The research team, led by Linying Ye, Xuelian Cheng, Guichao Xiao and corresponding authors Chao Liu and Ling Chen, set out to address a well-known weakness in forensic genetics. Human genetic markers such as short tandem repeats are powerful tools for identifying individuals, but they perform poorly when samples are trace-sized or badly degraded, which is precisely the condition of much of the biological material collected in real casework. Microbes, by contrast, are present in enormous numbers, differ markedly from person to person, and change over time in structured ways. These three properties, high quantity, individual specificity and temporal succession, make the microbiome an attractive complement to conventional DNA profiling.

To test how the vaginal microbiome behaves once it leaves the body, the investigators exposed vaginal swabs to a controlled, constant-temperature environment for a full 90 days. Rather than sampling at a handful of arbitrary intervals, they collected data at 13 separate time points across the exposure period, giving them an unusually dense picture of how the microbial community evolved. The study was approved by the Biomedical Ethics Committee of Southern Medical University, and informed consent was obtained from all participants, an important consideration given the intimate nature of the samples involved.

The analytical engine of the study was high-throughput absolute quantitative sequencing, a technique that goes beyond the relative abundance measures used in most microbiome research. Traditional amplicon sequencing tells scientists what percentage of a community each bacterial species occupies, but it says nothing about how many microbial cells were actually present. Absolute quantification adds that missing dimension, allowing researchers to track the total microbial load of a sample as it decays. This distinction proved central to the study’s success, because the two kinds of information, community composition and absolute abundance, turned out to answer two entirely different forensic questions.

When the team examined community structure, they found something initially counterintuitive. Even after three months of exposure to the environment, the overall architecture of the vaginal microbial community remained relatively stable. The broad ecological profile of each sample did not collapse or converge toward a single decayed state. However, within that stable framework, the composition of amplicon sequence variants, the fine-grained taxonomic units that can distinguish closely related bacterial strains, showed significant individual specificity. In other words, each donor’s sample carried a microbial fingerprint that persisted through time, separate from the fingerprints of every other donor in the study.

This persistence of individual signatures opened the door to source attribution. The researchers built a Random Forest model, a machine learning algorithm that combines the votes of many decision trees trained on random subsets of the data, and used it to match exposed samples back to their individual sources. Random Forest classifiers have become a workhorse of forensic microbiology because they handle high-dimensional compositional data well and are resistant to overfitting when properly tuned. In this study, the model was able to accurately identify the individual origin of the samples, demonstrating that the microbial fingerprint survives environmental exposure long enough to be forensically useful.

The second half of the study tackled time since deposition, a parameter that can be critical for reconstructing the timeline of a crime. Here the absolute abundance data took center stage. The researchers observed that the absolute quantity of microorganisms in the exposed swabs decreased over the 90-day period in a clear, time-dependent pattern. Microbial populations do not simply vanish the moment a secretion leaves the body; they decline gradually as cells die off under environmental conditions. By modeling this decline, the team converted microbial load into a kind of molecular clock.

To read that clock, the investigators turned to a Support Vector Regression model, a machine learning approach that fits a continuous output, in this case the number of days since deposition, rather than a categorical label. The SVR model, built on the dynamics of microbial load, achieved a mean absolute error of 8.01 days with a coefficient of determination, or R-squared, of 0.859. In practical terms, the model could estimate the age of a deposited sample to within roughly eight days on average across a three-month window, while explaining about 86 percent of the variance in the true deposition times. For an investigative tool that operates on environmentally exposed trace evidence, that level of precision represents a meaningful advance.

The study fits into a rapidly growing body of forensic microbiome research. Previous work from the same group and others has explored microbial time-of-deposition markers in saliva, semen and bloodstains, and has shown that species-level taxonomic characterization can sharpen the inferential power of stain microbiota. Other laboratories have applied targeted bacterial DNA approaches to saliva, RNA markers to crime scene traces, and spectroscopic or metabolomic methods to blood, urine and semen. What distinguishes the new work is its dual strategy: using community structure for identification and absolute abundance for dating, and applying both to vaginal secretions, a sample type that is crucial in sexual assault cases but has received less TsD-focused attention than blood or saliva.

The authors conclude that their findings clarify the dynamic evolutionary characteristics of both community structure and biomass in vaginal microorganisms after environmental exposure, providing a theoretical basis and practical analytical strategies for the forensic use of vaginal secretion samples. Challenges remain before such models reach the courtroom: real crime scenes present fluctuating temperatures, humidity and substrate effects that a constant-temperature laboratory design cannot fully capture, and larger donor cohorts will be needed to confirm that individual signatures remain distinguishable across populations. Nevertheless, the combination of stable community fingerprints and a quantifiable microbial decay curve offers investigators a genuinely new pair of tools, one that could help attribute intimate trace evidence to a specific person and place it on a timeline, even when human DNA is too degraded to tell the story on its own.

Subject of Research: Forensic application of the exposed vaginal microbiome for individual identification and time since deposition estimation

Article Title: Individual identification and time since deposition estimation based on community structure and absolute abundance of the exposed vaginal microbiome

Article References: Individual identification and time since deposition estimation based on community structure and absolute abundance of the exposed vaginal microbiome. (n.d.). https://doi.org/10.1007/s00414-026-04015-5

Image Credits: AI Generated

DOI: 10.1007/s00414-026-04015-5

Keywords: forensic science, vaginal microbiome, time since deposition, individual identification, absolute quantitative sequencing, Random Forest, Support Vector Regression, microbial load, sexual assault evidence, machine learning, International Journal of Legal Medicine, microbial decay