Genetic introgression from archaic hominins has had a major impact on modern human genetic diversity and adaptive potential. However, producing a complete catalog of introgressed sequences has proven difficult, especially for segments that lie within structurally complex regions of the genome. Conventional, reference-based detection approaches tend to miss sequences in those regions and may fragment intact archaic segments. The authors address this gap by developing an assembly-centered detection framework designed to recover more complete archaic-derived sequences.
The study introduces ASMaid (ASseMbly-based archaic introgression detector), a computational framework based on a Hidden Markov Model (HMM). Rather than relying solely on a single linear reference genome, ASMaid leverages haplotype-resolved pangenome assemblies. Using assemblies permits direct comparison of contiguous haplotype sequences, which improves resolution in regions with structural complexity or divergence from the reference. The HMM provides a probabilistic model to assign genomic segments as archaic-derived versus non-archaic based on sequence evidence present in the assembled haplotypes.
A distinguishing feature of ASMaid is that it jointly integrates two classes of genetic evidence: single-nucleotide genotype information and structural variation (SV) signals. By combining SNV-level genotype signatures with SV presence and structure, the method aims to identify intact archaic segments that would be fragmented or overlooked when considering only single-nucleotide patterns or reference alignments. The approach is therefore tailored to capture both small-scale and larger, structurally mediated traces of archaic ancestry.
ASMaid was applied to a global panel comprising 610 phased human genome assemblies. Using this assembly-based pipeline, the authors report that non-African individuals carry approximately 79.8 Mbp of Neanderthal-derived sequence and 8.3 Mbp of Denisovan-derived sequence. The abstract specifies that these values represent substantial increases over previous estimates derived from conventional reference-based methods. No further breakdown by population, length distribution, or confidence intervals is provided in the available excerpt.
One notable outcome reported is the detection of several centromere-spanning archaic segments, genomic regions that are often refractory to detection with reference-based approaches due to repetitive sequence and structural complexity. The authors highlight East Asian (EAS)-specific calls on chromosomes 5 and 7 as examples of such recovered segments, indicating that the assembly-based strategy can access previously inaccessible parts of the genome and reveal population-specific introgression patterns.
In addition to sequence blocks, ASMaid uncovered 1,701 archaic-derived structural variants across the analyzed assemblies. This finding emphasizes a previously underappreciated layer of archaic contribution in the form of SVs, which can have distinct functional consequences compared with single-nucleotide variants. The source text does not provide a catalog of these SVs, their genomic locations, frequencies, or predicted functional effects in the excerpt provided.
The abstract states that high-frequency introgressed loci are enriched in pathways, suggesting functional or adaptive significance of some archaic-derived sequences. However, the provided source excerpt truncates mid-sentence and does not report which pathways, the nature of the enrichment analysis, statistical measures, or specific loci involved. Therefore, pathway identities, enrichment magnitudes, and downstream interpretations were not reported in the available text.
Using haplotype-resolved pangenome assemblies combined with a probabilistic HMM and dual SNV+SV evidence enables recovery of a larger and more intact set of archaic-derived sequences compared with conventional reference-based methods. Strengths explicitly reported include the ability to recover centromere-spanning segments and to identify a substantial number of archaic-derived structural variants. Limitations in the provided excerpt include the absence of detailed results beyond the headline Mbp estimates and SV count, and the abstract’s truncation before completing the description of pathway enrichments. Additionally, because this work appears as a preprint, the abstract notes it has not yet been certified by peer review.
ASMaid demonstrates that an assembly-focused, HMM-driven detection framework that integrates structural variation and single-nucleotide signals can substantially expand the detectable set of archaic introgressed sequences in modern human genomes. The reported totals—approximately 79.8 Mbp of Neanderthal and 8.3 Mbp of Denisovan sequence in non-Africans—and the identification of 1,701 archaic-derived SVs indicate a larger archaic genetic legacy than previously estimated. The source excerpt does not provide further methodological details, per-population breakdowns, or the specific pathway enrichments referenced, so readers should consult the full preprint or subsequent peer-reviewed publication for complete data, analyses, and interpretations.