which assumptions are built into the calculation by the analyst.

The Statistics, Briefly, Because They Matter

When a forensic laboratory reports a match between a crime scene sample and a reference sample, the quantity that actually matters is not the bare statement that the two profiles match, it is the statement of how rare the matching profile is in the relevant population, because a match between two profiles is meaningful only to the extent that the profile in question is not expected to occur at significant frequency in the population from which a hypothetical alternative contributor might be drawn. The relevant statistical quantity for single-source samples is the Random Match Probability, which expresses the likelihood that a randomly selected, unrelated member of the reference population would share the observed profile purely by chance, and for a complete single-source profile measured across sixteen to twenty-four STR loci, this probability is typically expressed as one in ten to the fifteenth power or smaller, numbers so extreme that they exceed the total human population of the planet by several orders of magnitude.

These numbers, however stunning they sound when read out in court, depend on a series of assumptions about the underlying population, about the independence of the loci, about the absence of close relatives among the candidates, and about the integrity of the sample and the laboratory process, and they are valid when those assumptions hold and essentially meaningless when they do not. The statistical framework also changes fundamentally when the sample in question is a mixture, because the comparison is no longer between a single measured profile and a reference but between a set of profiles jointly consistent with the observed data and a set of hypotheses about which contributors might be present, so that the appropriate statistical quantity for mixture cases is the Likelihood Ratio, which compares the probability of observing the measured data under the hypothesis that the suspect is a contributor against the probability of observing the same data under the alternative hypothesis that the suspect is not a contributor and that the mixture derives instead from unrelated unknowns.

The calculation of a Likelihood Ratio for a complex mixture is not something a forensic analyst can do by hand, it requires specialized software, and the field today relies on a family of tools known as probabilistic genotyping software, of which the most widely used systems include STRmix, developed in New Zealand and Australia, TrueAllele, developed by Cybergenetics in the United States, EuroForMix, developed in Norway, and several others of more limited geographic distribution. These systems use Markov Chain Monte Carlo techniques to model the probability distribution of possible contributor genotypes, and they have been validated across thousands of casework examples, representing a genuine advance in the ability of laboratories to extract meaningful statistical information from samples that, under the interpretive techniques of fifteen years ago, would have been dismissed as inconclusive and set aside.

What is less widely understood, and what anyone reading about a DNA prosecution genuinely needs to know, is that these systems are not interchangeable with one another. In a 2023 case study published in the Journal of Forensic Sciences by William Thompson and colleagues, the identical item of DNA evidence from a federal criminal case was analyzed using both STRmix and TrueAllele, under the same protocol and with the same reference data, and the two systems produced dramatically different results from the same underlying input. STRmix reported a Likelihood Ratio of twenty-four in favor of the non-contributor hypothesis, meaning that the data were twenty-four times more likely if the suspect had not contributed to the mixture than if he had, while TrueAllele reported a Likelihood Ratio ranging from 1.2 million to 16.7 million in the opposite direction, depending on the reference population used, meaning that the data were between a million and seventeen million times more likely if the suspect had contributed to the mixture than if he had not. The two systems, examining the same physical sample, produced conclusions that differed by a factor of roughly forty million, and the case study traced the discrepancy back to subtle differences in modeling parameters, in analytic thresholds, and in mixture ratio assumptions, with the authors concluding that the analysis of complex mixtures rests on a lattice of contestable assumptions that need to be examined on a case-by-case basis rather than assumed to be standardized across the field.

The German national guidelines, published by the Spurenkommission, the nationally recognized forensic genetics commission responsible for technical standards in Germany, have been updated several times over the past two decades to incorporate these developments, and they now explicitly recommend fully continuous probabilistic genotyping models for the interpretation of mixture traces, including consideration of peak height information in the electropherogram itself. This is the framework under which modern German DNA casework is meant to operate, and it is considerably more conservative and better disciplined than the framework that applied when the field was still in its earlier phases, which is part of the reason why older cases sometimes need to be recalculated under current methodology before their evidentiary weight can be fairly assessed in the present.

The Robbery at IKEA, and What a Mixture Trace Looks Like in Practice

In the Frankfurt armored-transport robbery of the ninth of November 2019, an armed group attacked a money-transport guard at the parking lot of an IKEA store, took a sum of cash, and fled in a vehicle that was subsequently recovered in a burned-out condition some distance from the original scene. According to the public record of the proceedings before the Landgericht Koeln, investigators secured at the rear license plate of the burned-out vehicle a DNA mixture trace that contained material from more than one contributor, and the trace eventually became one of the lines of evidence cited in the prosecution of the principal accused, Thomas Drach, who stood trial on multiple counts including attempted murder and aggravated robbery, was convicted on the fourth of January 2024, and was sentenced to fifteen years of imprisonment with subsequent preventive detention, a sentence that means he will not be leaving custody at any point in the foreseeable future.

I was involved in this investigation during an earlier phase, in the investigative task force that was working through the video material collected from the various crime scenes and from the flight paths taken by the perpetrators, and my contribution to the case concerned a specific constructive feature of the getaway vehicle that became visible in the video record, a feature distinctive enough to substantially narrow the search for the vehicle itself, together with a biomechanical signature in the gait of one of the masked perpetrators, a genu valgus pattern that could be observed in movement and that, in combination with other identifying features, pointed with high likelihood to the same individual across multiple pieces of video material recovered during the investigation. I am not going to go into the technical specifics of either observation in this article, because the case is closed, the conviction is final, and the professional obligations attaching to my earlier role as the sworn expert in the proceedings remain in force, but what I am willing to say is that the DNA mixture trace at the rear license plate was, on the publicly available record, only one of several lines of evidence in the case, and that the convergence of independent forensic methods, including DNA analysis, biomechanical gait analysis, vehicle identification, and others, is what produced a record robust enough to survive the kind of intensive legal challenge that the defense team in this particular trial mounted against the prosecution.

This is the more general point that the proceedings illustrate, and it is the reason I am referring to the case at all in an article about DNA, because forensic identification in serious criminal investigations does not usually rest on a single DNA match and nothing else, but on a convergence of independent lines of evidence, each of which has its own sources of error and its own statistical weight, and which, taken together, produce a case whose reliability depends on the independence of those lines from one another. When a DNA trace is a mixture, and its statistical interpretation is contingent on the software used to analyze it, the weight that the trace can carry on its own is less than what the raw match probability would suggest in isolation, but when that trace is combined with biomechanical, video-forensic, vehicular, circumstantial, and other evidence pointing in the same direction, the combined weight can become genuinely substantial. This is how serious forensic casework is actually supposed to work, and it is also the reason why the popular image of DNA as a single unanswerable piece of evidence that alone determines guilt or innocence is a real distortion of how the field operates at its best.

I should add, because the point is genuinely relevant in this context, that after I withdrew from the proceedings on grounds of prejudice during the trial itself, my technical findings were reviewed by two independent successor experts appointed by the court, one of them an anthropologist and the other a sports scientist, both qualified to assess the biomechanical dimensions of the analysis I had originally prepared, and both arrived at conclusions consistent with those I had reached in my own work. The case is, from a forensic standpoint, an example of a converging multi-methodological reconstruction rather than a single-DNA prosecution, and the evidentiary record is correspondingly stronger for it.

When DNA Is Wrong: The Heilbronn Phantom

The single most instructive example of what can go wrong with DNA evidence in a major European jurisdiction is the case of what came to be known, in the German and international press, as the Phantom of Heilbronn, and the story is worth recounting in some detail because it captures, with unusual clarity, the structural problems that arise when DNA evidence is treated as a single decisive line rather than as one input among several. Between 1993 and 2009, investigators in Germany, Austria, and France recovered DNA from a single unknown female at no fewer than forty separate crime scenes, with the offenses spanning a range from minor burglaries to the execution-style murder of the police officer Michele Kiesewetter in Heilbronn on the twenty-fifth of April 2007. The sheer geographic and typological diversity of the crimes attracted increasing attention over time, police across three countries pooled their resources, a reward of three hundred thousand euros was offered, and the investigation eventually accumulated, by the time it finally collapsed, approximately two million euros in direct costs and sixteen thousand overtime hours distributed across multiple police agencies in three countries.

There was, in fact, no serial killer behind any of it. In March 2009, investigators trying to identify a burned corpse re-examined fingerprints that had been taken from a male asylum seeker's application document some years earlier, and they found the Phantom's female DNA profile sitting on those fingerprints, which was straightforwardly impossible on any sensible interpretation of the evidence. When the test was repeated using a different cotton swab from a different supplier, the Phantom's DNA profile no longer appeared in the result, and the source of the contamination was eventually traced to a factory operated by Greiner Bio-One International in Austria, where sterile cotton swabs used by many European police forces were packaged by a small workforce that included women of eastern European origin, the demographic that happened to be consistent with the mitochondrial lineage that had been inferred for the hypothetical serial killer. The sterilization process used during manufacture was sufficient to kill bacteria, viruses, and fungi, but it did not destroy human DNA at all, and the skin cells, saliva droplets, or other bodily residues that had been deposited during packaging remained on the swabs in quantities too small to see with the naked eye but well above the one-hundred-picogram threshold required to generate a forensic profile.

As a consequence of the Phantom of Heilbronn case, the International Organization for Standardization published in 2016 the standard ISO 18385, which establishes production requirements for forensic consumables certified free of human DNA contamination, and the standard is now widely applied in European and North American forensic supply chains, representing a structural response to a systemic problem that had remained invisible until the contamination became impossible to ignore any longer. The deeper lesson of the case, however, is not really a technical one about cotton-swab manufacturing, it is a lesson about the way forensic DNA evidence accumulates interpretive momentum, and about how difficult it becomes to reverse a narrative once investigators, prosecutors, and the press have committed themselves to it. For nearly fifteen years, every DNA hit that matched the Phantom's profile was interpreted as fresh evidence that a single offender was responsible for an increasingly bizarre range of crimes, because the profile was the only piece of physical evidence that linked the crimes to one another, and because the profile appeared, at the level of analysis available at the time, to be unambiguous. The possibility that the profile itself was the actual problem was systematically resisted by everyone involved, because acknowledging it would have required the investigators to abandon a commitment that they had been building for years, and the eventual confrontation with the truth came only when a test result was so manifestly impossible, the presence of a female profile on a male asylum applicant's fingerprints, that further resistance simply became untenable.

The full cost of that interpretive commitment, measured in euros, in hours, in misdirected resources, and, most importantly of all, in the forty genuine criminal investigations that were sidelined while the phantom was being pursued, has never been fully tallied in any official document. The Phantom of Heilbronn is not, in my own view, primarily a story about laboratory contamination at all, it is a story about institutional epistemology, about what happens when a single line of evidence is permitted to dictate the interpretive framework of an entire investigation, and about how much easier it always is to follow a false trail than to question the trail itself.

When DNA Is Wrong: The Paramedic, the Pulse Oximeter, and the Homeless Man

The second case worth examining in some detail is American, and it ran approximately concurrent with the later phases of the Heilbronn investigation, providing an almost laboratory-clean example of how secondary DNA transfer can produce a wrongful accusation against a person who could not possibly have committed the crime in question. On the twenty-ninth of November 2012, a Silicon Valley investor named Raveesh Kumra was bound, gagged, and killed during a home invasion at his residence in Monte Sereno, California, and the autopsy determined that he had died from asphyxiation caused by the duct tape that had been used to gag him. A DNA sample recovered from beneath his fingernails was analyzed shortly afterwards, and it matched the profile of a twenty-six-year-old homeless man named Lukis Anderson, who was living in San Jose, approximately ten miles from the crime scene, and Anderson was arrested on a charge of first-degree murder, an offense which, in California at that time, carried the possibility of the death penalty.

Anderson, a chronic alcoholic with episodic memory problems of long standing, initially told his public defender that he did not remember having committed the murder but could not entirely rule it out either, because he had been drinking heavily and did not remember the night in question at all, and his public defender, Kelley Kulick, did what public defenders are sometimes accused of failing to do, which was to take his alibi seriously anyway. Her investigator reconstructed Anderson's movements on the evening of the killing, using receipts, witness statements, and police dispatch records, and what they eventually found was that Anderson had been publicly intoxicated in front of a San Jose convenience store in the early evening hours, had collapsed on the sidewalk, and had been transported to the Santa Clara Valley Medical Center for medical treatment. He had been admitted to the hospital at approximately eight in the evening, was placed under continuous monitoring with bed checks at fifteen-minute intervals throughout the night, and was still in the hospital when Mr. Kumra was killed more than three hours after Anderson's own admission.

An alibi backed by continuous hospital monitoring is not the kind of alibi that a prosecutor can simply wave away as inconvenient, but the DNA evidence was equally unambiguous in the opposite direction, and the case was not going to close until the apparent contradiction had been fully resolved. The resolution came when the defense team examined the dispatch records of the local emergency medical services, and they discovered that the same two paramedics who had transported Anderson to the hospital that evening had, several hours later, responded to the Kumra residence when Kumra's body was discovered there. They had used the same equipment across both calls without any intervening decontamination, including a pulse oximeter that had been slipped onto Anderson's finger during the transport and was subsequently applied to Kumra's finger during the examination of the body at the crime scene some hours later. Anderson's skin cells, deposited on the pulse oximeter during his treatment, had been transferred from his finger onto Kumra's finger three hours after the original contact, from where they migrated to the underside of Kumra's fingernails during the struggle or during the pathologist's examination of the body, and they were then recovered in a DNA collection whose results pointed unambiguously at a man who had been unconscious in a hospital bed under continuous observation at the precise moment of the killing.

The charges against Anderson were dismissed once this transfer pathway was established, three other men were subsequently convicted of the Kumra murder, Anderson eventually returned to the streets, and the case was written up in the legal and forensic literature as an unusual example of secondary DNA transfer in the context of a wrongful accusation. It was, in fact, an example of exactly what the transfer research had been predicting for years already, and exactly what practitioners in the field had been warning about to anyone who was willing to listen, because the pulse oximeter was not a rare or exotic vector for DNA transfer at all, it was a routine piece of emergency medical equipment used across two calls on the same shift by the same two personnel, with DNA adhering to its surface in quantities well above the detection threshold of modern forensic techniques. What the Anderson case ultimately illustrates is not that DNA evidence is inherently unreliable, it is that the question of whose DNA is present at a crime scene and the question of how that DNA actually got there are two entirely separate questions, and that courts and juries have systematically confused them with one another. A DNA match tells you whose material is present at a particular location, it does not tell you anything at all about how that material arrived there, and in an era in which the detection threshold is measured in single-digit cell counts, the path of transfer is no longer a minor side detail that can simply be assumed away by the prosecution. It has become the actual question, and the failure to ask it is precisely how innocent people end up in custody for crimes that were committed by other people entirely.

The Petrous Bone, or Why the Dead Speak Longer Than Anyone Expected

The discussion up to this point has focused on DNA in the context of fresh or relatively recent biological material, which is the everyday terrain of contemporary forensic casework, but there is a second domain in which DNA analysis has undergone a quiet revolution over the past decade, and it has consequences for both archaeology and cold-case forensic practice that most readers will not have encountered in the popular coverage. The revolution concerns one specific anatomical structure within the human skull, and it begins with a question that initially sounds purely academic in nature: when a body has decomposed entirely, when the soft tissues are long gone, when even the bones themselves have begun to disintegrate under the pressure of years or centuries spent in the ground, is there anywhere within the remaining skeletal material where usable DNA can still be recovered with any reliability? The answer, arrived at through a sequence of landmark studies over the past decade, is yes, and the answer is specifically the petrous portion of the temporal bone, which in German is called the Felsenbein, the rock bone, a name that turns out, on closer examination, to be more of an understatement than an exaggeration.

The petrous portion of the temporal bone is the densest, the hardest, and the most heavily mineralized bone in the entire human body, sitting at the base of the skull between the sphenoid and the occipital bones, and its primary biological function is to house and to protect the delicate structures of the inner ear, the cochlea and the vestibular apparatus, which are responsible for hearing and for balance respectively. The extreme density of the petrous is an evolutionary adaptation to this protective function, since the inner ear structures cannot tolerate either deformation or impact, so the bone that surrounds them is built like nothing else in the entire human skeleton, and during life this density provides acoustic isolation and physical protection. After death, however, it turns out to provide something rather more interesting, namely an exceptionally stable environment for the preservation of endogenous DNA over timescales that can extend from decades all the way out to several millennia.

The foundational study was published in PLOS ONE in 2015 by Ron Pinhasi and his colleagues at University College Dublin, in collaboration with researchers at several institutions across Europe and the United States, and Pinhasi's group examined the distribution of endogenous DNA, meaning DNA originating from the deceased individual rather than from environmental or microbial contamination, across different regions of the petrous bone and across different bone elements within the same archaeological skeletons. What they found was that the dense cortical region surrounding the otic capsule, which is the bony housing of the inner ear itself, yielded up to one hundred and eighty-three times more endogenous DNA than other skeletal elements taken from the same individual, and that within the petrous bone itself, the densest central region of the otic capsule yielded up to sixty-five times more DNA than the less dense outer cortical layer and up to one hundred and seventy-seven times more than the spongy trabecular apex of the bone. The density of the bone, in other words, corresponds directly to the preservation of the DNA contained within it, and the petrous bone is anomalously dense even among the dense bones of the human body, which is the underlying reason why its DNA yields are so far out of proportion to what other skeletal elements can provide.

The implications of this finding have been worked out over the past decade in a sequence of methodological papers by the Pinhasi group itself, by the David Reich laboratory at Harvard, and by the ancient DNA laboratories at the Max Planck Institute for Evolutionary Anthropology in Leipzig, where Matthias Meyer and the late Svante Pääbo developed the extraction and sequencing techniques that made large-scale ancient genome recovery technically feasible. A 2019 paper by Daniela Gaudio and colleagues, published in Scientific Reports under the title "Genome-Wide DNA from Degraded Petrous Bones and the Assessment of Sex and Probable Geographic Origins of Forensic Cases," demonstrated that petrous bone sampling from the inner ear cochlear region, when combined with next-generation sequencing protocols, could yield endogenous DNA percentages between roughly fifteen and sixty-seven percent even in burned remains, a context that would render every other skeletal element effectively useless for DNA recovery purposes. A 2023 paper in Genome Research described a density-separation pretreatment that further increases the endogenous DNA yield from petrous powder by up to a factor of five relative to standard extraction protocols. The field is moving forward at considerable speed, and the implications for cold-case forensic work are substantial, because a skeletonized human body, recovered decades after death from conditions that would destroy any other form of biological identification evidence, retains within its temporal bone a reservoir of endogenous DNA sufficient to produce not just a forensic profile of the conventional kind but, with modern next-generation sequencing, a complete genome-wide genotype with population-ancestry resolution and the capability for close-kin matching as well.

This matters considerably for my other principal area of professional interest, which is the forensic and legal framework around the possession, the analysis, and the identification of historical human remains. Some readers will recall that I published earlier this year a guide on the legal situation around acquiring and possessing human skulls in Germany and in the United States, and that the guide discussed at some length the question of how a forensic practitioner distinguishes a historical specimen from a recent one in the absence of provenance documentation. The technical methods I described in that earlier guide, namely morphological assessment, dental analysis, histomorphometry, and radiocarbon dating, all remain the standard non-genetic toolkit for such questions, but what has changed in the past few years, and what is likely to change still further over the coming decade, is that a petrous-bone DNA extraction, which was once reserved exclusively for academic archaeology with unlimited time and budget, is becoming increasingly accessible as a routine forensic option for skeletal remains of uncertain date, of uncertain origin, or of uncertain identity. In practical terms, this means that the dead now speak considerably longer than any previous generation of forensic practitioners would have believed possible, and that the window for identifying individuals from ancient or long-buried remains is widening rather than closing as the underlying technology continues to mature.

The Legal Architecture in Germany, Briefly

Everything described up to this point in the article is technical in nature, but the question of what German law actually permits forensic laboratories to do with DNA evidence, and what it permits investigators to do with the resulting profiles afterwards, is a separate matter entirely, and it deserves at least a brief treatment here because the reader who wants to follow German criminal casework with any degree of comprehension needs to understand the basic statutory framework. The legal foundations for forensic DNA analysis in German criminal procedure are laid down in a small group of provisions within the German Code of Criminal Procedure, the Strafprozessordnung, principally Sections 81a, 81e, 81f, 81g, and 81h.

Section 81a paragraph 1 governs the physical examination of the accused, including the taking of blood and cellular samples for the purpose of establishing facts of relevance to the proceedings, and it permits such examination to be carried out without the consent of the accused, although this may only occur subject to a judicial order or, where there is imminent risk to the investigation, an order issued by the public prosecutor or by police investigative officers in their capacity under the relevant provisions.

Section 81e is the central provision that authorizes molecular-genetic analysis of biological material in criminal proceedings. Under Section 81e paragraph 1 first sentence, the analysis of material obtained through measures under Section 81a paragraph 1 or under Section 81c is permitted for the establishment of the DNA identification pattern, the parentage, and the sex of the person, with comparison against reference material insofar as this is necessary to investigate the facts of the case, and Section 81e paragraph 1 second sentence expressly prohibits any other findings, with any investigation directed at other genetic properties of the sample being declared inadmissible in clear statutory language. Under Section 81e paragraph 2 first sentence, the same kind of analysis may be conducted on material that has been found, secured, or seized as trace evidence, and under Section 81e paragraph 2 second sentence, where the source of the trace material is unknown to the investigators, additional findings as to the eye color, the hair color, and the skin color of the contributor, together with the contributor's age, may also be established, an extension of the original framework that was introduced by legislative amendment effective from the end of 2019.

Section 81f sets the procedural rules for the conduct of molecular-genetic analysis itself, with paragraph 1 establishing that, where the person from whom the material derives has not consented to the analysis, the procedure may be ordered only by a court, with the limited exception that, in cases of imminent risk to the investigation, the public prosecutor or police investigative officers may make the order themselves. Paragraph 2 of the same Section requires that the analysis be conducted by a court-appointed and sworn expert or by a person formally obligated under the relevant statute, that the expert be organizationally separate from the investigative authority itself, and, of particular importance for the protection of the person whose material is being analyzed, that the material be transmitted to the expert without disclosure of name, address, or full date of birth, so that the analysis is conducted on properly anonymized samples whenever this is at all possible.

Section 81g is, of the entire group, the most legally consequential, because it governs the establishment and storage of a DNA identification pattern for use in any future criminal proceedings against the same person. Under Section 81g paragraph 1, cellular material may be taken from a suspect of an offense of substantial significance, or of an offense against sexual self-determination, and the material may be analyzed in order to establish the DNA identification pattern and the sex of the person, where there is reason to assume, on the basis of the nature or the execution of the offense, of the personality of the suspect, or of other findings made during the investigation, that future criminal proceedings against the same suspect for offenses of substantial significance are to be expected. The repeated commission of other offenses can equate to an offense of substantial significance for the purposes of this assessment, and Section 81g paragraph 3 establishes a judicial reservation, with the same exception for imminent risk that applies in Section 81f, the provision specifying in considerable detail the requirements for the written justification of the order itself. Section 81h, finally, authorizes the form of mass DNA screening that is known in German practice as a DNA-Reihenuntersuchung, in which, under specific statutory conditions and judicial order, defined groups of persons may be invited to provide samples for elimination purposes in connection with serious offenses under investigation.

The national database that has been established under this overall framework is the DNA-Analyse-Datei, the DNA analysis database, abbreviated as DAD, which has been operated by the Federal Criminal Police Office, the Bundeskriminalamt, since the seventeenth of April 1998, and the database contains two distinct categories of records, namely person records, which consist of identification profiles of known individuals who meet the statutory criteria, and trace records, which consist of profiles obtained from crime scene samples where the original contributor remains unknown. The matching of profiles between these two categories, as well as the matching within each category individually, is the basic operation that the database performs, and published figures indicate that approximately one in three to one in four trace records entered into the database produces a match either to a person record or to another trace record. The general retention and deletion regime for personal data stored by the Bundeskriminalamt is laid down in Section 32 of the Bundeskriminalamtgesetz, which requires correction of inaccurate data and deletion of data whose storage is unlawful or whose knowledge is no longer necessary for the performance of the agency's tasks, with periodic review at intervals that have become customary in police practice of ten years for adult records and five years for juvenile records. The constitutional position of the broader storage architecture under the Bundeskriminalamtgesetz is itself currently in legislative motion, following the decision of the Federal Constitutional Court of the first of October 2024 in case 1 BvR 1160/19, which declared certain provisions of the Bundeskriminalamtgesetz to be incompatible with the Basic Law and required legislative correction by an extended deadline that now reaches into the spring of 2026.

The German framework is, in international comparison, distinctly restrictive in its overall character, since the DNA identification pattern that it permits to be stored is limited to non-coding STR markers, the scope of permitted analysis is narrow rather than broad, and the judicial oversight for storage is explicit rather than merely implicit. The United Kingdom's National DNA Database, by way of contrast, has historically operated under a far broader legal framework, with substantially longer retention periods and broader storage categories, and the American Combined DNA Index System, known as CODIS, operates under federal and state rules that vary considerably across jurisdictions, with many states permitting the retention of profiles obtained from individuals who were merely arrested but never actually charged, a practice that the German legal framework would not permit under any circumstances. Within the German system itself, the architecture provides meaningful safeguards against the most commonly cited civil liberties concerns about DNA databases, and it is considerably harder for a German citizen's profile to end up in the national database without a corresponding substantive legal basis having been established. Whether the framework is genuinely robust enough to address the new generation of privacy concerns that arise from commercial ancestry testing is a different question altogether, and it is one I will return to at the end of this article.

Familial Searching and Forensic Genetic Genealogy

There is one more technical category that needs to be addressed before the final section of this article, and it is perhaps the most consequential development in forensic DNA analysis of the past ten years in any jurisdiction. The category is commonly known in the United States as forensic genetic genealogy, although the technique is more precisely described as forensic investigative genetic genealogy, abbreviated as FIGG, and the method, which has been used to solve dozens of high-profile cold cases in the United States since 2018, operates on a principle that is mathematically straightforward in its construction but whose implications, for anyone who has not yet thought them through carefully, are profoundly disturbing in their reach.

The method works as follows in practice. A forensic laboratory obtains a usable DNA profile from a crime scene sample, typically using protocols that extend beyond the standard short-tandem-repeat panel used for routine casework and instead generate hundreds of thousands of single-nucleotide polymorphism markers across the genome, the kind of dense profile that is used by commercial ancestry-testing companies for their consumer reports. This expanded profile is then uploaded to one of several publicly accessible genealogy databases, historically primarily GEDmatch and FamilyTreeDNA, where it is compared against the profiles of users who have submitted their own DNA voluntarily for ancestry research purposes of their own. The comparison identifies shared DNA segments between the unknown crime scene sample and the registered users of the database, and from the length and the distribution of those shared segments, the degree of biological relatedness can be estimated with high precision, so that a shared segment pattern consistent with third or fourth cousins narrows the pool of possible relatives to a set of extended family trees that can then be constructed using conventional genealogical research methods. Skilled practitioners in the field have demonstrated the ability to identify the original contributor of a crime scene sample with remarkable specificity, often within only a few days of the initial upload to the genealogy database.

Importantly, the regulatory landscape around these databases has shifted considerably since the technique first became public, and the picture today is more nuanced than it was in the immediate aftermath of the Golden State Killer case. GEDmatch, in particular, moved several years ago to an opt-in model under which only users who have explicitly consented to law enforcement matching are made available for forensic searches, and FamilyTreeDNA introduced its own consent and notification framework after public controversy over the original arrangements. The major commercial ancestry-testing companies, including 23andMe and Ancestry, have generally taken the position that they do not voluntarily make their customer databases available to law enforcement except under court order, and they have published transparency reports listing the limited number of legal demands they receive each year. The structural problem, however, is that even with these consent and access restrictions in place, the underlying mathematics of biological relatedness does not change, and a user who has consented to forensic matching exposes the relatives who have not consented to the same downstream identification risk, because what the algorithm reads is genetic similarity rather than authorization.

The method was brought into broad public attention in April 2018, when it was used to identify Joseph DeAngelo, the so-called Golden State Killer, who had been responsible for thirteen murders and more than fifty rapes across California during the 1970s and 1980s, and the DNA recovered from the crime scenes of those decades-old offenses had matched no suspect in any conventional law enforcement database. Once it was uploaded to GEDmatch, however, it matched distant relatives whose publicly available family trees, when combined with investigative genealogy work, led within a matter of months to DeAngelo himself, who was seventy-two years old at the time of his eventual arrest. The case was widely celebrated as a breakthrough in cold-case investigation, and the technique has since been used to identify perpetrators in dozens of additional cold cases across the United States.

The statistical principle that makes this entire method work is the one that deserves real attention from anyone reading the present article, because it does not depend on the original contributor of the crime scene sample having submitted their own DNA to any database at any point. It depends instead on the simple presence, somewhere in the database, of their biological relatives. Each human being shares approximately fifty percent of their DNA with a parent or with a sibling, approximately twenty-five percent with a grandparent, an aunt, or an uncle, approximately twelve and a half percent with a first cousin, and diminishing but still detectable fractions all the way down through second, third, and fourth cousins. Published estimates from recent analyses indicate that once approximately two percent of the population of any given geographic region has submitted their DNA to a publicly searchable genealogy database, essentially every person in that region becomes identifiable through the biological relatives who are already present in the database. The two-percent threshold was crossed in the United States several years ago, and it is now being approached in several European countries as well, as commercial ancestry testing continues to spread through populations that have no institutional or cultural memory whatsoever of genetic privacy concerns.

The Box on Amazon, and Who Actually Pays the Price

Everything up to this point in the article has been forensic science, described as accurately and as dispassionately as I have been able to manage given the topic, but I am going to conclude on a less dispassionate note, because the final category of the issue I want to address is the one that affects not perpetrators or suspects but ordinary, law-abiding people, who have no reason to think about forensic DNA at all in their everyday lives, and who are nevertheless being drawn into the forensic system by a commercial market that they fundamentally do not understand. The market for direct-to-consumer ancestry testing, which is dominated in the United States by companies such as 23andMe, AncestryDNA, MyHeritage, and others of similar profile, offers consumers the opportunity to submit a saliva sample by mail and to receive in exchange a detailed report describing their genetic ancestry, their percentage of Neanderthal admixture, their predisposition to various medical conditions, their response to various medications, and any number of other genetic traits for which the company has developed an algorithmic assessment of one kind or another.

The kits are sold on Amazon, in pharmacies, and through the companies' own websites at prices typically ranging from fifty to two hundred dollars or euros, depending on the depth of analysis that has been requested by the customer, and the customer bases of these companies are by now genuinely enormous in scale, with 23andMe alone having accumulated genotyping data from approximately fifteen million customers at the peak of its market presence, and AncestryDNA operating a database of broadly comparable scale. It is worth noting at this stage that consumer ancestry tests do not, as is sometimes assumed in the popular press, produce a complete sequencing of the customer's genome, but instead generate a single-nucleotide polymorphism profile typically covering between six hundred thousand and roughly one million selected positions across the genome, sufficient for ancestry inference, for trait reporting, and crucially for genealogical kin-matching purposes, but not equivalent in scope to whole-genome sequencing. This distinction matters legally and ethically, but it does not soften the privacy implications, because the SNP positions selected for these panels are precisely the ones that carry the most relationship-relevant information, and a single-nucleotide polymorphism dataset is more than adequate to support familial searching of the kind described in the previous section. These numbers are not abstractions in any sense, they represent a substantial fraction of the adult populations of several developed countries, all of it stored in the commercial databases of for-profit corporations subject to the same commercial pressures that any for-profit corporation is subject to.

In March 2025, 23andMe filed for Chapter 11 bankruptcy protection in the United States, citing declining revenue, the lingering aftermath of a 2023 data breach that had compromised the genetic information of approximately seven million customers, and the failure of the company's subscription and research revenue models to sustain its operations over time. The bankruptcy proceedings placed the company's database, including the genotyping data of all fifteen million customers, into the hands of a court-appointed trustee, who was given the responsibility of evaluating bids for the company's assets, and the subsequent process unfolded over several months in a way that is worth recounting carefully because the public coverage and the actual outcome diverged significantly. In May 2025, Regeneron Pharmaceuticals was initially announced as the successful bidder in a sealed auction with an offer of approximately two hundred and fifty-six million dollars, and many commentators, including initial drafts of this article, treated the Regeneron acquisition as a settled matter. The bankruptcy court, however, reopened the auction at the request of Anne Wojcicki, the co-founder and former chief executive officer of 23andMe, who had organized a non-profit medical research organization called the TTAM Research Institute for the specific purpose of acquiring the assets, and in the reopened bidding TTAM submitted an offer of three hundred and five million dollars. Regeneron declined to raise its bid, the bankruptcy court approved the sale to TTAM at the end of June 2025, and the transaction closed on the fourteenth of July 2025, with 23andMe continuing operations as a non-profit under TTAM's ownership. Customers were given a window during which they could request the deletion of their personal data, and a substantial fraction of them did exercise that option, but the majority of the database transferred with the sale of the company. What TTAM intends to do with the data over the longer term, what its successor entities may do with it if the organization is itself reorganized or dissolved in the future, and what the data may eventually be used for in jurisdictions whose privacy frameworks permit uses that the original customer never contemplated, are all open questions to which there are no definitive answers at the present time, and the fact that the successor is structured as a non-profit committed to the existing privacy policies does not, in itself, alter the underlying mathematics of how genetic information moves through bankruptcy proceedings and corporate transitions.

I raise all of this not in order to argue for or against any particular commercial entity in the marketplace, but in order to illustrate the structural problem that direct-to-consumer DNA testing creates by its very existence, and to explain why the problem is qualitatively different from almost any other privacy concern that modern consumers encounter in their daily lives. Most people believe, reasonably enough, that they have a right to their own DNA, and they are entirely correct in that belief, but what almost nobody has told them is that they are also, by simple biological necessity, the custodians of the DNA of their children, of their grandchildren, and of their great-grandchildren as well, and that they surrendered that custodianship on the day they shipped a saliva sample to a commercial laboratory in exchange for a glossy report on their Italian heritage and their Neanderthal percentage. That data sits now in a database whose owners they cannot control, in a database that has already changed hands once through bankruptcy proceedings, and it will continue to sit there when, thirty or forty years from the present moment, a person who has not yet been born leaves a single cell at a crime scene somewhere in the world. The forensic familial search that is run against that cell will not ask whether the ancestor who contributed the original sample understood what he or she was doing at the time, it will not ask for the consent of the descendants whose identification it enables, the algorithm will simply draw the line through the family tree without pause, and the third-cousin match will pull a person who has never taken a DNA test in their entire life into a forensic investigation, all because their great-grandfather was once curious about his ancestry and decided to trust the privacy policy that happened to be active in 2019.

Anyone who imagines that this is paranoia has not actually looked at the mathematics, which has been available in the published literature for years already at this point. The Golden State Killer case demonstrated the entire method publicly in 2018, the academic research that established the two-percent population-saturation threshold, beyond which every member of a population becomes identifiable through their relatives alone, was published before that case was even in the news, and the 23andMe bankruptcy and the subsequent TTAM acquisition both occurred during the past year, as reported by Reuters, by Bloomberg, and by every other major financial news outlet that covers the biotechnology sector with any seriousness. None of this material is secret in any way, none of it is contested by anyone with relevant expertise, and yet none of it is adequately reflected in the consumer behavior around direct-to-consumer DNA testing, which continues to grow as a market despite the fact that every individual sale represents not the disclosure of one person's private genetic information but the permanent forfeiture of the genetic privacy of an entire multigenerational family line that was never asked to vote on the decision.

The particular feature of this situation that I find most difficult to accept, speaking now as a forensic practitioner who has spent most of his adult life thinking carefully about the consequences of genetic identification for the individuals whose samples pass through laboratories of various kinds, is that the legal framework designed to regulate forensic DNA in most jurisdictions, including in Germany, has almost nothing to say about the commercial channel through which the same genetic information now flows in such large quantities. The German Strafprozessordnung imposes careful constraints on what a public laboratory may do with a citizen's DNA, and on how long that information may be retained, and on which databases it may legitimately be placed, but none of those constraints apply to a German citizen who chooses to send a saliva sample to a commercial company in the United States, which itself operates under a different privacy framework, which may then change ownership through bankruptcy or acquisition, and whose records may eventually end up under the jurisdiction of a successor entity that did not even exist at the time of the original sample submission. The careful legal architecture that protects domestic forensic practice is bypassed entirely by the consumer who chooses to opt into a commercial system that operates outside of it, and the consumer is not told, not clearly, not in language they would actually understand, that the decision they are making applies not to their own privacy alone but to the privacy of the family they belong to today and to the privacy of the family they will eventually leave behind in the future.

Most people believe, as I have already noted, that they have a right to their own DNA, and they are correct in that belief, but what nobody has told them is that they have also become, through the simple act of submitting a saliva sample to a commercial company for ninety-nine dollars, the unilateral agent who has disposed of the genetic privacy of their children, of their grandchildren, and of their great-grandchildren as well, and that the disposition is now effectively permanent in nature, because the data has already been copied, sold, and distributed beyond any individual's capacity to recall it back into private hands. The data will sit in a database that these individuals do not control, and it will still be sitting there when, in forty years from now, a human being who has not yet been born will leave a single cell at a crime scene that happens to match the profile of a third or fourth cousin whose great-grandfather once wanted to know what percentage of him was Neanderthal in origin. The algorithm that compares those two profiles will not pause to ask any moral question about what it is doing, it will simply draw the line between them, and anyone who considers this scenario paranoid has not understood the mathematics that has been openly available in the scientific literature for years at this point, because curiosity about one's own genome is, at the end of the day, the cheapest possible route by which to sell the privacy of an entire family line that has never been consulted at any point about the transaction.

This article is intended for general informational purposes and represents the author's analysis of publicly available scientific literature and of cases whose facts are entirely within the public record. References to the proceedings before the Landgericht Koeln rely exclusively on publicly available reporting and do not disclose any information protected by the confidentiality obligations attaching to the author's earlier role as a sworn expert in the investigation. Nothing in this article constitutes a formal forensic opinion issued in connection with any active criminal proceeding.

References

• Bundeskriminalamt, DNA-Analyse-Datei (DAD), established 17 April 1998.
• Bundeskriminalamtgesetz (BKAG), Section 32 on correction, deletion, and blocking of personal data.
• Bundesverfassungsgericht, Decision of 1 October 2024, 1 BvR 1160/19, on the constitutionality of certain provisions of the Bundeskriminalamtgesetz.
• Fernandes, D.M., et al. (2023). Density separation of petrous bone powders for optimized ancient DNA yields. Genome Research, 33(4), 622.
• Gaudio, D., Fernandes, D.M., Schmidt, R., Cheronet, O., Mazzarelli, D., Mattia, M., et al. (2019). Genome-Wide DNA from Degraded Petrous Bones and the Assessment of Sex and Probable Geographic Origins of Forensic Cases. Scientific Reports, 9(1), 8226.
• Gill, P., Benschop, C., Buckleton, J., Bleka, O., Taylor, D. (2021). A Review of Probabilistic Genotyping Systems: EuroForMix, DNAStatistX and STRmix. Genes, 12(10), 1559.
• International Organization for Standardization (2016). ISO 18385:2016 — Minimizing the risk of human DNA contamination in products used to collect, store and analyze biological material for forensic purposes.
• Legal Tribune Online (2024). LG Koeln verurteilt Thomas Drach zu 15 Jahren Haft mit anschliessender Sicherungsverwahrung, 4 January 2024.
• Marshall Project (2018). Framed for Murder by His Own DNA, 19 April 2018.
• NPR (2025). Judge OKs sale of 23andMe to a nonprofit led by its founder, 30 June 2025.
• Pinhasi, R., Fernandes, D., Sirak, K., Novak, M., et al. (2015). Optimal Ancient DNA Yields from the Inner Ear Part of the Human Petrous Bone. PLOS ONE, 10(6), e0129102.
• Reich, D. (2018). Who We Are and How We Got Here. Interview with the US National Institute of General Medical Sciences.
• Reuters and Bloomberg coverage of 23andMe Chapter 11 bankruptcy, March 2025, the Regeneron asset purchase agreement of May 2025, and the subsequent acquisition of 23andMe by TTAM Research Institute, completed 14 July 2025.
• Scientific American (2016). When DNA Implicates the Innocent.
• Spurenkommission (German Forensic Genetics Commission). Allgemeine Empfehlungen zur Bewertung von DNA-Mischspuren.
• Strafprozessordnung, Section 81a (physical examination of the accused), Section 81e (molecular-genetic analysis), Section 81f (procedure for molecular-genetic analysis), Section 81g (DNA identification establishment for future proceedings), Section 81h (DNA mass screening).
• Thompson, W.C. (2023). Uncertainty in probabilistic genotyping of low template DNA: a case study comparing STRMix and TrueAllele. Journal of Forensic Sciences, 68(3).
• Time Magazine (2009). Germany's Phantom Serial Killer: A DNA Blunder, 27 March 2009.