Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.05.23.726879

Interpreting the WaveSeekerNet model to reveal the evolution and biology of influenza A virus

Nguyen, H.-H.·Rudar, J.·Mubareka, S.·Lapen, D.·Berhane, Y.·Leung, C. K.·Lung, O.
View original source ↗Export metadata ↓

Abstract

BackgroundInfluenza A virus (IAV) is a major public health burden, causing seasonal epidemics and occasional pandemics. Its transmission from avian species to mammals and subsequent spread requires adaptive changes in the viral genome. Understanding these molecular adaptations is essential for pandemic preparedness, and machine learning offers a powerful approach to uncover the evolution and biology of IAV. ResultsThis study established a well-calibrated WaveSeekerNet model that accurately predicted the host source across all 8 IAV segments (macro F1-score: 0.9728), significantly improving the reliability of predicted probabilities with calibration errors approaching zero. Model interpretation revealed that avian-adapted IAVs consistently activated G/C content, whereas mammalian-adapted IAVs generally activated A/T content. This distinction was confirmed by codon-level analysis, in which G/C-rich codons were rewarded for the avian hosts and A/T-rich codons for the mammalian hosts. In the feature space learned by WaveSeekerNet, we defined host-adaptive distance to quantify species barriers and proposed it as a risk-assessment metric. We hypothesized the Mammalian Adaptation Zone (MAZ), a zone where the virus is expected to adjust its host-adaptive distance to reach, thereby helping it establish persistent mammalian lineages. The analysis also revealed the Hard Distance of avian-origin viruses (e.g., H5Nx, H9N2), indicating they have not yet established persistent mammalian lineages. Finally, analysis of human H7N9 (2013, China) and non-human mammalian H5Nx (North America) viruses showed that WaveSeekerNet accurately identified key mammalian-adaptive mutations, including PB2-E627K and PB2-D701N. ConclusionsWaveSeekerNet elucidated IAV host-adaptation mechanisms in silico, providing insights into the underlying mechanisms of host adaptation and informing improved surveillance and intervention strategies.

Explore related subjects

genomicsPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Nguyen, H.-H., Rudar, J., Mubareka, S., Lapen, D., Berhane, Y., Leung, C. K., Lung, O.. 2026-05-25. Interpreting the WaveSeekerNet model to reveal the evolution and biology of influenza A virus. https://doi.org/10.64898/2026.05.23.726879

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Structural variation in repeat elements is widespread in normal human tissues and in tumorigenesis

Somatic mosaicism contributes to genomic variation, yet postzygotic structural variants remain under-characterized. We performed long- and short-read WGS from multiple individuals (n=47 normal tissues; n=168 samples) and identified mosaic structural variants in all individuals and germ layers, impacting a median 285.2 kb/genome. Nearly half of breakpoints were independently validated, with tissue distributions reflecting both early and late developmental origins. Most mosaic variants were repeat-mediated and 8.3% overlapped functional elements, an enrichment compared to germline variants. To extend these analyses in samples where long-read sequencing is infeasible, we measured repeat alterations from short-read sequencing, recapitulating mosaic tissue-specific differences. We characterized tumor- and tissue- specific variation in repeats across 15 cancer types and found tumor-related repeat variation to be similar in scale to that of normal mosaic variation. Tracking repeat changes in cell-free DNA provided a noninvasive approach for tumor monitoring. Our analyses revealed widespread repeat-driven structural variation in health and disease.

genomics↗
bioRxiv2026

RNA isoform-resolved multiplexed sequencing with bioorthogonal barcoding

RNA isoform dysregulation drives disease pathogenesis and is the target of FDA-approved splice-switching therapeutics. However, multiplexed sequencing methods discard splice junction information because only 3' termini are barcoded and counted. Here, we repurpose acylation and click chemistries to conjugate bioorthogonal barcodes (bobcodes) directly onto multiple internal positions along cellular RNAs. Bobcoded RNAs from multiple samples are pooled for multiplexed cDNA synthesis, during which reverse transcriptase switches from each RNA template onto its tethered bobcode with greater than 99% accuracy in species mixing experiments. Bobcode attachment intervals set cDNA insert sizes without a library fragmentation step, and priming with poly(dT) or random hexamers selects between 3'-end counting and full-length isoform capture. A bioorthogonal barcode-sequencing (BOB-seq v0.1) drug screen identifies transcriptome-wide on- and off-target RNA splicing effects and outperforms existing multiplexing RNA sequencing methods in workflow simplicity, sample-to-sample variability, and barcoding accuracy. Bobcodes add isoform resolution to scalable multiplexed RNA sequencing.

genomics↗
bioRxiv2026

Structural polymorphism and population-variable coding capacity of HERV-K(HML-2) in human pangenomes

Approximately 8% of the human genome is derived from ancient retroviral infections. The most recently integrated of these endogenous retroviruses is the HERV-K(HML-2) clade, whose expression has been associated with cancer, amyotrophic lateral sclerosis, and embryogenesis. Studies of HERV expression, particularly HML-2, have relied predominantly on short-read sequencing. However, the high similarity among HML-2 proviruses prevents many short reads from being assigned uniquely to individual loci. We therefore compared haplotype-resolved long-read genome assemblies from 292 donors to resolve variation in proviral structure and coding capacity. Several loci previously thought to be fixed were structurally polymorphic. Tandem arrays occurred at 13 loci and contained up to six proviral copies in a single array. At 8q11.23, we identified a previously undescribed full-length provirus in one haplotype. All 583 other haplotypes carried a solo-LTR. We found that standard reference genomes failed to represent the coding capacity retained in many individuals, whose proviruses contained intact open reading frames despite disruptive mutations in the reference sequences. Short-read genotypes left 32.5% of the tested donor-variant pairs unresolved at sites associated with viral reading frames. These findings show why HML-2 expression must be interpreted in the context of the structural and coding alleles each individual carries.

genomics↗