Next Article in Journal
Antiviral and Immunomodulatory Effects of 7-Deaza-2-methyladenosine (7DMA) in a Susceptible Mouse Model of Usutu Virus Infection
Previous Article in Journal
Quercetin Regulates Autophagy to Inhibit PRRSV Replication Through the PI3K/Akt/mTOR Signaling Pathway
Previous Article in Special Issue
Probenecid Treatment Inhibits Replication of the Edmonston Measles Virus Strain in Vero Cells
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Viruses
Volume 17
Issue 12
10.3390/v17121638
viruses-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

AI-Powered Identification of Human Cell Surface Protein Interactors of the Hemagglutinin Glycoprotein of High-Pandemic-Risk H5N1 Influenza Virus

by
Christian Poitras
and
Benoit Coulombe
*
Montreal Clinical Research Institute, 110 Pine Ave West, Montréal, QC H2W 1R7, Canada
*
Author to whom correspondence should be addressed.
Viruses 2025, 17(12), 1638; https://doi.org/10.3390/v17121638
Submission received: 9 November 2025 / Revised: 11 December 2025 / Accepted: 15 December 2025 / Published: 17 December 2025
(This article belongs to the Special Issue Virus–Host Protein Interactions)
Browse Figure
Versions Notes

Abstract

H5N1 is a highly pathogenic avian influenza virus of major global concern. Since 2023, it has circulated widely among wild and farmed birds, with increasing spillover into mammals, including minks, seals, and cattle, and sporadic infections in humans in Chile, the UK, and the USA. The risk of a future pandemic is considered high because ongoing viral evolution could enable efficient human-to-human transmission. The hemagglutinin (HA) glycoprotein is the principal determinant of host range, mediating viral attachment and entry through interactions with sialylated glycans and potentially additional host surface proteins. Here, we developed an artificial intelligence (AI)-based pipeline integrating structural modeling, protein–protein interaction prediction, and biological filtering to identify human cell surface proteins with high likelihood of interacting with H5N1 HA. These interactions may contribute to viral entry and tropism and therefore represent promising candidates for experimental validation and therapeutic targeting. Our findings highlight the utility of AI-driven pipelines in accelerating the discovery of host factors relevant to pandemic influenza viruses.

1. Introduction

The highly pathogenic avian influenza virus (HPAIV) H5N1 is one of the most concerning zoonotic pathogens due to its potential to cause a human pandemic. Since its emergence in 1997, H5N1 has repeatedly spilled over from birds into humans, often causing severe disease, though sustained human-to-human transmission has not yet been established [1,2]. The current wave of H5N1 circulation (2023–2025) has been marked by intense spread among wild and farmed birds, with occasional human cases reported in Chile, the UK, and the USA [3,4,5]. Alarmingly, infections have also been documented in several mammalian species, including minks, seals, and cattle, raising concerns about ongoing adaptation toward human hosts [6,7,8]. The World Health Organization has therefore classified the pandemic risk associated with H5N1 as “high” [9].
A key molecular determinant of influenza host specificity and tropism is the viral hemagglutinin (HA) surface glycoprotein. HA mediates viral entry by binding to sialylated glycans on host cells and subsequently triggering membrane fusion [10]. Mutations in HA that alter receptor-binding specificity or stability have historically underpinned host shifts in avian influenza viruses into humans [11,12]. For example, recent structural analyses demonstrated that specific substitutions in H5N1 HA can shift binding preference from avian-like (α2,3-linked) to human-like (α2,6-linked) sialic acids, potentially facilitating airborne transmissibility [13,14]. Other studies revealed that internal auto-bound glycans within HA must be displaced to permit receptor engagement, adding another layer of regulation to receptor binding [15].
While HA–glycan interactions are essential, they may not fully explain host cell entry. Evidence suggests that influenza viruses can exploit additional host cell surface proteins as attachment factors or co-receptors. For instance, HA has been shown to bind human immune cell subsets in ways not fully dependent on sialylated glycans [16]. Furthermore, multivalent filamentous HA–glycan arrangements can enhance avidity and alter binding geometry, suggesting complex binding modes at the cell surface [17]. However, systematic identification of human cell-surface proteins that directly interact with H5N1 HA remains limited. Although the A/goose/Guangdong/1/1996 strain used here is strictly avian and not expected to infect human airway epithelial cells due to its preference for α2,3-linked sialic acids, the goal of this study is not to model productive infection. Instead, we evaluate the direct structural compatibility between the HA ectodomain and human extracellular proteins. This approach is independent of viral entry competence and allows identification of potential host attachment factors that could be relevant across both avian and human-adapted H5N1 strains. The same workflow can be applied directly to circulating human H5N1 isolates, including clade 2.3.4.4b viruses.
Virus–host protein–protein interactions (PPIs) are crucial for viral replication, immune evasion, and pathogenesis [18]. Although several large-scale studies have mapped PPIs for influenza polymerase and other internal proteins [19], fewer efforts have targeted viral surface glycoproteins such as HA. Identifying host surface proteins that physically interact with HA could yield new insights into viral tropism and identify promising antiviral targets.
Recent advances in artificial intelligence (AI) and structural biology now enable predictive mapping of virus–host interactions at unprecedented resolution. AlphaFold-based pipelines and machine learning models have been successfully applied to predict viral PPIs and host tropism signatures [20,21,22,23]. For influenza, computational approaches including XGBoost, convolutional neural networks, and hybrid feature models have shown promise in predicting human-virus PPIs and classifying host range based on HA sequences [24,25,26]. Yet, most of these approaches have focused on intracellular interactions or broad PPI networks, rather than surface glycoprotein–host protein complexes.
In this study, we developed an AI-driven pipeline integrating structural modeling, docking, and biological filtering to identify human cell-surface proteins with high likelihood of interacting with H5N1 HA. We prioritized candidates based on structural compatibility, cell-surface localization, and relevance to respiratory epithelium. By applying this framework to contemporary H5N1 isolates, we propose a set of novel HA–host interactions that may influence infection dynamics and represent promising targets for antiviral discovery.

2. Materials and Methods

Sequence Data Acquisition—The protein sequence of the Influenza A virus (A/goose/Guangdong/1/1996(H5N1)) hemagglutinin (HA) gene was downloaded from NCBI database on 24 January 2024.
All human proteins sequences for cell surface GO annotation number 0009986 (a total of 958 proteins) were downloaded from UniProt database on 14 July 2023 (UniProt Consortium, 2023). Proteins not annotated under this GO term (e.g., intracellular or nuclear proteins such as DCTPP1 or RPSA) were intentionally excluded, as the objective was to identify direct biochemical compatibilities between extracellular prefusion HA and surface-exposed human proteins, not the intracellular interactome.
The AlphaFold 3 databases were downloaded on 1 June 2025.
Bait-Target Pair Generation—Protein sequences from hemagglutinin gene sequences (bait) and every individual human protein from the cell surface GO annotation (targets) were combined into bait-target pairs. For each bait-target pair, an AlphaFold3 JSON file was created.
AlphaFold 3 modeling—AlphaFold 3 [27] version 3.0.1 was run in “data pipeline only” mode for every bait-target JSON file using the parameters “--max_template_date=2025-06-01 --norun_inference”.
AlphaFold 3 was run in “model inference only” for every bait-target JSON file created by AlphaFold 3 data pipeline using the parameters “--norun_data_pipeline”.
The highest ipTM score for each bait-target pair was extracted from each summary confidences JSON file created by AlphaFold 3 model inference and was saved into an Excel file.
AFM-LIS [28] was used to compute Local Interface Score (LIS) and Local Interaction Area (LIA) for each bait-target using the output of AlphaFold 3 as an input for AFM-LIS. The result was saved into an Excel file.
Data Integration and Thresholding—Excel files containing LIS and LIA from AFM-LIS and ipTM scores from AlphaFold 3 were merged. A threshold of 0.203 for LIS and 3432 for LIA were used to color the cells in green to show the best potential interactors. The thresholds are calculated according to Kim et al., 2024 (see Supplementary Figure S8 in Ref. [28]). A positive PPI is suggested if either of the following conditions is met: Best LIS ≥ 0.203 AND Best LIA ≥ 3432, or Average LIS ≥ 0.073 AND Average LIA ≥ 1610.
Image generation—Images were generated using ChimeraX version 1.10, GIMP version 3.0.4 and Inkscape version 1.3.
Functional classification of host interactors—Candidate HA-interacting proteins were manually grouped into six biological categories (immune/inflammatory, adhesion/surface receptors, cytokines/growth factors, neural/synaptic proteins, signaling enzymes, and plasma/structural proteins). Classification was based on UniProt and Gene Ontology functional annotations, domain architecture, tissue expression, and known relevance to viral tropism, immune modulation, and epithelial/endothelial biology.

3. Results

3.1. Identification of the Top 5 High-Confidence HA–Host Protein Interactions

Using AlphaFold 3 (AF3), we screened 987 human extracellular and membrane-associated proteins for potential interactions with the prefusion H5N1 hemagglutinin (HA) protein. AF3 returned 30 high-confidence candidate complexes, among which the top five showed ipTM larger than 0.75 and well-defined interface geometries (Table 1). These top candidates include BMP2, PROCR, ICAM1, LY6G6D, and IGHV2-5. All five are accessible to extracellular HA and play roles in immune regulation, vascular biology, or antigen recognition, processes central to H5N1 pathogenesis. Figure 1 shows structural models of high-confidence human proteins interacting with HA.

3.2. The H5N1 HA-Interacting Proteins Form Six Functional Groups

The H5N1 HA-interacting proteins can be regrouped in six functional groups shown in Table 1, namely Immune/Inflammatory, Surface/Adhesion, Growth Factors, Neural/Synaptic, Signaling Enzymes, and Plasma/Structural. These groups reflect major processes disrupted during severe influenza and H5N1 disease Details for each group are provided in Table 1. Of note, the six categories represent biological interpretation and are not derived from a clustering algorithm.

3.3. Structural and Functional Annotation of the Remaining 26 Predicted Interactors

The remaining 25 predicted HA-binding proteins were annotated for function, disease links, expression, structural features, and, importantly, subcellular localization (Table 2). Most proteins are located at the plasma membrane or secreted into the extracellular space, consistent with accessibility to intact HA during viral attachment and early entry.

3.4. Subset of Predicted Interactors Overlaps with Known Influenza Host Factors

Although few human proteins have been experimentally shown to bind HA directly (e.g., NKp44, NKp46), several AF3-predicted partners have strong independent evidence linking them to influenza A infection, host restriction, or influenza-induced lung injury. These proteins -BTN3A3, ICAM1, CEACAM19 (via CEACAM1), CSF3R, LBP, and HRG- form a coherent subset of high biological plausibility (Table 3). Their presence within the predicted interactome reinforces that AF3 captures relevant immunological and pathophysiological themes, not random structural matches.

3.5. Localization Strengthens Biological Plausibility

We systematically added subcellular localization for all host interactors to refine biological plausibility. The majority of predicted interactors are plasma membrane or secreted proteins, and therefore physically accessible to intact prefusion HA during the initial attachment phase. This is consistent with the goal of modeling extracellular HA binding rather than intracellular viral processes. Predicted intracellular interactors are interpreted cautiously, as possible downstream signaling interfaces, post-entry partners, or structurally compatible false positives.
This information is now integrated throughout Table 1, Table 2 and Table 3.
These classes are not random; instead, they map onto known influenza mechanisms including epithelial damage, vascular leakage, neutrophil-driven pathology, cytokine amplification, and coagulopathy.

4. Discussion

The SARS-CoV-2 (COVID 19) pandemic revealed how unprepared our medical systems were to face large scale viral outbreaks, in this particular case leading to dramatic issues at various levels of our societies. Strong of this lesson, we now understand that it is essential to be prepared for the next such pandemics, in case it occurs. The H5N1 virus has been identified as a putative threat to humanity because it already replicates in animal and human cells. We are possibly at only a few mutations away of a highly dangerous infection spreading across the world.
One valuable method to prepare for H5N1 spread is to identify and characterize surface proteins in human cells that can be targeted by the virus, namely its surface antigens such as HA. In this study, we have used the software package AlphaFold 3, a performant software for determining the structures of proteins as well as protein–protein interactions. All human surface proteins were virtually paired with H5N1 HA and interactions found positive were selected (see Section 2). In our conditions, a total of 30 human proteins were found positive in this assay. Five of them have interaction scores that overpass the 25 others. We were able to manually separate interaction pairs into six categories (see Table 1). A detailed analysis of the selected proteins emphasizes the functional variety of HA interactors. Overall, these interactors are specific proteins in distinct cell types, and they have many specific structural domains (also see Figure 1). Moreover, their role in H5N1 infection was in many cases predictable. This last point is important as pharmacologists and clinicians can find in this work a wealth of information useful to develop drugs and treatments for facing a possible H5N1 outbreak in humans. Similar studies involving PPI maps have revealed drug candidates for SARS-CoV-2 [58].
AF3 reveals a structured, biologically coherent HA–host interactome. Although AF3 predictions do not establish physical binding, several aspects of our results indicate that the predicted interactome is biologically meaningful rather than random. High ipTM scores for the top-ranked complexes suggest well-formed HA–host interfaces. The proteins fall into clear functional groups consistent with processes known to be critical in H5N1 pathogenesis: immune sensing, epithelial barrier control, vascular integrity, neutrophil biology, and coagulation. This alignment argues that AF3 is capturing real structural compatibilities relevant to influenza infection.
Avian HA and human-cell infectivity: scope of the structural predictions. Because the HA used in this study derives from an avian (goose) H5N1 isolate, it is not expected to efficiently enter human epithelial cells due to its specificity for α2,3-linked sialic acids. However, the aim of our computational screening was not to model productive infection, but to evaluate direct structural compatibility between extracellular prefusion HA and human surface proteins. Protein–protein docking scores are not dependent on viral entry efficiency, and the prefusion HA fold is highly conserved across avian and emerging human-lineage H5N1 viruses. Our predictions therefore reflect potential biochemical compatibilities that may become relevant if human-adapting mutations arise. We also note that the same workflow can be applied without modification to human H5N1 isolates as sequence data become available.
Independent host-factor evidence reinforces several predictions. Six AF3-predicted proteins already have independent experimental links to influenza A infection (Table 3). BTN3A3 restricts avian IAV replication in primate airways; ICAM1 and CEACAM1 are induced in influenza-infected cells; CSF3R signaling exacerbates influenza-associated lung injury; LBP mediates lethal TLR4-dependent inflammation; HRG regulates vascular and coagulation pathways disrupted in severe flu. The presence of this non-random overlap strengthens the credibility of AF3-generated predictions.
Complementarity between AI predictions and cellular proteomics. A recent proteomic study [59] identified DCTPP1 and RPSA as HA-associated host proteins. These were not present in our predicted set because they were not included in the initial protein pool, which was restricted to the Gene Ontology category “cell surface proteins.” DCTPP1 and RPSA are intracellular proteins and therefore fall outside the biological scope of our extracellular HA binding analysis. Their detection in proteomics highlights interacting partners that may be relevant after viral internalization or as part of intracellular multi-protein complexes. Thus, proteomic and structural AI-based predictions capture distinct but complementary layers of the HA interactome.
HA processing and antigen presentation: scope and limitations. HA is not intact throughout infection. HA0 is cleaved into HA1 and HA2, undergoes pH-triggered rearrangements during endocytosis, and is ultimately digested into peptides for MHC I/II presentation. Our AF3 analysis intentionally focuses on extracellular prefusion HA, which is the relevant conformation for virus attachment and entry. We explicitly do not model post-fusion HA, HA degradation, MHC loading, or T-cell recognition. This limitation is acknowledged.
In addition, several predicted interactors, such as immunoglobulin variable segments, CEACAM family members, and plasma proteins including HRG, are expected to bind intact HA ectodomain rather than peptide fragments derived from antigen processing. Our analysis therefore focuses on the extracellular, prefusion form of HA relevant for receptor engagement, immune attachment factors, or soluble scavenging proteins. Epitope-level screening of HA peptides represents a valuable future extension of our computational framework.
Localization considerations clarify which interactions are most plausible. We added a Localization column to Table 2 and Table 3. Most high-confidence hits are extracellular or plasma-membrane proteins, i.e., physically accessible to HA. These interactions are the most plausible. Intracellular predictions are interpreted as lower-confidence hypotheses or possible downstream effectors. Localization thus adds an important layer of biological filtering.
Constraints on experimental validation (BSL-3 requirements). Validating and confirming 30 HA–host interactions involving highly pathogenic H5N1 is not feasible without BSL-3 containment and a dedicated virology program. Our study serves as a computational roadmap, identifying priority targets for future biochemical and virological validation rather than delivering validated interactions. Given the above constraints, our AF3 structural predictions provide: (i) High-confidence candidate complexes that can guide targeted experimental programs, (ii) Functional clustering consistent with known influenza biology, (iii) A subset of host factors already validated in independent influenza studies, and (iv) A structurally guided hypothesis framework for viral entry, immune evasion, and pathogenesis. We present these findings as a preliminary but biologically grounded interactome map, offering new testable hypotheses for future influenza research.

5. Conclusions

The computational workflow presented here (integrating sequence retrieval, structural prediction, and systematic screening of human surface proteins) offers a rapid approach to identify putative extracellular interactors of viral glycoproteins. While our results do not establish physical binding and must be validated experimentally, they provide a structured and biologically coherent set of hypotheses for future influenza research. Importantly, the workflow is, in principle, applicable to other viral surface proteins, although such extensions remain to be demonstrated experimentally. Our findings thus serve as a foundation for targeted virological studies aimed at clarifying mechanisms of viral attachment, tropism, and pathogenesis in emerging influenza strains.

Author Contributions

Conceptualization, B.C.; Bioinformatic analysis, C.P.; Figure preparation, C.P. and B.C.; Writing—Original Draft Preparation, C.P. and B.C.; Writing—Review and Editing, C.P. and B.C.; Supervision, B.C.; All authors have read and agreed to the published version of the manuscript.

Funding

This work is fully supported by the Montreal Clinical Research Institute.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.

Acknowledgments

Authors wish to thank members of their laboratory for helpful discussions. We also acknowledge the use of OpenAI’s ChatGPT (GPT-5) for editorial assistance, language refinement, and formatting support during the preparation of this manuscript. This research was possible with the support of Calcul Québec (calculquebec.ca) and Digital Research Alliance of Canada (alliancecan.ca).

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AF3AlphaFold 3
AIArtificial Intelligence
GOGene Ontology
HAHemagglutinin Protein
IgImmunoglobulin
ILInterleukin
LIALocal Interface Area
LISLocal Interaction Score
MHCMajor Histocompatibility Complex
PPIProtein–Protein Interaction

References

  1. Claas, E.C.J.; Osterhaus, A.D.M.E.; van Beek, R.; De Jong, J.C.; Rimmelzwaan, G.F.; Senne, D.A.; Krauss, S.; Shortridge, K.F.; Webster, R.G. Human influenza A H5N1 virus related to a highly pathogenic avian influenza virus. Lancet 1998, 351, 472–477. [Google Scholar] [CrossRef] [PubMed]
  2. Subbarao, K.; Klimov, A.; Katz, J.; Regnery, H.; Lim, W.; Hall, H.; Perdue, M.; Swayne, D.; Bender, C.; Huang, J.; et al. Characterization of an avian influenza A (H5N1) virus isolated from a child with a fatal respiratory illness. Science 1998, 279, 393–396. [Google Scholar] [CrossRef] [PubMed]
  3. WHO. Human Infection caused by Avian Influenza A(H5N1)—Chile. Disease Outbreak News. 21 April 2023. Available online: https://www.who.int/emergencies/disease-outbreak-news/item/2023-DON461 (accessed on 9 November 2025).
  4. UK Health Security Agency. Avian Influenza (H5N1) Confirmed in UK; UKHSA Technical Briefing; UK Health Security Agency: London, UK, 2023. [Google Scholar]
  5. Garg, S.; Reed, C.; Davis, C.T.; Uyeki, T.M.; Behravesh, C.B.; Kniss, K.; Budd, A.; Biggerstaff, M.; Adjemian, J.; Barnes, J.R.; et al. Outbreak of Highly Pathogenic Avian Influenza A(H5N1) Viruses in U.S. Dairy Cattle and Detection of Two Human Cases-United States, 2024. MMWR Morb. Mortal. Wkly. Rep. 2024, 73, 501–505. [Google Scholar] [CrossRef]
  6. Agüero, M.; Monne, I.; Sánchez, A.; Zecchin, B.; Fusaro, A.; Ruano, M.J.; del Valle Arrojo, M.; Fernández-Antonio, R.; Suoto, A.M.; Tordable, P.; et al. Highly pathogenic avian influenza A(H5N1) virus infection in farmed minks, Spain, October 2022. Euro Surveill. 2023, 28, 2300009. [Google Scholar] [CrossRef]
  7. Puryear, W.; Sawatzki, K.; Hill, N.; Foss, A.; Stone, J.; Doughty, L.; Walk, D.; Gilbert, K.; Murray, M.; Cox, E.; et al. Outbreak of highly pathogenic avian influenza H5N1 in New England seals. Nature 2023, 619, 440–445. [Google Scholar]
  8. National Veterinary Services Laboratories (NVSL). Detection of Highly Pathogenic Avian Influenza A(H5N1) in Dairy Cattle—United States; National Veterinary Services Laboratories: Ames, IA, USA, 2024. [Google Scholar]
  9. WHO. Avian Influenza A(H5N1): Situation Update; WHO: Geneva, Switzerland, 2024. [Google Scholar]
  10. Skehel, J.J.; Wiley, D.C. Receptor binding and membrane fusion in virus entry: The influenza hemagglutinin. Annu. Rev. Biochem. 2000, 69, 531–569. [Google Scholar] [CrossRef]
  11. Russell, C.J.; Hu, M.; Okda, F.A. Influenza hemagglutinin protein stability, activation, and pandemic risk. Trends Microbiol. 2018, 26, 841–853. [Google Scholar] [CrossRef] [PubMed]
  12. Herfst, S.; Schrauwen, E.J.A.; Linster, M.; Chutinimitkul, S.; de Wit, E.; Munster, V.J.; Sorrell, E.M.; Bestebroer, T.M.; Burke, D.F.; Smith, D.J.; et al. Airborne transmission of influenza A/H5N1 virus between ferrets. Science 2012, 336, 1534–1541. [Google Scholar] [CrossRef]
  13. Yang, H.; Carney, P.J.; Chang, J.C.; Villanueva, J.M.; Stevens, J. Structure and receptor binding preferences of recombinant hemagglutinins from avian and human H5N1 influenza viruses. J. Virol. 2015, 89, 4612–4623. [Google Scholar] [CrossRef]
  14. Wang, J.; Yang, P.; Zhang, Y.; Li, H.; Zhu, W.; Liu, L.; Wang, Q.; Chen, Y.; Gao, R.; Shu, Y.; et al. A single HA mutation switches receptor specificity of bovine H5N1 toward human-like sialylated glycans. Science 2024, 384, eadt0180. [Google Scholar]
  15. Benton, D.J.; Gamblin, S.J.; Skehel, J.J. Structural insights into influenza virus HA–glycan interactions. Structure 2025, 33, 1–13. [Google Scholar]
  16. Fu, Y.; Zhang, Y.; Gu, H.; Qin, C.; Hu, Y.; Chen, Y.; Wang, J.; Yang, W.; Liu, J.; Gao, G.F.; et al. Binding of influenza A (H5N1) virus hemagglutinin to human γδ T cells. PLoS ONE 2014, 9, e98176. [Google Scholar]
  17. Liu, S.; Liu, X.; Wu, N.C.; Das, S.R.; McBride, R.; He, J.; Xu, R.; Wilson, I.A.; Paulson, J.C.; Ward, A.B.; et al. Multivalent glycan-mediated binding of influenza HA filaments enhances avidity. Nat. Commun. 2024, 15, 1509. [Google Scholar]
  18. Watanabe, T.; Watanabe, S.; Kawaoka, Y. Influenza virus–host interactions: Mechanisms of viral pathogenesis. Nat. Rev. Microbiol. 2010, 8, 267–278. [Google Scholar]
  19. Tripathi, S.; Batra, J.; Lal, S.K. Interplay between influenza A virus and host factors: Targets for antiviral intervention. Front. Microbiol. 2017, 8, 739. [Google Scholar] [CrossRef] [PubMed]
  20. Jumper, J.; Evans, R.; Pritzel, A.; Green, T.; Figurnov, M.; Ronneberger, O.; Tunyasuvunakool, K.; Bates, R.; Žídek, A.; Potapenko, A.; et al. Highly accurate protein structure prediction with AlphaFold. Nature 2021, 596, 583–589. [Google Scholar] [CrossRef]
  21. Bryant, P.; Pozzati, G.; Elofsson, A. Improved prediction of protein–protein interactions using AlphaFold2. Nat. Commun. 2022, 13, 1265. [Google Scholar] [CrossRef]
  22. Gao, W.; Zhao, J.; Wang, L.; Li, X.; Chen, H.; Zhang, Y.; Liu, Y.; Sun, Z.; Zhou, X.; Li, Y.; et al. AI-based prediction of virus–host protein interactions. Brief. Bioinform. 2024, 25, bbae034. [Google Scholar]
  23. Rao, R.; Liu, J.; Verkuil, R.; Meier, J.; Canny, J.; Abbeel, P.; Sercu, T.; Rives, A.; Marks, D.S.; Hopkins, M.; et al. Computational prediction of protein complexes with AlphaFold-Multimer. Science 2022, 375, 780–785. [Google Scholar]
  24. Li, B.; Li, X.; Li, X.; Wang, L.; Lu, J.; Wang, J. Predicting influenza A virus–human protein–protein interactions using machine learning. PeerJ 2024, 12, e18863. [Google Scholar]
  25. Wang, H.; Lin, Y.; Luo, Z.; Sun, Y.; Chen, J.; Zhang, Y.; Li, X.; Zhao, Y.; Liu, H.; Zhou, Y.; et al. Host tropism prediction for influenza viruses based on HA and NA sequences using deep learning. Front. Immunol. 2022, 13, 834617. [Google Scholar]
  26. Yang, C.; Chen, J.; Luo, F.; Li, H.; Zhang, Y.; Wang, J.; Liu, X.; Zhou, Y.; Sun, Z.; Li, Y.; et al. Multi-channel neural networks for influenza virus classification. Bioinformatics 2021, 37, 4012–4020. [Google Scholar]
  27. Abramson, J.; Adler, J.; Dunger, J.; Evans, R.; Green, T.; Pritzel, A.; Ronneberger, O.; Willmore, L.; Ballard, A.J.; Bambrick, J.; et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 2024, 630, 493–500. [Google Scholar] [CrossRef]
  28. Kim, A.R.; Hu, Y.; Comjean, A.; Rodiger, J.; Mohr, S.E.; Perrimon, N. Enhanced protein–protein interaction discovery via AlphaFold-Multimer. bioRxiv 2024. preprint. [Google Scholar] [CrossRef]
  29. Lefranc, M.-P.; Giudicelli, V.; Ginestoux, C.; Jabado-Michaloud, J.; Folch, G.; Bellahcene, F.; Wu, Y.; Gemrot, E.; Brochet, X.; Lane, J.; et al. IMGT®, the international ImMunoGeneTics information system®. Nucleic Acids Res. 2009, 37, D1006–D1012. [Google Scholar] [CrossRef]
  30. Jelkmann, W. Regulation of erythropoietin production. Physiol. Rev. 2012, 92, 449–489. [Google Scholar] [CrossRef]
  31. de Beaucoudrey, L.; Puel, A.; Filipe-Santos, O.; Cobat, A.; Ghandil, P.; Chrabieh, M.; Feinberg, J.; von Bernuth, H.; Samarina, A.; Jannière, L.; et al. Mutations in STAT3 and IL12RB1 impair the development of human IL-17–producing T cells. Semin. Immunol. 2007, 19, 331–340. [Google Scholar] [CrossRef]
  32. Waldmann, T.A. The biology of interleukin-15: Implications for cancer therapy and immune regulation. Nat. Rev. Immunol. 2006, 6, 595–601. [Google Scholar] [CrossRef]
  33. Wang, Z.; Shen, D.; Parsons, D.W.; Bardelli, A.; Sager, J.; Szabo, S.; Ptak, J.; Silliman, N.; Peters, B.A.; van der Heijden, M.S.; et al. Mutational analysis of the tyrosine phosphatome in colorectal cancers. Science 2004, 304, 1164–1166. [Google Scholar] [CrossRef] [PubMed]
  34. Rhodes, D.A.; Reith, W.; Trowsdale, J. Regulation of immunity by butyrophilins. J. Immunol. 2005, 174, 3045–3052. [Google Scholar] [CrossRef] [PubMed]
  35. de Wit, J.; Ghosh, A. Specification of synaptic connectivity by cell surface interactions. Nat. Rev. Neurosci. 2016, 17, 22–35. [Google Scholar] [CrossRef] [PubMed]
  36. Poon, I.K.H.; Patel, K.K.; Davis, D.S.; Parish, C.R.; Hulett, M.D. Histidine-rich glycoprotein: The Swiss Army knife of mammalian plasma. J. Thromb. Haemost. 2011, 9, 203–213. [Google Scholar] [CrossRef] [PubMed]
  37. Kong, Y.-Y.; Yoshida, H.; Sarosi, I.; Tan, H.-L.; Timms, E.; Capparelli, C.; Morony, S.; Oliveira-dos-Santos, A.J.; Van, G.; Itie, A.; et al. OPGL is a key regulator of osteoclastogenesis, lymphocyte development and lymph-node organogenesis. Nature 1999, 397, 315–323. [Google Scholar] [CrossRef]
  38. Magnuson, B.; Ekim, B.; Fingar, D.C. Regulation and function of ribosomal protein S6 kinase (S6K) within mTOR signalling networks. Cell Metab. 2012, 15, 283–297. [Google Scholar] [CrossRef]
  39. Mullighan, C.G.; Collins-Underwood, J.R.; Phillips, L.A.; Loudin, M.G.; Liu, W.; Zhang, J.; Ma, J.; Coustan-Smith, E.; Harvey, R.C.; Willman, C.L.; et al. Rearrangement of CRLF2 in B-progenitor– and Down syndrome–associated acute lymphoblastic leukemia. Nat. Genet. 2009, 41, 1243–1246. [Google Scholar] [CrossRef]
  40. Kuespert, K.; Pils, S.; Hauck, C.R. CEACAMs: Their role in physiology and pathophysiology. Nat. Rev. Microbiol. 2006, 4, 611–621. [Google Scholar] [CrossRef]
  41. Südhof, T.C. Neuroligins and neurexins link synaptic function to cognitive disease. Nature 2008, 455, 903–911. [Google Scholar] [CrossRef]
  42. Kurima, K.; Peters, L.M.; Yang, Y.; Riazuddin, S.; Ahmed, Z.M.; Naz, S.; Arnaud, D.; Drury, S.; Mo, J.; Makishima, T.; et al. Dominant and recessive deafness caused by mutations of TMC1. Nat. Genet. 2002, 30, 277–284. [Google Scholar] [CrossRef]
  43. Imai, Y.; Soda, M.; Takahashi, R. Parkin suppresses unfolded protein stress-induced cell death through its E3 ubiquitin-protein ligase activity. J. Neurochem. 2000, 75, 2513–2520. [Google Scholar] [CrossRef]
  44. Wright, G.J.; Cherwinski, H.; Foster-Cuevas, M.; Brooke, G.; Puklavec, M.J.; Bigler, M.; Song, Y.; Jenmalm, M.; Gorman, D.; McClanahan, T.; et al. Characterization of the CD200 receptor family in mice and humans. Immunology 2009, 128, 1–13. [Google Scholar]
  45. Lanier, L.L.; Chang, C.; Phillips, J.H. Human NKR-P1: A C-type lectin superfamily member expressed by a subset of NK and T lymphocytes. J. Immunol. 1993, 150, 458–468. [Google Scholar]
  46. Wright, S.D.; Ramos, R.A.; Tobias, P.S.; Ulevitch, R.J.; Mathison, J.C. CD14, a receptor for complexes of lipopolysaccharide (LPS) and LPS-binding protein. Science 1990, 249, 1431–1433. [Google Scholar] [CrossRef] [PubMed]
  47. Maxson, J.E.; Gotlib, J.; Pollyea, D.A.; Fleischman, A.G.; Agarwal, A.; Eide, C.A.; Bottomly, D.; Wilmot, B.; McWeeney, S.K.; Tognon, C.E.; et al. Oncogenic CSF3R mutations in chronic neutrophilic leukemia and atypical CML. N. Engl. J. Med. 2013, 368, 1781–1790. [Google Scholar] [CrossRef] [PubMed]
  48. Chuang, P.T.; McMahon, A.P. Vertebrate Hedgehog signalling modulated by induction of a Hedgehog-binding protein. Nature 1999, 397, 617–621. [Google Scholar] [CrossRef]
  49. Tatsuo, H.; Ono, N.; Tanaka, K.; Yanagi, Y. SLAM (CDw150) is a cellular receptor for measles virus. Nature 2000, 406, 893–897. [Google Scholar] [CrossRef]
  50. Weisel, J.W. Fibrinogen and fibrin. Adv. Protein Chem. 2005, 70, 247–299. [Google Scholar] [CrossRef]
  51. Park, S.G.; Ewalt, K.L.; Kim, S. Functional expansion of aminoacyl-tRNA synthetases and their interacting factors: New perspectives on housekeepers. Trends Biochem. Sci. 2005, 30, 569–574. [Google Scholar] [CrossRef] [PubMed]
  52. Pinto, R.M.; Bakshi, S.; Lytras, S.; Zakaria, M.K.; Swingler, S.; Worrell, J.C.; Herder, V.; Hargrave, K.E.; Varjak, M.; Cameron-Ruiz, N.; et al. BTN3A3 evasion promotes the zoonotic potential of influenza A viruses. Nature 2023, 619, 338–347. [Google Scholar] [CrossRef]
  53. Othumpangat, S.; Timour, M.S.; Verma, V.; Piedimonte, G. Influenza virus infection increases expression of adhesion molecules in bronchial epithelial cells. Virology 2016, 496, 1–10. [Google Scholar]
  54. Ye, Z.-W.; Yuan, S.; Chan, J.F.-W.; Zhang, A.J.; Yu, C.-Y.; Ong, C.P.; Chan, K.-H.; Yuen, K.-Y.; Kok, K.-H.; Jin, D.-Y.; et al. CEACAM1 regulates cytokine responses in primary human alveolar epithelial cells infected with H5N1 influenza virus. Sci. Rep. 2018, 8, 16924. [Google Scholar] [CrossRef]
  55. McGrath, J.C.; Fajardo, I.; Guven, N.; Shan, M.; Wong, C.; Daigneault, A.; Chen, Y.; Kotas, M.E.; McNagny, K.M.; Hogg, J.C.; et al. Neutrophil-driven alveolar barrier disruption in influenza requires CSF3–CSF3R signaling. Eur. Respir. J. 2022, 59, 2102977. [Google Scholar]
  56. Shirey, K.A.; Lai, W.; Scott, A.J.; Lipsky, M.; Mistry, P.; Pletneva, L.M.; Karp, C.L.; McAlees, J.; Gioannini, T.L.; Weiss, J.; et al. The TLR4 antagonist Eritoran protects mice from lethal influenza infection. Nature 2013, 497, 498–502. [Google Scholar] [CrossRef] [PubMed]
  57. Poon, I.K.H.; Patel, K.K.; Davis, D.S.; Parish, C.R.; Hulett, M.D. Histidine-rich glycoprotein: The Swiss Army knife of mammalian plasma. Blood 2011, 117, 2093–2101. [Google Scholar] [CrossRef] [PubMed]
  58. Gordon, D.E.; Jang, G.M.; Bouhaddou, M.; Xu, J.; Obernier, K.; White, K.M.; O’Meara, M.J.; Rezelj, V.V.; Guo, J.Z.; Swaney, D.L.; et al. A SARS-CoV-2 protein interaction map reveals targets for drug repurposing. Nature 2020, 583, 459–468. [Google Scholar] [CrossRef]
  59. Haas, K.M.; McGregor, M.J.; Bouhaddou, M.; Polacco, B.J.; Kim, E.-Y.; Nugyan, T.T.; Newton, B.W.; Urbanowski, M.; Kim, H.; Williams, M.A.P.; et al. Proteomic and genetic analyses of influenza A viruses identify pan-viral host targets. Nat. Commun. 2023, 14, 6030. [Google Scholar] [CrossRef]
Viruses 17 01638 g001
Figure 1. Molecular models of the high-confidence human proteins interacting with HA. The color code indicates the confidence for each domain.
Figure 1. Molecular models of the high-confidence human proteins interacting with HA. The color code indicates the confidence for each domain.
Viruses 17 01638 g001
Table 1. Functional grouping of predicted HA–host interactors.
Table 1. Functional grouping of predicted HA–host interactors.
GroupRepresentative ProteinsFunctional AxisH5N1 Connection
Immune/InflammatoryICAM1, IL12RB1, IL15, SLAMF1, KLRB1Cytokine and immune signalingCytokine storm, immune modulation
Surface/AdhesionPROCR, CEACAM19, LY6G6DViral entry and cell bindingPotential alternative HA interactions
Growth FactorsBMP2, AIMP1, EPOTissue repair, host growth regulationHost remodeling post infection
Neural/SynapticLRFN2, GPR37Neurotropic or cross-reactive pathwaysPotential neurological sequelae
Signaling EnzymesRPS6KB1, PTPRTIntracellular signaling regulationVirus-induced signaling reprogramming
Plasma/StructuralFGB, HRG, CSF3RCoagulation, barrier integrityVascular leakage, sepsis-like features
Table 2. Remaining 26 significant HA interactors.
Table 2. Remaining 26 significant HA interactors.
SymbolFunction/RoleDisease LinksExpressionLocalizationStructural NotesKey Reference
IGHV1-3Immunoglobulin heavy variable segment shaping antigen specificity.Autoimmunity, infection, B-cell malignancy.B cells/plasma cellsPlasma membrane/secretedIg V-domain; CDR loops define binding.[29]
IGHV3-30-3Ig heavy variable segment (antigen binding).As above.B-cell lineagePlasma membrane/secretedIg V-domain[29]
EPOCytokine driving erythropoiesis via EPOR.Anemia, polycythemia, CKD; tumors.Kidney interstitial cells; secreted to plasmaSecreted4-helix bundle cytokine[30]
IGHV3-30Ig heavy variable segment.As above.B-cell lineagePlasma membrane/secretedIg V-domain[29]
IGHV3-30-5Ig heavy variable segment.As above.B-cell lineagePlasma membrane/secretedIg V-domain[29]
IL12RB1IL-12/23 receptor β1 subunit; Th1/NK signaling.MSMD immunodeficiency.T cells, NK cells, DCsPlasma membrane (type I receptor)FN3 domains[31]
IL15NK and CD8 T-cell homeostasis; transpresentation by IL15RA.Autoimmunity, inflammation, cancer.APCs, stromal and epithelial cellsSecreted4-helix bundle cytokine[32]
PTPRTReceptor-type tyrosine phosphatase; tumor suppressor.CRC, brain tumors.Neuronal and epithelialPlasma membrane (Ig/FN3 ectodomain)Dual PTP domains[33]
BTN3A3Butyrophilin family immune modulator.Viral and tumor immunity.APCs, epitheliaPlasma membrane (Ig-like ectodomain)B30.2/PRY-SPRY domain[34]
LRFN2Synaptic adhesion; excitatory synapse assembly.Neurodevelopmental disorders.NeuronsPlasma membrane (LRR/FN3)TM, short tail[35]
HRGHistidine-rich glycoprotein; coagulation/immunity.Sepsis, thrombosis, cancer.Liver plasma proteinSecretedCystatin-like domains; Zn2+[36]
TNFRSF11A (RANK)RANK receptor; bone and immune signaling.Osteopetrosis, Paget’s disease, metastasis.Osteoclast precursors, DCs, epitheliaPlasma membrane (TNFR)Cysteine-rich domains[37]
RPS6KB1p70S6K; mTORC1 effector.Cancer (oncogenic activation).Broad, high in proliferating cellsCytosolicSer/Thr kinase (Thr389)[38]
CRLF2TSLP receptor subunit; B-cell development.B-ALL; allergy.Hematopoietic progenitors, thymic stromaPlasma membraneType I cytokine receptor[39]
CEACAM19CEACAM family adhesion molecule.Epithelial cancers.Epithelium (predicted)Plasma membraneIg-like ectodomains[40]
LRFN5Synaptic adhesion; neuronal connectivity.Neurodevelopmental traits.NeuronsPlasma membrane (LRR)TM[41]
TMC1Mechanotransduction channel component.DFNB7/11 deafness.Cochlear hair cellsMulti-pass membraneChannel-like protein[42]
GPR37GPCR in neuronal/myelin biology; parkin substrate.Parkinson’s, leukodystrophy.Oligodendrocytes, neuronsPlasma membrane (GPCR)Class A, 7TM[43]
CD200R1LCD200 receptor-like immune modulator.Immune regulation.Myeloid lineage (pred.)Plasma membraneIg-like ectodomains; TM[44]
KLRB1C-type lectin-like NK/T-cell receptor.Autoimmunity, infections, cancer.NK, NKT, CD4/CD8 T-cellsPlasma membrane (type II)CTLR homodimer[45]
LBPLPS-binding protein for TLR4 signaling.Sepsis, infection susceptibility.Liver, plasma; GI epitheliaSecretedLipid-binding protein[46]
CSF3RG-CSF receptor; granulopoiesis.CNL/AML; neutropenia.Myeloid progenitors, neutrophilsPlasma membrane (type I receptor)Cytokine receptor[47]
HHIPHedgehog-interacting protein.COPD, cancer, development.Lung, cartilageSecretedHh-binding protein[48]
SLAMF1SLAM-family lymphocyte activator.Measles entry receptor; immune regulation.Activated T/B cells, DCs, macrophagesPlasma membrane (IgV/IgC)ITSM motifs[49]
FGBFibrinogen β-chain.Dysfibrinogenemia, thrombosis, CVD.Liver → plasmaSecretedFibrinogen domains[50]
AIMP1tRNA synthetase scaffold; secreted stress cytokine-like factor.Inflammation, angiogenesis, cancer.Ubiquitous; stress secretionSecreted/cytosolicAdaptor; extracellular processed[51]
Table 3. Host factors with independent influenza-related evidence.
Table 3. Host factors with independent influenza-related evidence.
ProteinLocalizationInfluenza-Related EvidenceKey Reference
BTN3A3Plasma membranePrimate-specific antiviral restriction factor inhibiting avian IAV replication; expressed in human airway cells.[52]
ICAM1Plasma membraneUpregulated in influenza-infected airway epithelial cells; modulates viral survival and cytokine responses.[53]
CEACAM19 (via CEACAM1)Plasma membraneCEACAM1 highly induced in H5N1-infected alveolar type II cells; enhances inflammatory signaling.[54]
CSF3RPlasma membraneCSF3–CSF3R axis drives neutrophilic lung injury and barrier disruption in influenza.[55]
LBPSecreted/plasmaTLR4–LBP pathway mediates lethal influenza-induced acute lung injury; Eritoran protection in vivo.[56]
HRGSecreted/plasmaRegulates coagulation, vascular integrity, and immune responses—pathways disrupted in severe influenza.[57]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Poitras, C.; Coulombe, B. AI-Powered Identification of Human Cell Surface Protein Interactors of the Hemagglutinin Glycoprotein of High-Pandemic-Risk H5N1 Influenza Virus. Viruses 2025, 17, 1638. https://doi.org/10.3390/v17121638

AMA Style

Poitras C, Coulombe B. AI-Powered Identification of Human Cell Surface Protein Interactors of the Hemagglutinin Glycoprotein of High-Pandemic-Risk H5N1 Influenza Virus. Viruses. 2025; 17(12):1638. https://doi.org/10.3390/v17121638

Chicago/Turabian Style

Poitras, Christian, and Benoit Coulombe. 2025. "AI-Powered Identification of Human Cell Surface Protein Interactors of the Hemagglutinin Glycoprotein of High-Pandemic-Risk H5N1 Influenza Virus" Viruses 17, no. 12: 1638. https://doi.org/10.3390/v17121638

APA Style

Poitras, C., & Coulombe, B. (2025). AI-Powered Identification of Human Cell Surface Protein Interactors of the Hemagglutinin Glycoprotein of High-Pandemic-Risk H5N1 Influenza Virus. Viruses, 17(12), 1638. https://doi.org/10.3390/v17121638

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop