Human rights investigators are frequently overwhelmed by massive, physically degraded volumes of visual evidence. This review evaluates the technological frameworks driving computational forensics in transitional justice. By analyzing multimodal deep learning methodologies, spanning remote sensing, forensic anthropology, text recognition, and biometric identification, we demonstrate how specialized artificial intelligence (AI) pipelines extract legally actionable intelligence from highly fragmented archives. However, our findings highlight acute evidentiary risks, particularly demographic algorithmic bias and the legal inadmissibility of generative AI hallucinations as substantive standalone evidence. Consequently, our primary contribution is the synthesis of a legally defensible operational framework. To satisfy rigid judicial admissibility thresholds, human-in-the-loop methodologies must be implemented not merely as a technical safeguard, but as a mandatory forensic protocol. We ground this framework in a case study of the NuestramemorIA project, utilizing AI to explore the Chilean dictatorship (1973–1990) archives, ultimately illustrating how combining computational scale with independent human validation establishes robust evidentiary links for legal accountability.
Background Autopsy data regarding fatal interpersonal violence in Morocco and North Africa remain limited. This study describes the demographic, injury, thanatological, and procedural profiles of autopsied fatal assaults in Casablanca and explores factors associated with death within 24hours. Methods Between 2018 and 2025, 4,816 medico-legal autopsies were performed at the institute; 162 deaths met the operational definition of a fatal assault involving the action or omission of a third party, all of which underwent a complete medico-legal autopsy. The principal assault category, dominant lethal injury region, multiregional injury, dominant terminal mechanism, survival, hospitalization, and autopsy interval were analyzed. Rare cells and complete separation phenomena justified the use of exact methods and Firth penalized logistic regression for adjusted analysis. Results The 162 fatal assaults represented 3.4% of the 4,816 autopsies performed. Victims were predominantly male (136/162; 84.0%), and 45.1% were aged 20–39 years. Principal categories were blunt-force trauma (40.1%), sharp-force/penetrating trauma (28.4%), burns (10.5%), mixed mechanisms (8.6%), exposure/neglect (6.8%), mechanical asphyxia (4.9%), and firearm (0.6%). Among 134 cases with a focal traumatic injury, the head was the dominant lethal region in 48.5%, and 35/134 (26.1%) were multiregional. Fatal hemorrhage (40.7%) and fatal cranioencephalic injuries (32.1%) predominated. Fifty-five victims (34.0%) died within 24hours. Multiregional injury was associated with early death in unadjusted analysis among focal cases (19/35 vs. 30/99; OR 2.73; 95% CI 1.24–6.03; p=0.015). Firth penalized regression applied to the 134 focal cases and adjusted for age and sex yielded an aOR=3.23 (95% CI 1.42–7.56; p=0.005) for multiregionality. Regarding modalities, mechanical asphyxia, characterized by 8 early deaths out of 8 cases, produced complete separation; Firth yielded a finite but highly imprecise estimate compared with blunt-force trauma (OR=40.54; 95% CI 4.66–5,337.49; p<0.001). Conclusion This institutional autopsy series is characterized by a young male predominance, blunt-force and sharp-force/penetrating mechanisms, and deaths dominated by hemorrhage and cranioencephalic injuries. The frequency of early deaths and the diversity of delayed complications underscore the need to integrate injury extent, clinical trajectory, and the entire causal chain into medico-legal interpretation.
Forensic DNA methylation-based donor age prediction is currently one of the most researched methods in the field of forensic genetics. Many tools have already been developed for this purpose, focusing on different body fluids, target markers and DNA methylation detection technologies. One such tool created by the VISAGE consortium allows to predict donor age from three different somatic cells, namely blood, buccal cells and bones. This VISAGE tool is based on massively parallel sequencing and requires bioinformatic expertise to analyse the data. Although the VISAGE consortium developed a software for automatic data analysis, this software does not automatically perform quality control steps the user can look at, no visual outputs (such as graphical plots) are generated and it is not conveniently usable if many samples are being analyzed. In this study, we therefore propose an alternative easy-to-use data analysis pipeline consisting of two shiny apps for the VISAGE somatic age prediction tool. The first Python-based shiny app allows for bam file generation and read count extraction from raw data and the second R-based shiny app performs quality controls, visualizations, and allows the user to access all intermediary analysis files. Although age prediction itself can only be performed if the end user has statistical models suitable for the VISAGE data available, these apps facilitate the analysis of the VISAGE age prediction tool, requiring only minimal bioinformatics skills and a high level of transparency how the results were generated.
The retrospective estimation of the immersion post-mortem interval (IPMI) requires continuous water temperature data routinely unavailable in many case series. This study tests the hypothesis that daily water temperature in the regulated Neuquén-Limay-Negro river system (38–40°S, northern Patagonia, Argentina) can be reliably estimated from ERA5-Land reanalysis air temperature. A linear regression was fitted between daily mean water temperature (Twater; Autoridad Interjurisdiccional de las Cuencas, AIC, 2007–2020) and ERA5-Land air temperature (Tair; Copernicus C3S, 2016–2020 calibration period), yielding Twater = 5.45 + 0.60·Tair (R² = 0.811; RMSE = RMSE_cv = 2.04 °C; n = 314). Non-linear alternatives did not improve fit. Seasonal error ranges from RMSE = 1.65 °C (winter) to 2.41 °C (spring). Model coefficients and uncertainty estimates are integrated into an accumulated degree-hour (ADH) framework for IPMI estimation, with analytical uncertainty propagation illustrated by a worked example.
Background Blunt force trauma is a major cause of accidental death worldwide. Falls from height and motor-vehicle collisions represent two common but biomechanically distinct mechanisms in which differences in force direction, load pathways, and energy dissipation are expected to produce characteristic injury patterns. Autopsy-based studies directly comparing these mechanisms within a unified biomechanical framework remain limited.MethodsA retrospective analysis of 50 adult autopsies performed at the El Paso County Office of the Medical Examiner (2016–2024) was conducted. Twenty-five deaths resulted from vertical blunt trauma due to falls from height and 25 from horizontal blunt trauma due to motor-vehicle collisions. Demographic variables were recorded, and fractures were documented by anatomic region and analyzed as binary outcomes. Comparisons between groups were performed using chi-square and Welch’s t-tests, with multivariable logistic regression including trauma mechanism, age, sex, and body mass index.ResultsDemographic characteristics were similar between groups. Vertical trauma demonstrated a significantly higher prevalence of cranial fractures (56% vs 8%, p < 0.001), whereas horizontal trauma showed a higher frequency of rib and sternal fractures (64% vs 28%, p = 0.02). Pelvic fractures were more common in vertical trauma (48% vs 24%) but did not reach statistical significance. No significant differences were identified in femoral, spinal, tibial, or upper extremity fractures. Descriptive analyses suggested that fracture burden increased with body mass index in vertical trauma and decreased in horizontal trauma, while higher fracture counts were observed in females and in decedents aged ≥ 50 years.ConclusionsVertical and horizontal blunt force trauma produce distinct and reproducible skeletal injury signatures that reflect differences in axial versus lateral loading and mechanism-specific energy dissipation. Interpreting regional injury distributions within a biomechanical framework allows autopsy findings to function as an independent tool for mechanism inference when scene data are incomplete and provides a structural basis for understanding the modifying effects of body mass and age in fatal trauma.
The search for unmarked graves at former Indian Residential School (IRS) sites in Canada can be assisted by geophysical surveys, yet these surveys face several challenges. Many of these graves are over a century old and become increasingly difficult to distinguish from the surrounding soil, because as decomposition advances and the disturbed backfill re-equilibrates with the host soil, the dielectric, conductive, and density contrasts that produce detectable anomalies progressively diminish. This study addresses these challenges through a geophysical monitoring program at the Sage Test Facility near Brandon, Manitoba, established and led with the Sioux Valley Dakota Nation under Indigenous protocols, where nine test units were constructed to approximate documented IRS burial styles. These test units ranged from 0.5 to 2.0 m in depth and contained wrapped and unwrapped faunal remains, empty wooden boxes, superabsorbent polymer bundles, or soil only, and each excavation was backfilled with its own excavated soil, homogenized across depths to interrupt the natural stratigraphy. Repeated, time-lapse ground-penetrating radar (GPR), electromagnetic induction (EMI), and Light Detection and Ranging (LiDAR) surveys were conducted over 624 days and were integrated to evaluate how target observability and soil moisture around the test units changed over time. The results demonstrate that soil moisture variations, both natural seasonal changes and an induced water addition over the test units relative to the surrounding undisturbed ground, play a critical role in the discernibility of burial features. For example, shallow targets were detected in up to 75% of cases during the coldest surveys (temperature quartile T1) but in as few as 12.5% during the warmest, driest surveys (T4), and detectability declined markedly following heavy rainfall. The combined GPR, EMI, and LiDAR framework supports the refinement of geophysical workflows for subtle target detection while honoring the Indigenous-led protocols under which the facility was established. This research advances burial detection by demonstrating the benefits of surveying during periods of high soil moisture or frozen ground, of refilling surface depressions to improve GPR coupling, of optimizing GPR frequency and survey-mode selection, and of integrating EMI and LiDAR surveys. These refinements contribute meaningfully to community-led searches, to the identification of unmarked graves, and to honoring the children who disappeared from IRS sites.
Aim The study aimed to analyze the biodemographic, temporal, and joint postural determinants of handwriting speed and legibility, and to explore their potential relevance to forensic handwriting research. Methods A cross-sectional study was conducted among 209 medical students (mean age 21.4 ± 1.5 years; 64% female) at a tertiary teaching hospital in India. Participants copied a standardized text under uniform conditions. High-resolution photographs captured wrist, thumb, and index finger joint angles during writing. Writing speed, fluency, and legibility were quantified using validated metrics. Correlation and multivariable regression analyses identified independent predictors of handwriting performance. Results Mean writing speed was 127.4 ± 22.1 letters per minute, and mechanical fluency was 120.4 ± 22 words per five minutes. Writing speed correlated significantly with the second proximal interphalangeal (PIP) joint angle, whereas legibility correlated with the thumb interphalangeal (IP) joint angle. After adjusting for confounders, the PIP joint angle remained an independent predictor of writing speed (β = 0.32, 95% CI: 0.14–0.51, p < 0.001), and the thumb IP angle predicted legibility (β = 0.27, 95% CI: 0.10–0.45, p < 0.001). Medium of instruction and academic performance were also significant predictors of writing speed (p < 0.001). Conclusion Handwriting performance is governed by joint mechanics and biodemographic variables influencing speed and legibility. These findings provide baseline postural data for Indian writers and may offer preliminary biomechanical reference observations relevant to future forensic handwriting research. However, further validation using authentic forensic handwriting samples is required before practical forensic applicability can be established.
Differential extraction is widely used to recover male DNA from sexual assault evidence, but manual workflows can introduce variability and contamination risk. This validation evaluated the Maxwell® DE System, a semi-automated approach to differential extraction, through multi-laboratory studies conducted in accordance with SWGDAM guidelines at three forensic laboratories. Analytical performance was assessed across sensitivity, reproducibility, stability, workflow flexibility, post-coital interval, and case-type studies. Sperm fraction (Fraction 2) male DNA recovery increased with semen input and reflected the availability of intact sperm cells, with informative autosomal STR profiles obtained at low input levels. At lower inputs tested and extended post-coital intervals, autosomal results were limited; however, partial Y-STR profiles were recovered, supporting their use when male DNA is minimal. Intra-run variability within sites was low less than 30%. Under the conditions evaluated, the system reduced the impact of tested inhibitors and produced comparable results across lysis durations and Fraction 1 processing approaches. Performance was assessed across a range of sample types and reagent blanks. All reagent blanks were below the PowerQuant® System analytical threshold, with no exogenous human DNA detected. Performance was assessed across a range of sample types, and no evidence of exogenous human DNA introduced by the workflow was identified. These findings support the use of the Maxwell® DE System as a semi-automated approach for differential extraction in forensic workflows.
Handprint and footprint evidence recovered from crime scenes can be used to estimate stature and support biological profiling when ridge-detail comparison is not possible. Stature estimation standards for the Punjabi population of Peninsular Malaysia, a culturally distinct minority population with potentially different anthropometric characteristics, have not previously been established. This study developed sex-specific and combined regression models for stature estimation from handprint and footprint dimensions in this population. Ninety-four healthy adult Punjabis (40 males, 54 females; aged 18–50 years) were included following verification of sex classification and age-eligibility criteria. Bilateral handprints and footprints were collected using standardised ink-on-paper methods. Composite foot-length, foot-breadth, hand-length, and hand-breadth scores were derived via principal component analysis to address multicollinearity among outline-length variables, and models were validated using leave-one-out cross-validation. Significant positive correlations were observed between stature and print parameters (p < 0.05; R = 0.355–0.807 in males, 0.398–0.714 in females). The foot-based model explained 68.5% of stature variance in males (R² = 0.685, cross-validated R² = 0.624, SEE = 4.35 cm) and 49.8% in females (R² = 0.498, cross-validated R² = 0.432, SEE = 3.93 cm). Logistic regression classified sex from print dimensions with 93.6% accuracy (AUC = 0.960), supporting sex-specific equations when sex can be inferred from available evidence. These findings represent a population-specific forensic anthropometric reference for Punjabis in Peninsular Malaysia; further validation in independent samples is recommended prior to routine forensic application.
This study investigated the morphological and positional relationships between the mastoid process and the external ear in a Korean population to improve ear modeling in forensic craniofacial reconstruction/approximation. Post-mortem computed tomography (PMCT) scans of 142 Korean adults (106 males and 36 females; mean age, 46.18 years) were analyzed using three-dimensional cranioanthropometric landmarks and reference planes. Correlation analyses and sex-specific simple linear regression models were used to develop prediction equations for external ear dimensions based on mastoid process measurements. The developed models were further evaluated using an independent external validation dataset. Significant correlations were identified between mastoid vertical length and external ear dimensions, as well as between supramastoid crest protrusion and lateral ear protrusion. External validation demonstrated mean absolute errors (MAEs) ranging from 1.15 to 5.06 mm, with most anatomical sections showing prediction accuracies exceeding 80%. Although sex-based differences were significant, age-related differences were relatively small (approximately 1–2 mm), supporting the use of simplified sex-specific models without age stratification. These findings provide population-specific data and a validated methodological framework for predicting external ear morphology in Korean adults. The proposed approach may contribute to more complete and anatomically informed craniofacial reconstruction/approximation for forensic human identification.
Some forensic pathology cause-of-death opinions concern deaths in which the total evidentiary field supports fatal interpersonal violence, although no more specific anatomical or physiological mechanism is certifiable. The formulation “homicidal violence” is used for this problem category, but its evidentiary warrant has not previously been articulated as a case-series-derived forensic pathology opinion-framework. This retrospective descriptive case series aimed to derive an evidentiary-warrant framework for certifying homicidal violence from anonymised forensic pathology records held by the Office of the State Forensic Pathologist, Zambia. The primary analytic dataset comprised cases in which homicidal violence, or an accepted wording variant, was the operative cause-of-death formulation and no more specific mechanism was certifiable. Of 111 records reviewed, 103 met the primary analytic definition. Evidentiary features, post-mortem contribution, ancillary status, exclusionary considerations, and documented limitations were analysed descriptively. Preservation-related or post-burial limitation was present in 79 cases. Mechanism-adjacent detail was present in 15 cases but did not amount to a certifiable mechanism. Post-mortem contribution ranged from positively supportive findings in 66 cases to mainly non-contradictory or limiting findings in 37. Exclusionary considerations were explicit in 42 cases and limited, implicit, or sparse in 61. Six recurrent evidentiary-warrant domains were derived: recorded fatal-violence context, linkage to event, post-mortem contribution, ancillary status and contribution, exclusionary considerations, and documented limitations. The framework organises evidentiary domains relevant to warrant homicidal violence certification across the total evidentiary field. It is not a scoring tool, decision rule, diagnostic instrument, or validation instrument, and requires prospective evaluation.
Next generation sequencing (NGS) has slowly been gaining traction internationally over the last decade in forensic science. The last 5 years has seen an uptick in the usage in laboratories in the United States with the release of numerous commercially available forensic systems, and even National DNA Index System (NDIS) approvals of some of these systems [1]. This research sought to demonstrate concordance and reproducibility between replicate samples processed at three different accredited forensic laboratories. The data demonstrated comparable data can be achieved using the ForenSeq Signature Prep Primer Set B profiling system between laboratories, even when different parameters and thresholds are employed. Additionally, the varying sensitivity of these systems when different quantities of libraries are pooled was also confirmed. No significant differences were noted across laboratories and concordance was achieved with expected profiles for autosomal STRs, Y-STRs, X-STRs, and SNPs.
Nitazene analogues are highly potent new synthetic opioids increasingly implicated in fatal intoxications worldwide. Protonitazepyne (N-pyrrolidino protonitazene) is a recently identified derivative for which toxicological and postmortem data remain extremely limited, particularly in mono-intoxication cases. We report the first fatal mono-intoxication with protonitazepyne in Belgium involving a 22-year-old man. Scene investigation revealed a suicide note, drug paraphernalia consistent with intranasal administration, and powder labeled as protonitazepyne purchased online as a “research chemical.” Postmortem examination demonstrated cerebral and pulmonary edema, right-sided cardiac dilatation, and generalized organ congestion, consistent with severe central respiratory depression. Comprehensive toxicological screening excluded other psychoactive substances. Quantitative analysis using a newly developed and validated LC-QTOF/MS method identified protonitazepyne concentrations of 34.6 ng/mL in peripheral blood, 62.2 ng/mL in gastric contents, and 476.0 ng/mL in bile. The absence of the parent compound and metabolites in urine suggests predominant hepatobiliary excretion. This case also provides the first experimental confirmation of protonitazepyne acid as a metabolite. It confirms the lethal potential of protonitazepyne as a sole agent, expand knowledge of its postmortem concentration ranges, matrix distribution and metabolic pathways. These findings highlight forensic and public health challenges posed by unregulated nitazene analogues and the need for enhanced oversight.
Background In some jurisdictions, courts request hymenal findings to be interpreted as evidence of “virginity” in sexual-offence investigations. Current professional guidance emphasises that hymenal appearance cannot determine sexual history and that the absence of visible injury does not exclude sexual contact or penetration, particularly after delayed presentation. We present an illustrative court-referred adolescent pregnancy case from the Baghdad Medico-Legal Directorate, Iraq, contextualised with aggregated institutional workload data, and propose a structured medico-legal reporting framework for gross-examination settings. Case presentation A 14-year-old student was referred for medico-legal assessment after pregnancy was identified at approximately seven months’ gestation. Gross anogenital examination under bright illumination demonstrated an annular hymen with a smooth free edge and a dominant central opening; a small inferior midline opening was also visible. No acute laceration, active bleeding, or gross discontinuity was identified on inspection. Colposcopy was unavailable, and forensic swabs were not collected because of the prolonged interval since the reported event. Obstetric ultrasonography confirmed an ongoing intrauterine pregnancy. Aggregated administrative workload summaries from the Baghdad Medico-Legal Directorate documented sustained court-referred caseloads from 2020 to 2024, female predominance in years with available sex stratification, and a substantial proportion of minors in 2024. Conclusions This case illustrates a recurrent court-communication problem: pregnancy and delayed presentation may coexist with non-diagnostic hymenal findings, while courts may request categorical “virginity” opinions. A structured medico-legal framework can separate objective findings from interpretation, document timing and method limitations, incorporate broader forensic evidence considerations, and avoid unsupported conclusions about sexual history or legally framed acts.
The potential of handheld, non-destructive devices for detecting ivory from Proboscidea is advancing rapidly. The elemental and structural composition of ivory provides insights into the animals' environments, living conditions, and origins, all of which are highly relevant to forensic science, conservation, and provenance research. This paper aims to provide a comparative literature review of studies on this topic, highlighting current trends and significant correlations. For this purpose, a review of the literature over the last ten years was conducted, with a particular focus on X-ray fluorescence (XRF), which has so far been the primary method used in research on the identification of ivory using handheld devices. As other methods, such as Raman spectroscopy, show promising potential for identifying ivory as well, results from handheld and benchtop instruments have also been included in this review, and the differentiation must be considered. Studies using handheld XRF analyze elemental composition and cover a broader range of topics, including distinguishing ivory from fake ivory, identifying carved ivory, detecting elemental differences in length and cross-section, and assessing sexual dimorphism, whereas research with Raman spectroscopy relies on vibrational modes and focuses more on species identification and age determination. In addition to XRF and Raman spectroscopy, complementary methods were reviewed, including Laser-Induced Breakdown Spectroscopy (LIBS) and Fourier Transform Infrared Spectroscopy (FTIR). Rather than providing an exhaustive overview, this review focuses on illuminating current trends and the most promising approaches in the field.
Objectives To estimate the prevalence of genitoanal injuries and analyze their potential association with age, reported sexual act type (penetrative vs. non-penetrative), and time from assault to forensic examination among female survivors of police-reported sexual assault cases with laboratory evidence compatible with recent sexual contact. Methods Medico-legal reports and laboratory evidence compatible with recent sexual contact (identification of spermatozoa and/or a positive qualitative prostate-specific antigen test) from cases treated at the Bem-Me-Quer Program (São Paulo, Brazil) in 2019 were analyzed. Association analyses were performed using bivariate tests and multivariable logistic regression with covariates selected a priori and a complete-case approach. Results From 297 cases with positive laboratory analysis (microscopic sperm identification and/or PSA detection), genitoanal injuries were documented in 47 cases (15.8%). Most examinations were performed within 24 h of the reported event, and 76.6% of the documented injuries occurred in this subgroup. Genitoanal injury was not significantly associated with age or reported type of sexual act. In an exploratory analysis restricted to cases with reported timing, a ≥ 24 h vs. < 24 h comparison suggested an association, but this finding did not persist after multivariable adjustment. Conclusion Genitoanal injuries were documented in a minority of cases. No independent association was detected between genitoanal injury and age, reported sexual act type, or time from assault to forensic examination; however, this finding, particularly regarding elapsed time, should be interpreted cautiously because of limited statistical power and sparse data in some categories. These findings reinforce the interpretive limits of genitoanal injuries in sexual assault forensics and indicate that they should not be used in isolation to establish or exclude the occurrence of sexual assault.
The widespread proliferation of embedded devices across many critical sectors has made them prime targets for exploitation, creating a clear need for practical firmware forensics with industrial focus. The methodology enables analysis of complex embedded systems, focusing on adaptability and efficiency when time-sensitive analysis is required, and is intended for a broad technical audience. This article (1) introduces hardware-focused forensic methodology for embedded firmware extraction, (2) provides an approach for rapid interface identification via instrumentation taxonomy, and (3) establishes reproducibility criteria for the presented firmware extraction process. By bridging digital forensics with hardware reverse engineering, investigators can efficiently extract and analyze the underlying system data to recover forensic evidence via accessing standardized debugging ports and interacting directly with physical memory chips. The outlined techniques not only encompass both non-destructive and destructive approaches, but also highlight best practices while documenting the extractions for forensic integrity. These presented methods will enable investigators to systematically extract firmware without requiring deep hardware-specific expertise, providing crucial capabilities for uncovering malicious activity and supporting incident response.
Objective To identify drugs in blood and urine samples from victims of sexual assault suspecting exposure to covert administration of drugs, and compare with substances reported taken by the victim. Methods We analyzed forensic blood and urine samples from patients 16 years or older presenting at the Oslo Sexual Assault Centre, Norway, suspecting exposure to covert administration of drugs in relation to the sexual assault. We recorded psychoactive substances reported taken before suspected exposure. Blood samples were analyzed for 64 drugs using ultra-high performance liquid chromatography – tandem mass spectrometry (UHPLC-MS/MS) and an automated enzymatic method for ethanol. Urine samples were analyzed for 39 drugs using liquid chromatography – tandem mass spectrometry (LC-MS/MS), and an additional nine drugs using immunoassay methods. Results We included 50 patients, median age 23 years, 49 were women. Median time from suspected exposure to toxicological sampling was 22 h. Drugs not reported taken were found in 17 (34%) patients, combined with drugs reported taken in nine. In 15 patients we only found drugs reported taken. Unreported drugs found in blood were cocaine in three patients, alprazolam in two, methylenedioxymethamphetamine/methylenedioxyamphetamine (MDMA/MDA) in two, tetrahydrocannabinol (THC) in one, and amphetamine in one. Unreported drugs found in urine were cocaine in eleven patients, MDMA/MDA in three, alprazolam in two, codeine in two, cannabis in one, oxazepam in one, amphetamine in one, and zolpidem in one. Gammahydroxybutyrate (GHB) was not detected in either matrix. Conclusion Unreported drugs were found in 34% of sexual assault victims suspecting covert administration of drugs.
Accurate estimation of the postmortem interval (PMI) is one of the most critical and least reliable applications in forensic science. Methods based on the cooling of the body’s tissues rely on the memoryless property of heat transfer in a constant ambient temperature environment, which can introduce biases into the PMI estimation process. We introduce a memory-based fractional model for PMI estimation that accounts for these biases and provides uncertainty quantification in the model’s predictions. Our model accounts for the effects of body mass, clothing insulation, humidity, ambient airflow, and ground contact in a Monte Carlo process to calculate PMI with uncertainty. A sensitivity analysis helps identify the most significant sources of uncertainty in the model’s estimates of PMI. A case study based on a real forensic investigation indicates the potential value of the model in a forensic investigation while remaining transparent to the courts that oversee those investigations.
Post-PCR purification enhances forensic DNA analysis, particularly for low-template samples. This study evaluated Amplicon Rx purification on PCR products amplified with GlobalFiler. Key interpretation parameters: stochastic, analytical, and stutter thresholds, plus peak height ratios were established for purified samples. Stochastic thresholds for purified samples exceeded 3000 RFU versus < 700 RFU for unpurified samples, underscoring the need for tailored thresholds.Unlike previous studies on single-source samples, this study assesses post-PCR purification in two-person DNA mixtures. We evaluated mixture ratios from 1:19–19:1 at 50 pg and 25 pg total DNA input, representing typical low-template quantities in forensic casework. Using a consensus approach for data evaluation (alleles observed in ≥2 of 3 replicates deemed reliable), purification and concentration via Amplicon Rx consistently reduced allelic dropout and improved peak heights.At 25 pg, unpurified samples exhibited extensive dropout preventing interpretation, while purified samples retained sufficient loci for major contributor typing at ratios ≥ 4:1. At 50 pg, purification enabled recovery of approximately 9 complete heterozygous loci from minor contributors at ratios as low as 1:4 (10 pg minor contributor DNA) versus 3–4 loci unpurified. This improvement is operationally meaningful: ≥ 8 CODIS core loci may enable database searching (NDIS eligibility), whereas profiles below this threshold are typically non-searchable. Major contributors at 50 pg consistently yielded 16–18 complete heterozygous loci following purification.These findings suggest that Amplicon Rx purification can substantially improve allele recovery and profile interpretability in low-template mixtures, increasing the likelihood of generating database-searchable DNA profiles. This supports the potential integration of post-PCR purification into forensic workflows where DNA quantity is limiting.
