Beneath the Plaster: Ancient DNA Rewrites the Last Hours of Pompeii

Scientists Were WRONG About Pompeii | Here's What The DNA Shows

Genome-wide sequencing of skeletal fragments embedded in the famous 19th-century body casts overturns long-held assumptions about who Vesuvius killed — and reveals an Imperial Roman city far more cosmopolitan than the family portraits we have imagined for 150 years.

BOTTOM LINE UP FRONT — 

In November 2024, an international team led by the University of Florence, Harvard Medical School, and the Max Planck Institute for Evolutionary Anthropology reported in Current Biology the first genome-wide ancient DNA recovered from skeletal fragments embedded in Pompeii's plaster casts [1, 14]. Of 14 casts sampled during ongoing restoration of 86 specimens, five yielded sufficient nuclear DNA for genome-wide analysis. The genetic results contradict three iconic interpretations: the "mother and child" of the House of the Golden Bracelet was an unrelated adult male holding an unrelated child; the "embracing sisters" of the House of the Cryptoporticus included at least one male and were not maternally related; and four victims long presented as a nuclear family at the Golden Bracelet shared no first-, second-, or third-degree biological kinship. All five sequenced individuals derive most of their ancestry from the eastern Mediterranean — chiefly Anatolian and Levantine Neolithic farming populations — consistent with parallel results from Imperial Rome and confirming the cosmopolitan, mobile character of the Roman Empire in the first century CE. The findings do not erase a tragedy; they correct a Victorian-era story that researchers themselves now describe as "modern assumptions about gendered behaviors" projected onto the dead.

A morning in autumn, 79 CE

The traditional date for the eruption of Mount Vesuvius — August 24, 79 CE — comes from a single source: a letter written by Pliny the Younger to the historian Tacitus roughly a quarter century after the event. That date has been steadily eroded by physical evidence. Autumn fruits, heavier clothing on victims, a coin in the House of the Golden Bracelet whose mint date may post-date August, and most strikingly a charcoal graffito found in 2018 in Regio V dated XVI K(alendas) Nov(embris) — October 17 — together point to a late-October eruption, most often given as October 24–25, 79 CE [1, 2]. The August date persists in popular accounts (and in the YouTube transcript that prompted this article) because the surviving medieval manuscripts of Pliny preserve it, and because a 2022 reassessment by Pedar Foss of the manuscript tradition has reopened, rather than closed, the debate [13]. The Pompeii Archaeological Park itself has oscillated. What is no longer in dispute is that the city died in a single day's worth of sequential pyroclastic events.

The killing mechanism has also been revised. For more than a century, ash suffocation was the standard textbook cause of death. Multidisciplinary work by Giuseppe Mastrolorenzo and colleagues at Italy's National Institute of Geophysics and Volcanology, beginning with a 2010 PLOS One paper, demonstrated that the pyroclastic density currents (PDCs) that reached Pompeii — roughly 10 km from the vent — carried temperatures sufficient to cause instantaneous thermal death even inside buildings [3]. Pompeii's surges struck with temperatures estimated between roughly 250 °C and 300 °C; closer in, at Herculaneum and Oplontis, peripheral surge temperatures reached 500–600 °C. The contorted, "pugilistic" postures preserved in the casts are now interpreted as cadaveric heat spasm, not protracted agony.

That thermal picture was sharpened again in February 2025, when Guido Giordano of Roma Tre University and colleagues published in Scientific Reports the calorimetric analysis of vitrified organic matter recovered from inside the skull and spinal column of a young man found in the Collegium Augustalium at Herculaneum [4]. Differential scanning calorimetry indicated his brain tissue had been heated above 510 °C and then cooled rapidly enough to bypass crystallization — the first reported instance of natural high-temperature vitrification of soft animal tissue. The team attributes the event to a brief, dilute, very hot ash cloud that preceded the main pyroclastic flow. (Skepticism remains: Alexandra Morton-Hayward of Oxford has questioned both the identification and the thermal scenario.) For Pompeii's lower-altitude victims, however, the killing was less exotic: hot enough, fast enough, fatal in seconds.

Fiorelli's invention — and its limits

The plaster casts themselves are an artifact of 19th-century method, not of antiquity. Excavations of Pompeii began in 1748 under the Bourbon kings of Naples but proceeded haphazardly until Giuseppe Fiorelli took charge in 1863. Fiorelli realized that the cavities in the hardened ash were the negative impressions of bodies whose soft tissue had decomposed, and he devised the technique of pouring liquid plaster into those voids through small bore-holes, then chipping away the surrounding matrix. Roughly 104 casts were eventually produced from the estimated 1,000-plus victims recovered at the site.

What Fiorelli's technique did not do was preserve a complete osteological record. When the Pompeii Archaeological Park began a systematic restoration of 86 casts in 2015 and subjected 26 of them to CT scanning or X-ray imaging, the results were sobering. None contained a complete skeleton. Many had been "creatively restored" in the past — bones removed, metal armatures inserted, postures adjusted. The Pilli et al. paper notes dryly that the casts have served as "vehicles for storytelling" and that "stylistic variations between casts in part reflect aesthetic preferences of the periods in which they were made" [5]. The most famous example: a "pregnant woman" whose distended abdomen turned out, on imaging, to be bunched-up garments.

This means the genetic study did not have to overturn a settled scientific record so much as a popular and museological one. The narratives the new DNA contradicts were largely constructed, restored, and curated; they were not derived from rigorous bioarchaeological assessment.

What the sequencers actually did

The methodology, described in the Current Biology paper and its STAR Methods supplement, is forensic in its caution [5]. Sampling occurred at the Pompeii Archaeological Park during cast restoration, accessing fragmented bone and teeth through pre-existing damage in the casts rather than breaching them. Samples were processed at the Molecular Anthropology Unit of the University of Florence — a dedicated ancient-DNA clean facility — with outer surfaces mechanically abraded and ultraviolet-irradiated to suppress modern contamination.

DNA extracts of the first set of six samples were quantified using the Quantifiler Trio kit. Illumina sequencing libraries were prepared in two formats: non-UDG-treated (preserving the deamination damage patterns that authenticate ancient DNA) and partial-UDG-treated (cleaner, suitable for capture). At Harvard Medical School, libraries were enriched in solution for the mitochondrial genome plus roughly 3,000 nuclear screening SNPs, then for those that passed, for the standard 1,237,207-SNP "1240K" capture panel. Sequencing ran on Illumina MiSeq and NextSeq 500 instruments. Of 14 sampled casts, seven yielded enough DNA to attempt 1240K capture; five produced data covered on more than 50,000 SNPs and were retained for population genetic analysis [5].

For one individual — Cast 25 from the Villa of the Mysteries, the best-preserved cast in the set — a lower premolar was also processed at the University of Florida for strontium and oxygen isotope ratios via thermal ionization and isotope-ratio mass spectrometry, providing a complementary signal of childhood residency [5].

Three corrected stories

The House of the Golden Bracelet. Excavated in 1974, this terraced villa in Insula 17 of Regio VI yielded four victims long presented as a nuclear family fleeing toward the seafront. Cast 52 — the adult on whom a 6.1-gram gold bracelet was found, giving the house its name — was traditionally identified as a mother because of the bracelet and because a young child (Cast 51) appeared to be on the adult's hip. Cast 50, an adjacent adult, was cast as the father; Cast 53, a four-year-old child found nearby, as a son. DNA quantification using the Quantifiler Trio kit's Y-target showed all four were biologically male. Where nuclear coverage was sufficient (Casts 51, 52, 53), the result is unambiguous: XY karyotypes. Crucially, BREADR and KIN relatedness analyses found no biological kinship up to the third degree among any of the four. They were not a family. Mitochondrial haplogroups (U1a1 for Cast 52; T2c1c for Cast 51; H for Cast 53) further rule out maternal lineage [5]. The bracelet, in this revised reading, simply reflects what historians of Roman material culture have long noted: high-status Roman men wore gold.

The House of the Cryptoporticus. Excavated in 1914, this house in Insula 6 of Regio I produced a pair of victims — Casts 21 and 22 — found in what archaeologists described as an embrace. The narrative of "two sisters," "mother and daughter," or "lovers" entered the popular literature without any osteological sex determination. CT analysis estimated Cast 21 at 14–19 years old and Cast 22 as a young adult, but produced no reliable sex attribution. Genetic analysis succeeded for Cast 22, identifying him as male (Y-haplogroup J2b2a1, mtDNA N1b1a1). Cast 21 yielded only mitochondrial data, but its haplogroup R is incompatible with N1b1a1, ruling out a mother–daughter relationship [5]. The intimacy of the pose, in other words, is a fact of physical proximity in the moment of death, not of biological or necessarily social relationship.

The Villa of the Mysteries. Cast 25, found alone on a layer of ash on the upper floor of the farm wing with an iron ring, five bronze coins, and a whip, has been interpreted since excavation as the villa's faithful steward. Genetics confirmed male sex (Y-haplogroup E1b1b1b1b, mtDNA H). Strontium analysis returned 87Sr/86Sr = 0.7084729 ± 0.00001, compatible with the southern Campanian plain (0.7075–0.7085); δ¹⁸O of 26.77‰ VSMOW is consistent with coastal central Italy [5]. The isotopes do not exclude a Pompeian childhood, but neither do they require it; similar geochemical signatures recur across the Mediterranean. His genome-wide ancestry suggests a mixed Eastern Mediterranean and European origin, distinct from the other four sequenced individuals.

"At two of the villas we analyzed, individuals previously assumed to be women, in absence of careful osteological assessment, were found to be genetically male... These discoveries challenge longstanding interpretations, such as associating jewelry with femininity or interpreting physical closeness as indicators of biological relationships." — Pilli et al., Current Biology, 2024 [1, 12]

A cosmopolitan port

The ancestry results align with a picture that has been emerging from ancient-genomics work on Imperial Rome over the past five years. On principal component analysis projected against modern West Eurasian and worldwide reference panels, the five Pompeian genomes plot away from modern Italians, Iron Age Italians, and contemporaneous Etruscans, and cluster instead with eastern Mediterranean and Levantine populations. ADMIXTURE analysis at k=6 places them close to Imperial Roman individuals from central Italy and to contemporaneous individuals from the Aegean and Anatolia [5].

Formal qpAdm modeling using distal source populations attributes 48–75% of ancestry in each individual to Anatolian and/or Levantine Neolithic farmers, with most of the remainder from Iranian/Zagros Neolithic farmers. Cast 52 is an outlier: he is best modeled as roughly 58% Levantine Pre-Pottery Neolithic and 42% Iranian Neolithic, with no Anatolian Neolithic contribution — a profile most parsimoniously explained by recent Levantine ancestry, possibly Hellenistic Egyptian. Phenotype prediction using HIrisPlex-S indicates Cast 52 likely had black hair and dark skin; Casts 25, 51, and 53 likely had brown eyes [5]. The single individual (Cast 52) with sufficient genome coverage to evaluate runs of homozygosity (ROH) showed only one short ROH — inconsistent with consanguinity or origin in a small founding population, and consistent with a large, mixed urban gene pool.

This matches what historians have long inferred from inscriptions, trade goods, and the writings of authors such as Strabo and Tacitus: that Pompeii, a port at the mouth of the Sarno river, was a node in an empire whose population was constantly redistributed by commerce, military service, and the Roman institution of slavery. The previously published whole genome from a victim recovered in the Casa del Fabbro [6] showed the same eastward-shifted ancestry, as did Antonio et al.'s landmark 2019 Science paper on the genomic history of Imperial Rome [7].

Pompeii is still being uncovered

The DNA paper landed in the middle of an unusually productive period for Pompeii's archaeologists, working under park director Gabriel Zuchtriegel — himself a co-author of the genetic study and the official authority who must approve sampling. A non-exhaustive list of recent finds:

• January 2025: Excavation of an unusually large private bathhouse complex in Regio IX, capable of accommodating around 30 people across calidarium, tepidarium, frigidarium, and a cold plunge pool [8].
• February 2025: Discovery of a near-life-size Dionysian frieze on the walls of a banquet room, depicting a Bacchic mystery cult procession.
• April–May 2025: The "House of Helle and Phrixus" yielded the remains of four victims, including a child, who had barricaded a bedroom door with a bed frame against the inrushing lapilli [9].
• August 2025: Excavations in the Insula Meridionalis, published in the E-Journal of the Excavations of Pompeii, documented sustained reoccupation of the city's upper floors and cellars from the late first through fifth centuries CE — including a ceramic lamp bearing an early Christian symbol — overturning the long-held assumption that the site was simply abandoned after 79 CE [10, 11]. This finding has a forensic implication for the genetic and isotopic work: not every organic trace recovered above the destruction layer necessarily belongs to a 79 CE victim.
• December 2025: Reconstruction work at the Casa del Tiaso (House of the Thiasos) and CNN reporting indicated evidence of a multi-story tower in a luxury residence — architecturally unprecedented for Pompeii [15]. New finds at the Villa Poppaea at Oplontis, traditionally associated with Nero's wife Poppaea Sabina, included a peahen fresco, an Atellan-comedy theatrical mask of Pappus, and four previously unknown rooms [16].
• April 2026: The Pompeii park, in collaboration with the University of Padua, released an AI-assisted facial reconstruction of a victim from the Porta Stabia necropolis who had attempted to shield his head from falling lapilli with a terracotta mortar [17].

None of these finds individually overturns the genetic results, but together they reinforce the methodological argument the Pilli et al. team made implicitly: any single interpretive lens — archaeological, osteological, genetic, isotopic — gives a partial view. The casts read as a tableau of Roman family life; the genomes read as a port city of strangers; the new excavations read as an "invisible" post-eruption favela. All three are true at once.

The instrument behind the result: Illumina sequencing in ancient DNA

The Pompeii study is one application of a now-standard toolkit — instruments, chemistries, library protocols, capture reagents, bioinformatic pipelines, and curated databases — that has reshaped the study of the human past since roughly 2010. None of the science in the Current Biology paper is possible without that infrastructure, and a few sentences of background help explain both why the result took so long to obtain and why it is trustworthy.

Sequencing chemistry. Illumina's platforms all use sequencing-by-synthesis (SBS) with reversible-terminator chemistry, originally developed at Solexa and acquired by Illumina in 2007. DNA fragments are immobilized on a flow cell, amplified into clonal clusters by bridge PCR, and then sequenced by sequential incorporation of fluorescently labeled, 3′-blocked nucleotides; after each cycle the fluorescent tag and the blocking group are cleaved and the next base added [18]. Because each cluster is read in parallel, a single run yields hundreds of millions to tens of billions of short reads — typically 75–150 bp paired-end. The newer NovaSeq X and NextSeq 1000/2000 platforms run on Illumina's "XLEAP-SBS" chemistry, introduced in 2023, which improves read quality and throughput per run.

Ancient DNA work, however, almost never needs the highest-throughput platform. Endogenous DNA in archaeological bone is usually fragmented to fewer than 100 base pairs, often fewer than 50, and is present in vanishingly small quantities relative to environmental and microbial DNA. The Pilli et al. study used the Illumina MiSeq for the initial mitochondrial-capture screen at Florence and the Illumina NextSeq 500 at Harvard for genome-scale capture data — both v2 chemistry, paired-end 2 × 76 cycles [5]. These are mid-tier benchtop instruments, well matched to the read lengths the input material can support.

Library preparation. Before any sequencing happens, fragmented ancient DNA must be converted into a "library" — a set of molecules with the right sequencing adapters at each end so the flow cell can capture them. Two protocol families dominate aDNA work. The double-stranded protocol of Meyer and Kircher (2010, Cold Spring Harbor Protocols) is what Pompeii's samples received: blunt-end repair, ligation of two adapters in a single reaction, and a fill-in step, with unique dual indexes added by PCR for sample identification [19]. The Pilli team used both fully untreated libraries (which preserve the C-to-T deamination damage at fragment ends that authenticates ancient DNA) and partial-UDG-treated libraries, in which uracil-DNA glycosylase removes most damaged uracils from the interior of reads while leaving terminal damage intact for authentication [5, 20]. The alternative single-stranded protocol developed by Gansauge and Meyer at the Max Planck Institute for Evolutionary Anthropology (2013, refined 2017) uses CircLigase II or T4 DNA ligase to attach adapters strand-by-strand, recovering ultra-short fragments that double-stranded methods miss; it is the workhorse for Neanderthal, Denisovan, and Sima de los Huesos hominin sequencing [21]. Pompeii didn't need it — the bone fragments inside the casts, while degraded, were young enough (ca. 1,950 years) to yield workable double-stranded libraries.

In-solution SNP capture. Direct shotgun sequencing of a Pompeian library would waste roughly 99% of reads on environmental DNA. The standard solution since 2015 is targeted enrichment: synthetic biotinylated RNA or DNA "baits" complementary to a panel of informative human SNPs are mixed with the library, captured on streptavidin beads, and washed clean of off-target sequence. The reference panel that the Reich Lab established and that has been used in roughly 70% of all published aDNA studies to date is the so-called 1240K reagent — 1,237,207 SNPs across the autosomes plus targeted Y-chromosomal and ancestry-informative sites [22]. The Pilli team used two rounds of 1240K capture on partial-UDG libraries, after a smaller pre-screen on roughly 3,000 SNPs plus a mitochondrial bait set originally described by Maricic et al. (2010) [5].

Because synthesizing 1240K baits in-house was prohibitively expensive for most laboratories, in 2021 two companies — Daicel Arbor Biosciences and Twist Bioscience — released commercial assays targeting the same core SNPs. Rohland et al. (2022, Genome Research) benchmarked all three on 27 common libraries and found Twist's panel produced the most uniform coverage, the lowest allelic bias, and the cleanest co-analysis with shotgun data; Harvard has since transitioned to Twist for new captures [22]. The 1240K dataset, however, remains the lingua franca of the field because it is what existing published genomes were genotyped against.

Authenticating that the DNA is actually ancient. Modern human DNA is the universal contaminant in any aDNA lab. Three signals together establish authenticity: (1) the characteristic deamination damage pattern — elevated C-to-T misincorporations at the 5′ end of reads (and G-to-A at the 3′ end for double-stranded libraries), quantified by tools such as mapDamage 2.0 [23] and DamageProfiler; (2) X-chromosomal heterozygosity in putative males, estimated by ANGSD's contamination module — Pilli et al. reported below 4% in the two Pompeii samples with sufficient X-chromosome coverage [5]; and (3) mitochondrial contamination estimates from contamMix or Schmutzi. Reads also carry expected fragment-length distributions (most molecules under 100 bp). Anything with full-length reads and no terminal damage is contamination, and is rejected.

Bioinformatics pipelines. The Florence group used EAGER (Efficient Ancient Genome Reconstruction; Peltzer et al. 2016) for read processing in the mitochondrial-capture phase: AdapterRemoval/Clip&Merge for adapter trimming and paired-end merging, BWA-aln (a short-read-tuned variant of the Burrows-Wheeler aligner) for mapping to the revised Cambridge Reference Sequence (rCRS) using CircularMapper to handle mtDNA's circular topology, and DeDup for paired-end-aware duplicate removal [24]. The current community-standard successor is nf-core/eager, a Nextflow pipeline released in 2021 with Docker/Singularity containers, automated benchmarking, and integration with most major HPC schedulers — designed so that any researcher in any institution can reproduce a Reich Lab- or Max Planck-quality analysis from raw FASTQ to genotype calls [25].

Downstream population-genetic analysis in the Pompeii paper used the Reich Lab's ADMIXTOOLS suite — qpAdm for fitting admixture models, qpWave for source-rank tests, and pre-computed f-statistics via qpfstats; EIGENSOFT/smartpca for principal component analysis; ADMIXTURE for unsupervised clustering at k = 2 through 15; KIN and BREADR for relatedness inference up to the third degree; hapROH for runs-of-homozygosity (which test for parental relatedness or small-population effects); sexDetERRmine for genetic sex determination from X/Y read ratios; and HIrisPlex-S for skin, eye, and hair pigmentation prediction [5]. Haplogrep3 handles mitochondrial haplogroup assignment against the PhyloTree reference; Y-chromosomal haplogroups were assigned against the YFull v8.09 phylogeny.

Reference databases. Two repositories anchor the field. Raw sequencing reads from virtually all published aDNA studies — including the Pompeii data, accession PRJEB74999 — go into the European Nucleotide Archive (ENA) or the U.S. Sequence Read Archive (SRA), where they are publicly accessible to anyone [29]. Curated, uniformly reprocessed genotype calls live in the Allen Ancient DNA Resource (AADR), maintained at Harvard Medical School by Mallick, Micco, Mah, Ringbauer, Lazaridis, Olalde, Patterson, and Reich [26]. The AADR's most recent public release at the time of this writing — version 9.0 / v62.0, dated 16 September 2024 — contains 13,571 ancient individuals plus 4,054 present-day reference samples drawn from the 1000 Genomes Project, the Simons Genome Diversity Project, and the Human Genome Diversity Project, all genotyped at the same core 1.23-million-SNP positions and aligned to the hg19 reference genome [27]. The dataset crossed 10,000 ancient individuals at the end of 2022 and is updated roughly twice a year; an ArcGIS-based AADR Visualizer went online in 2025 [28]. Without this infrastructure, comparing Cast 52 to Hellenistic Egyptians, Iron Age Italians, and Imperial Roman individuals from Latium would not be a one-day analysis — it would be a multi-year project.

What this means for the Pompeii result. The 14 cast samples Pilli et al. drilled were processed through this entire pipeline twice — once at Florence's Molecular Anthropology Unit and once at Harvard Medical School — using independent library preparations and independent capture protocols. The mitochondrial haplogroups recovered from each laboratory matched. The X-chromosomal contamination estimates were below 4%. The damage patterns were as expected for ~2,000-year-old DNA. The five usable genomes were then projected onto the AADR's reference space and modeled with qpAdm against deeply curated source populations. None of this proves the result correct in any philosophical sense, but it does mean that for a popular narrative — "the mother in the Golden Bracelet" — to survive contact with this evidence, every one of these independent checks would have to fail in the same direction. They didn't.

What the data do not say

It is worth being clear about the limits of the genetic evidence, because public summaries (including the YouTube transcript that prompted this article) tend to overshoot. The study analyzed five individuals — not the population of Pompeii. The ancestry signal is robust and aligns with parallel work on Imperial Rome, but the headline figure sometimes given as "70% eastern Mediterranean ancestry" is a description of these specific genomes, not an estimate of the city's population structure. Cast 50 yielded only mitochondrial data, so a paternal-line relationship to the other Golden Bracelet individuals cannot be formally excluded. Phenotypic predictions for skin, eye, and hair color were possible only because the HIrisPlex-S panel is small and well-calibrated; predictions of disease susceptibility from the same low-coverage data were not reliable. And the absence of biological kinship does not preclude social, household, or affective relationships among the dead. The genome is silent on whether the man holding the child loved him.

What the study does establish, with the rigor characteristic of David Reich's group and the Italian aDNA consortia, is that the popular narratives attached to these casts since the 19th century cannot be defended as factual claims. They were inferences from posture, ornament, and Victorian sentiment. They are now testable — and several have failed the test.

A note on dates and details. Some popular accounts of this story (including the YouTube transcript that prompted this article) state that the eruption began on August 25, 79 CE, that the casts were initially called "the Lady of Vesuvius," and that the first cast was made in 1863. The August date is the traditional one but is increasingly displaced by an October date based on the 2018 charcoal graffito and supporting evidence; Fiorelli's first cast is conventionally dated to 1863, but his systematic excavation began in 1860 and the casting technique was refined over several years. "Lady of Vesuvius" is not standard terminology in the peer-reviewed literature for Fiorelli's earliest casts. Temperature figures of "500 °C at Pompeii" should be read with care: the 500–600 °C estimates apply to Herculaneum and Oplontis, not to Pompeii, where surge temperatures were closer to 250–300 °C.

Sources

1. Pilli, E., Vai, S., Moses, V. C., et al. "Ancient DNA challenges prevailing interpretations of the Pompeii plaster casts." Current Biology 34(22): 5307–5318.e7 (November 18, 2024). DOI: 10.1016/j.cub.2024.10.007. Open-access manuscript at PubMed Central.
   https://www.cell.com/current-biology/fulltext/S0960-9822(24)01361-7 · https://pmc.ncbi.nlm.nih.gov/articles/PMC11627482/
2. Osanna, M., et al. Statement on charcoal inscription dated XVI K. Nov. (October 17, 79 CE) found in Regio V, Pompeii, October 16, 2018. Reported in Daley, J., "Ancient graffiti shifts date of Pompeii's destruction by 2 months," Science, October 17, 2018.
   https://www.science.org/content/article/ancient-graffiti-shifts-date-pompeii-s-destruction-back-2-months · https://phys.org/news/2018-10-pompeii-evidence-rewrites-vesuvius-eruption.html
3. Mastrolorenzo, G., Petrone, P., Pappalardo, L., and Guarino, F. M. "Lethal Thermal Impact at Periphery of Pyroclastic Surges: Evidences at Pompeii." PLOS ONE 5(6): e11127 (June 2010). DOI: 10.1371/journal.pone.0011127.
   https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0011127
4. Giordano, G., Pensa, A., Vona, A., et al. "Unique formation of organic glass from a human brain in the Vesuvius eruption of 79 CE." Scientific Reports 15: 4567 (February 27, 2025). DOI: 10.1038/s41598-025-88894-5. See also Petrone, P., et al., "Heat-Induced Brain Vitrification from the Vesuvius Eruption in c.e. 79," NEJM 382: 383–384 (2020), DOI: 10.1056/NEJMc1909867.
   https://www.nature.com/articles/s41598-025-88894-5 · https://www.nejm.org/doi/full/10.1056/NEJMc1909867
5. Pilli et al. 2024 (full STAR Methods, supplementary tables, and dataset descriptions). Same DOI as [1]. All quantitative claims about cast individuals, mtDNA/Y haplogroups, sex determination, qpAdm models, ROH, HIrisPlex-S phenotypes, and strontium/oxygen isotope values are drawn from this source.
6. Scorrano, G., Viva, S., Pinotti, T., et al. "Bioarchaeological and palaeogenomic portrait of two Pompeians that died during the eruption of Vesuvius in 79 AD." Scientific Reports 12: 6468 (May 26, 2022). DOI: 10.1038/s41598-022-10899-1. The first published genome from a Pompeian victim (the Casa del Fabbro individual).
   https://www.nature.com/articles/s41598-022-10899-1
7. Antonio, M. L., Gao, Z., Moots, H. M., et al. "Ancient Rome: A genetic crossroads of Europe and the Mediterranean." Science 366(6466): 708–714 (November 8, 2019). DOI: 10.1126/science.aay6826. The reference work establishing eastern-Mediterranean-shifted ancestry in Imperial Rome.
   https://www.science.org/doi/10.1126/science.aay6826
8. Associated Press / CBS News. "Pompeii excavation reveals large private bathhouse built 2,000 years ago." January 17, 2025.
   https://www.cbsnews.com/news/pompeii-excavation-thermal-complex-discovery/
9. Archaeology Magazine. "New Pompeii Excavations Reveal How One Family Tried to Save Themselves" (House of Helle and Phrixus). May 5, 2025.
   https://archaeology.org/news/2025/05/05/new-pompeii-excavations-reveal-how-one-family-tried-to-save-themselves/
10. Parco Archeologico di Pompei. "Pompei fu rioccupata dopo la distruzione del 79 d.C." Press release and article in the E-Journal degli Scavi di Pompei, August 6, 2025.
    https://pompeiisites.org/
11. CBS News / CNN. Coverage of the Insula Meridionalis reoccupation finding, August 6–14, 2025.
    https://www.cbsnews.com/news/pompeii-vesuvius-eruption-survivors-reoccupation-precarious-conditions/ · https://www.cnn.com/2025/08/14/science/pompeii-reoccupation-after-eruption
12. Krause, J. (commentary). "Archaeogenetics: Four letters from Pompeii." Current Biology 34(22): R1107 (November 18, 2024). DOI: 10.1016/j.cub.2024.10.009. Companion dispatch to Pilli et al.
    https://www.sciencedirect.com/science/article/pii/S0960982224013630
13. Foss, P. W. Pliny and the Eruption of Vesuvius. Routledge, 2022. Reassessment of the manuscript tradition for Pliny the Younger's letters and the eruption date. ISBN 978-1-032-00131-7.
14. Max Planck Institute for Evolutionary Anthropology / Cell Press. "DNA evidence rewrites histories for people buried in volcanic eruption in ancient Pompeii." Press release, EurekAlert!, November 7, 2024.
    https://www.eurekalert.org/news-releases/1063333
15. Strickland, A. (CNN). "An ancient stone staircase seemed to lead nowhere. It's revealing the 'lost Pompeii.'" December 7, 2025. Coverage of the Casa del Tiaso tower reconstruction.
    https://www.cnn.com/2025/12/07/science/pompeii-tower-digital-archaeology
16. GreekReporter. "Pompeii Excavation Reveals Lavish New Finds in Villa Linked to Nero's Wife." December 20, 2025. Villa Poppaea / Oplontis findings.
    https://greekreporter.com/2025/12/20/pompeii-excavation-villa-nero-wife-new-finds/
17. CBS News. "Archaeologists at Pompeii use AI to reveal the face of a victim trying to flee the Mount Vesuvius eruption." April 2026. Pompeii Archaeological Park / University of Padua reconstruction.
    https://www.cbsnews.com/news/pompeii-ai-face-victim-mount-vesuvius-eruption/
18. Bentley, D. R., Balasubramanian, S., Swerdlow, H. P., et al. "Accurate Whole Human Genome Sequencing using Reversible Terminator Chemistry." Nature 456: 53–59 (2008). DOI: 10.1038/nature07517. The foundational paper for Illumina sequencing-by-synthesis. See also Illumina's technology overview.
    https://www.illumina.com/science/technology/next-generation-sequencing/sequencing-technology.html
19. Meyer, M., and Kircher, M. "Illumina sequencing library preparation for highly multiplexed target capture and sequencing." Cold Spring Harbor Protocols 2010(6): pdb.prot5448 (June 2010). DOI: 10.1101/pdb.prot5448. The double-stranded library preparation protocol used in the Pompeii study and adopted as a community standard for ancient DNA.
    http://cshprotocols.cshlp.org/content/2010/6/pdb.prot5448
20. Rohland, N., Harney, E., Mallick, S., Nordenfelt, S., and Reich, D. "Partial uracil–DNA–glycosylase treatment for screening of ancient DNA." Philosophical Transactions of the Royal Society B 370(1660): 20130624 (January 19, 2015). DOI: 10.1098/rstb.2013.0624. Establishes the partial-UDG protocol used to retain terminal damage signal while removing internal uracils.
    https://royalsocietypublishing.org/doi/10.1098/rstb.2013.0624
21. Gansauge, M.-T., and Meyer, M. "Single-stranded DNA library preparation for the sequencing of ancient or damaged DNA." Nature Protocols 8: 737–748 (2013). DOI: 10.1038/nprot.2013.038. Updated in Gansauge, M.-T., et al., "Single-stranded DNA library preparation from highly degraded DNA using T4 DNA ligase," Nucleic Acids Research 45(10): e79 (2017). DOI: 10.1093/nar/gkx033.
    https://www.nature.com/articles/nprot.2013.038 · https://pmc.ncbi.nlm.nih.gov/articles/PMC5449542/
22. Rohland, N., Mallick, S., Mah, M., Maier, R., Patterson, N., and Reich, D. "Three assays for in-solution enrichment of ancient human DNA at more than a million SNPs." Genome Research 32: 2068–2078 (2022). DOI: 10.1101/gr.276728.122. Benchmarks the Reich Lab 1240K reagent against commercial Daicel Arbor Biosciences and Twist Bioscience panels. See also the foundational 1240K probe design in Mathieson, I., et al., Nature 528: 499–503 (2015), DOI: 10.1038/nature16152, and Haak, W., et al., Nature 522: 207–211 (2015), DOI: 10.1038/nature14317.
    https://pmc.ncbi.nlm.nih.gov/articles/PMC9808625/
23. Jónsson, H., Ginolhac, A., Schubert, M., Johnson, P. L. F., and Orlando, L. "mapDamage2.0: fast approximate Bayesian estimates of ancient DNA damage parameters." Bioinformatics 29(13): 1682–1684 (2013). DOI: 10.1093/bioinformatics/btt193. The standard tool for quantifying terminal C-to-T deamination patterns that authenticate ancient DNA.
    https://academic.oup.com/bioinformatics/article/29/13/1682/184965
24. Peltzer, A., Jäger, G., Herbig, A., Seitz, A., Kniep, C., Krause, J., and Nieselt, K. "EAGER: efficient ancient genome reconstruction." Genome Biology 17: 60 (2016). DOI: 10.1186/s13059-016-0918-z. The pipeline used by the Florence laboratory in the Pompeii study.
    https://genomebiology.biomedcentral.com/articles/10.1186/s13059-016-0918-z
25. Fellows Yates, J. A., Lamnidis, T. C., Borry, M., et al. "Reproducible, portable, and efficient ancient genome reconstruction with nf-core/eager." PeerJ 9: e10947 (March 2021). DOI: 10.7717/peerj.10947. The Nextflow successor to EAGER, the current community-standard pipeline. Project repository at github.com/nf-core/eager.
    https://peerj.com/articles/10947/
26. Mallick, S., Micco, A., Mah, M., Ringbauer, H., Lazaridis, I., Olalde, I., Patterson, N., and Reich, D. "The Allen Ancient DNA Resource (AADR): a curated compendium of ancient human genomes." Scientific Data 11: 182 (February 10, 2024). DOI: 10.1038/s41597-024-03031-7.
    https://www.nature.com/articles/s41597-024-03031-7
27. Allen Ancient DNA Resource (AADR), version 9.0 / v62.0. Released September 16, 2024. Harvard Dataverse, DOI: 10.7910/DVN/FFIDCW. Contains 13,571 ancient and 4,054 present-day individuals. Release page: https://reich.hms.harvard.edu/allen-ancient-dna-resource-aadr-downloadable-genotypes-present-day-and-ancient-dna-data · Versioned releases: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/FFIDCW
28. Cheronet, O., et al. "AADR Visualizer: an ArcGIS online visualizer for ancient human DNA from the Allen Ancient DNA Resource." Bioinformatics Advances 5(1): vbaf199 (August 2025). DOI: 10.1093/bioadv/vbaf199.
    https://academic.oup.com/bioinformaticsadvances/article/5/1/vbaf199/8238351
29. European Nucleotide Archive (ENA). Sequencing data for Pilli et al. 2024, accession PRJEB74999. Reich Lab genotype data: https://reich.hms.harvard.edu/datasets.
    https://www.ebi.ac.uk/ena/browser/view/PRJEB74999