Using a mutual information-based site transition network to map the genetic evolution of influenza A/H3N2 virus

TEMPO: A transformer-based mutation prediction framework for SARS-CoV-2 evolution

The widespread of SARS-CoV-2 presents a significant threat to human society, as well as public health and economic development. Extensive efforts have been undertaken to battle against the pandemic, whereas effective approaches such as vaccination would be weakened by the continuous mutations, leading to considerable attention being attracted to the mutation prediction. However, most previous studies lack attention to phylogenetics. In this paper, we propose a novel and effective model TEMPO for predicting the mutation of SARS-CoV-2 evolution. Specifically, we design a phylogenetic tree-based sampling method to generate sequence evolution data. Then, a transformer-based model is presented for the site mutation prediction after learning the high-level representation of these sequence data. We conduct experiments to verify the effectiveness of TEMPO, leveraging a large-scale SARS-CoV- 2 dataset. Experimental results show that TEMPO is effective for mutation prediction of SARS- CoV-2 evolution and outperforms several state-of-the-art baseline methods. We further perform mutation prediction experiments of other infectious viruses, to explore the feasibility and robustness of TEMPO, and experimental results verify its superiority. The codes and datasets are freely available at https://github.com/ZJUDataIntelligence/TEMPO.

Anticipating time-dependent antigenic variants of influenza A (H3N2) viruses

Frequent variations in influenza vaccines are necessary to match antigenic variants which appear in influenza epidemics. Antigenic variants of influenza viruses result from frequent mutations in amino acid residues located on their hemagglutinin (HA) proteins. Knowledge of specific changes in these amino acids helps to characterize distinct antigenic variants. In this paper, statistical models are developed and used to investigate changes in amino acids which accompany antigenic variants of epidemiological importance. Amino acid sequences of the HA proteins of influenza A (H3N2) strains isolated from 1968 to 2015 were obtained. The sequences were aligned using Clustal Omega and the number of differences in amino acid residues located on annotated positions of antigenic sites of the HA protein between pairs of strains were determined. These were linked in the statistical models and used to assess the relationship between any pair of influenza strains. The results revealed that both antigenic similarity between strains and the amino acid changes are affected by the time of isolation of the strains. Furthermore, the models predicted that rates of changes in amino acids located on the antigenic sites ranged between 5% and 6% per site per year. The findings of the study suggest that time-dependent antigenic variants of influenza A (H3N2) strains may occur as they evolve. The study has the potential to greatly improve influenza surveillance in as much as it supports vaccine designs.

In Silico Vaccine Strain Prediction for Human Influenza Viruses

Vaccines preventing seasonal influenza infections save many lives every year; however, due to rapid viral evolution, they have to be updated frequently to remain effective. To identify appropriate vaccine strains, the World Health Organization (WHO) operates a global program that continually generates and interprets surveillance data. Over the past decade, sophisticated computational techniques, drawing from multiple theoretical disciplines, have been developed that predict viral lineages rising to predominance, assess their suitability as vaccine strains, link genetic to antigenic alterations, as well as integrate and visualize genetic, epidemiological, structural, and antigenic data. These could form the basis of an objective and reproducible vaccine strain-selection procedure utilizing the complex, large-scale data types from surveillance. To this end, computational techniques should already be incorporated into the vaccine-selection process in an independent, parallel track, and their performance continuously evaluated.

Immune history and influenza virus susceptibility

Antibody responses to influenza viruses are critical for protection, but the ways in which repeated viral exposures shape antibody evolution and effectiveness over time remain controversial. Early observations demonstrated that viral exposure history has a profound effect on the specificity and magnitude of antibody responses to a new viral strain, a phenomenon called 'original antigenic sin.' Although 'sin' might suppress some aspects of the immune response, so far there is little indication that hosts with pre-existing immunity are more susceptible to viral infections compared to naïve hosts. However, the tendency of the immune response to focus on previously recognized conserved epitopes when encountering new viral strains can create an opportunity cost when mutations arise in these conserved epitopes. Hosts with different exposure histories may continue to experience distinct patterns of infection over time, which may influence influenza viruses' continued antigenic evolution. Understanding the dynamics of B cell competition that underlie the development of antibody responses might help explain the low effectiveness of current influenza vaccines and lead to better vaccination strategies.

Rules of co-occurring mutations characterize the antigenic evolution of human influenza A/H3N2, A/H1N1 and B viruses

BACKGROUND

The human influenza viruses undergo rapid evolution (especially in hemagglutinin (HA), a glycoprotein on the surface of the virus), which enables the virus population to constantly evade the human immune system. Therefore, the vaccine has to be updated every year to stay effective. There is a need to characterize the evolution of influenza viruses for better selection of vaccine candidates and the prediction of pandemic strains. Studies have shown that the influenza hemagglutinin evolution is driven by the simultaneous mutations at antigenic sites. Here, we analyze simultaneous or co-occurring mutations in the HA protein of human influenza A/H3N2, A/H1N1 and B viruses to predict potential mutations, characterizing the antigenic evolution.

METHODS

We obtain the rules of mutation co-occurrence using association rule mining after extracting HA1 sequences and detect co-mutation sites under strong selective pressure. Then we predict the potential drifts with specific mutations of the viruses based on the rules and compare the results with the "observed" mutations in different years.

RESULTS

The sites under frequent mutations are in antigenic regions (epitopes) or receptor binding sites.

CONCLUSIONS

Our study demonstrates the co-occurring site mutations obtained by rule mining can capture the evolution of influenza viruses, and confirms that cooperative interactions among sites of HA1 protein drive the influenza antigenic evolution.

Influenza infection in human host: challenges in making a better influenza vaccine

Influenza is a ubiquitous infection with a spectrum ranging from mild to severe. The mystery regarding such variability in the clinical spectrum has not been fully unravelled, although a role for the complex interplay among virus characteristics, host immune response and environmental factors has been suggested. Antivirals and current vaccines have a limited role in prophylaxis and treatment because they primarily target surface glycoproteins which undergo antigenic/genetic changes under host immune pressure. Targeting conserved internal proteins could lead the way to a universal vaccine which can be used against various types/subtypes. However, this is on the distant horizon, so in the meantime, developing improved vaccines should be given high priority. In this review, we discuss where the current influenza research stands in terms of vaccines, adjuvants, and how we can better predict the vaccine strains for upcoming influenza seasons by understanding complex phenomena which drive the continuous antigenic evolution.

Expected Shannon Entropy and Shannon Differentiation between Subpopulations for Neutral Genes under the Finite Island Model

Shannon entropy H and related measures are increasingly used in molecular ecology and population genetics because (1) unlike measures based on heterozygosity or allele number, these measures weigh alleles in proportion to their population fraction, thus capturing a previously-ignored aspect of allele frequency distributions that may be important in many applications; (2) these measures connect directly to the rich predictive mathematics of information theory; (3) Shannon entropy is completely additive and has an explicitly hierarchical nature; and (4) Shannon entropy-based differentiation measures obey strong monotonicity properties that heterozygosity-based measures lack. We derive simple new expressions for the expected values of the Shannon entropy of the equilibrium allele distribution at a neutral locus in a single isolated population under two models of mutation: the infinite allele model and the stepwise mutation model. Surprisingly, this complex stochastic system for each model has an entropy expressable as a simple combination of well-known mathematical functions. Moreover, entropy- and heterozygosity-based measures for each model are linked by simple relationships that are shown by simulations to be approximately valid even far from equilibrium. We also identify a bridge between the two models of mutation. We apply our approach to subdivided populations which follow the finite island model, obtaining the Shannon entropy of the equilibrium allele distributions of the subpopulations and of the total population. We also derive the expected mutual information and normalized mutual information ("Shannon differentiation") between subpopulations at equilibrium, and identify the model parameters that determine them. We apply our measures to data from the common starling (Sturnus vulgaris) in Australia. Our measures provide a test for neutrality that is robust to violations of equilibrium assumptions, as verified on real world data from starlings.

Selecting vaccine strains for H3N2 human influenza A virus

H3N2 human influenza A virus causes epidemics of influenza mainly in the winter season in temperate regions. Since the antigenicity of this virus evolves rapidly, several attempts have been made to predict the major amino acid sequence of hemagglutinin 1 (HA1) in the target season of vaccination. However, the usefulness of predicted sequence was unclear because its relationship to the antigenicity was unknown. Here the antigenic model for estimating the degree of antigenic difference (antigenic distance) between amino acid sequences of HA1 was integrated into the process of selecting vaccine strains for H3N2 human influenza A virus. When the effectiveness of a potential vaccine strain for a target season was evaluated retrospectively using the average antigenic distance between the strain and the epidemic viruses sampled in the target season, the most effective vaccine strain was identified mostly in the season one year before the target season (pre-target season). Effectiveness of actual vaccines appeared to be lower than that of the strains randomly chosen in the pre-target season on average. It was recommended to replace the vaccine strain for every target season with the strain having the smallest average antigenic distance to the others in the pre-target season. The procedure of selecting vaccine strains for future epidemic seasons described in the present study was implemented in the influenza virus forecasting system (INFLUCAST) (http://www.nsc.nagoya-cu.ac.jp/~yossuzuk/influcast.html).

Predictability of antigenic evolution for H3N2 human influenza A virus

Influenza A virus continues to pose a threat to public health. Since this virus can evolve escape mutants rapidly, it is desirable to predict the antigenic evolution for developing effective vaccines. Although empirical methods have been proposed and reported to predict the antigenic evolution more or less accurately, they did not provide much insight into the effects of unobserved mutations and the mechanisms of antigenic evolution. Here a theoretical method was introduced to predict the antigenic evolution of H3N2 human influenza A virus by evaluating de novo mutations through estimating the antigenic distance. The antigenic distance defined with the hemagglutination inhibition (HI) titer was estimated with antigenic models taking into account the volume, isoelectric point, relative solvent accessibility, and distances from receptor-binding sites (RBS) and N-linked glycosylation sites (NGS) for amino acids in hemagglutinin 1 (HA1). When the best model with the optimized parameter values was used to predict the antigenic evolution for the dominant strains, the prediction accuracy was relatively low. However, there appeared to be an overall tendency that the amino acid sites with larger potential net effect on antigenicity were more likely to evolve and the amino acid changes with larger potential effect were more likely to take place, suggesting that natural selection may operate to enhance the antigenic evolution of H3N2 human influenza A virus.

Computational prediction of vaccine strains for human influenza A (H3N2) viruses

Human influenza A viruses are rapidly evolving pathogens that cause substantial morbidity and mortality in seasonal epidemics around the globe. To ensure continued protection, the strains used for the production of the seasonal influenza vaccine have to be regularly updated, which involves data collection and analysis by numerous experts worldwide. Computer-guided analysis is becoming increasingly important in this problem due to the vast amounts of generated data. We here describe a computational method for selecting a suitable strain for production of the human influenza A virus vaccine. It interprets available antigenic and genomic sequence data based on measures of antigenic novelty and rate of propagation of the viral strains throughout the population. For viral isolates sampled between 2002 and 2007, we used this method to predict the antigenic evolution of the H3N2 viruses in retrospective testing scenarios. When seasons were scored as true or false predictions, our method returned six true positives, three false negatives, eight true negatives, and one false positive, or 78% accuracy overall. In comparison to the recommendations by the WHO, we identified the correct antigenic variant once at the same time and twice one season ahead. Even though it cannot be ruled out that practical reasons such as lack of a sufficiently well-growing candidate strain may in some cases have prevented recommendation of the best-matching strain by the WHO, our computational decision procedure allows quantitative interpretation of the growing amounts of data and may help to match the vaccine better to predominating strains in seasonal influenza epidemics. Importance: Human influenza A viruses continuously change antigenically to circumvent the immune protection evoked by vaccination or previously circulating viral strains. To maintain vaccine protection and thereby reduce the mortality and morbidity caused by infections, regular updates of the vaccine strains are required. We have developed a data-driven framework for vaccine strain prediction which facilitates the computational analysis of genetic and antigenic data and does not rely on explicit evolutionary models. Our computational decision procedure generated good matches of the vaccine strain to the circulating predominant strain for most seasons and could be used to support the expert-guided prediction made by the WHO; it thus may allow an increase in vaccine efficacy.

Predicting the emergence of H3N2 influenza viruses reveals contrasted modes of evolution of HA and NA antigens

Vaccine design for rapidly changing viruses is based on empirical surveillance of strains circulating in a given season to assess those that will most likely spread during the next season. The choice of which strains to include in the vaccine is critical, as an erroneous decision can lead to a nonimmunized human population that will then be at risk in the face of an epidemic or, worse, a pandemic. Here, we present the first steps toward a very general phylogenetic approach to predict the emergence of novel viruses. Our genomic model builds upon natural features of viral evolution such as selection and recombination / reassortment, and incorporates episodic bursts of evolution and or of recombination. As a proof-of-concept, we assess the performance of this model in a retrospective study, focusing: (i) on the emergence of an unexpected H3N2 influenza strain in 2007, and (ii) on a longitudinal design. Based on the analysis of hemagglutinin (HA) and neuraminidase (NA) genes, our results show a lack of predictive power in both experimental designs, but shed light on the mode of evolution of these two antigens: (i) supporting the lack of significance of recombination in the evolution of this influenza virus, and (ii) showing that HA evolves episodically while NA changes gradually.

A novel empirical mutual information approach to identify co-evolving amino acid positions of influenza A viruses

Mutual information (MI) is an approach commonly used to estimate the evolutionary correlation of 2 amino acid sites. Although several MI methods exist, prior to our contribution no systematic method had been developed to assess their performance, or to establish numerical thresholds to detect co-evolving amino acid sites. The current study performed a Markov chain Monte Carlo (MCMC) algorithm on influenza viral sequences to capture their evolutionary characteristics. A consensus maximum clade credibility (MCC) tree was estimated from the samples, together with their amino acid substitution statistics, from which we generated synthetic sequences of known dependent and independent paired amino acid sites. A pair-to-pair and influenza-specific amino acid substitution matrix (P2PFLU) incorporated into Bayesian Evolutionary Analysis Sampling Trees (BEAST) enumerated these synthetic sequences. The sequences inherited evolutionary features and co-varying characteristics from the real viral sequences, rendering these synthetic data ideal for exploring their co-evolving features. For the MI measure, we proposed a novel metric called the empirical MI (MI(Em)), which outperformed other MI measures in analysis of receiver operating characteristics (ROC). We implemented our approach on 1086 all-time PB2 sequences of influenza A H5N1 viruses, in which we found 97 sites exhibiting co-evolutionary substitution of one or more amino acid sites. In particular, PB2 451, along with eight other PB2 sites of various MI(Em) scores, was found to co-evolve with PB2 627, a known species-associated amino acid residue which plays a critical role in influenza virus replication.

Inference of genotype-phenotype relationships in the antigenic evolution of human influenza A (H3N2) viruses

Distinguishing mutations that determine an organism's phenotype from (near-) neutral ‘hitchhikers’ is a fundamental challenge in genome research, and is relevant for numerous medical and biotechnological applications. For human influenza viruses, recognizing changes in the antigenic phenotype and a strains' capability to evade pre-existing host immunity is important for the production of efficient vaccines. We have developed a method for inferring ‘antigenic trees’ for the major viral surface protein hemagglutinin. In the antigenic tree, antigenic weights are assigned to all tree branches, which allows us to resolve the antigenic impact of the associated amino acid changes. Our technique predicted antigenic distances with comparable accuracy to antigenic cartography. Additionally, it identified both known and novel sites, and amino acid changes with antigenic impact in the evolution of influenza A (H3N2) viruses from 1968 to 2003. The technique can also be applied for inference of ‘phenotype trees’ and genotype–phenotype relationships from other types of pairwise phenotype distances.

The molecular evolution of any organism is described by changes in the genotype resulting from genetic drift or selection to maintain or establish fitness under the given environmental conditions. Identification of phenotype-defining changes and their distinction from (near-) neutral (‘hitchhikers’) ones is a fundamental challenge in genome research. The standard approach involves time- and cost-intensive mutation experiments, which are typically low throughput, due to their experimental nature. We have developed a computational method for the inference of phenotypic impact of genotypic changes that is applicable to any system, within or across species, where homologous genetic sequences and associated pairwise phenotype distances are available. We demonstrate the accuracy of our method by application to the human influenza A (H3N2) virus. This exemplary system is of particular interest, as recognizing changes in the antigenic phenotype and a viral strains' capability to evade pre-existing host immunity is important for the production of efficient vaccines. We accurately identified known sites and amino acid changes with antigenic impact over 35 years of evolution, and provide further details on individual antigenically relevant changes in the evolution of influenza A (H3N2) viruses.

Free-energy simulations reveal that both hydrophobic and polar interactions are important for influenza hemagglutinin antibody binding

Antibodies binding to conserved epitopes can provide a broad range of neutralization to existing influenza subtypes and may also prevent the propagation of potential pandemic viruses by fighting against emerging strands. Here we propose a computational framework to study structural binding patterns and detailed molecular mechanisms of viral surface glycoprotein hemagglutinin (HA) binding with a broad spectrum of neutralizing monoclonal antibody fragments (Fab). We used rigorous free-energy perturbation (FEP) methods to calculate the antigen-antibody binding affinities, with an aggregate underlying molecular-dynamics simulation time of several microseconds (∼2 μs) using all-atom, explicit-solvent models. We achieved a high accuracy in the validation of our FEP protocol against a series of known binding affinities for this complex system, with <0.5 kcal/mol errors on average. We then introduced what to our knowledge are novel mutations into the interfacial region to further study the binding mechanism. We found that the stacking interaction between Trp-21 in HA2 and Phe-55 in the CDR-H2 of Fab is crucial to the antibody-antigen association. A single mutation of either W21A or F55A can cause a binding affinity decrease of ΔΔG > 4.0 kcal/mol (equivalent to an ∼1000-fold increase in the dissociation constant K(d)). Moreover, for group 1 HA subtypes (which include both the H1N1 swine flu and the H5N1 bird flu), the relative binding affinities change only slightly (< ±1 kcal/mol) when nonpolar residues at the αA helix of HA mutate to conservative amino acids of similar size, which explains the broad neutralization capability of antibodies such as F10 and CR6261. Finally, we found that the hydrogen-bonding network between His-38 (in HA1) and Ser-30/Gln-64 (in Fab) is important for preserving the strong binding of Fab against group 1 HAs, whereas the lack of such hydrogen bonds with Asn-38 in most group 2 HAs may be responsible for the escape of antibody neutralization. These large-scale simulations may provide new insight into the antigen-antibody binding mechanism at the atomic level, which could be essential for designing more-effective vaccines for influenza.

Mapping of H3N2 influenza antigenic evolution in China reveals a strategy for vaccine strain recommendation

One of the primary efforts in influenza vaccine strain recommendation is to monitor through gene sequencing the viral surface protein haemagglutinin (HA) variants that lead to viral antigenic changes. Here we have developed a computational method, denoted as PREDAC, to predict antigenic clusters of influenza A (H3N2) viruses with high accuracy from viral HA sequences. Application of PREDAC to large-scale HA sequence data of H3N2 viruses isolated from diverse regions of Mainland China identified 17 antigenic clusters that have dominated for at least one season between 1968 and 2010. By tracking the dynamics of the dominant antigenic clusters, we not only find that dominant antigenic clusters change more frequently in China than in the United States/Europe, but also characterize the antigenic patterns of seasonal H3N2 viruses within China. Furthermore, we demonstrate that the coupling of large-scale HA sequencing with PREDAC can significantly improve vaccine strain recommendation for China.

Rohlin distance and the evolution of influenza A virus: weak attractors and precursors

The evolution of the hemagglutinin amino acids sequences of Influenza A virus is studied by a method based on an informational metrics, originally introduced by Rohlin for partitions in abstract probability spaces. This metrics does not require any previous functional or syntactic knowledge about the sequences and it is sensitive to the correlated variations in the characters disposition. Its efficiency is improved by algorithmic tools, designed to enhance the detection of the novelty and to reduce the noise of useless mutations. We focus on the USA data from 1993/94 to 2010/2011 for A/H3N2 and on USA data from 2006/07 to 2010/2011 for A/H1N1. We show that the clusterization of the distance matrix gives strong evidence to a structure of domains in the sequence space, acting as weak attractors for the evolution, in very good agreement with the epidemiological history of the virus. The structure proves very robust with respect to the variations of the clusterization parameters, and extremely coherent when restricting the observation window. The results suggest an efficient strategy in the vaccine forecast, based on the presence of "precursors" (or "buds") populating the most recent attractor.

Structural basis and sequence co-evolution analysis of the hemagglutinin protein of pandemic influenza A/H1N1 (2009) virus

Severe pandemic influenza A H1N1 (2009) infection, especially in the lower respiratory tract, is often associated with the virus carrying a D222G substitution in the hemagglutinin (HA) protein of the virus. The mechanism for this association has not been fully explored. In the in vitro binding assay, it was found that clinical isolates carrying D222G substitution exhibit higher binding avidity to 2,3-linked sialic acids than the wild-type virus. The receptor binding pocket of the pandemic influenza (H1N1) HA was found to be smaller than those of other influenza A strains, allowing tighter binding of the virus with the receptor, yet also inducing steric stress for the binding. Our homology modeling and molecular docking calculations implicated that residue 222 may affect the positioning of the conserved Q223 residue, hence modulating flexibility of the binding pocket and steric hindrance during receptor binding. The molecular property of residue 222 can also directly influence the 'lysine fence' via the polarity of the amino acid residue where D222G substitution will enhance the electrostatic interactions between the receptor and the protein. The potential importance of residue 222 was illustrated by evolutionary analysis, which showed that this site is under intense selection pressure during adaptation of the virus to human host. Our findings provide a useful reference for follow-up studies in monitoring the ongoing evolution of the pandemic influenza A H1N1 (2009) virus.

Allele dynamics plots for the study of evolutionary dynamics in viral populations

Phylodynamic techniques combine epidemiological and genetic information to analyze the evolutionary and spatiotemporal dynamics of rapidly evolving pathogens, such as influenza A or human immunodeficiency viruses. We introduce ‘allele dynamics plots’ (AD plots) as a method for visualizing the evolutionary dynamics of a gene in a population. Using AD plots, we propose how to identify the alleles that are likely to be subject to directional selection. We analyze the method’s merits with a detailed study of the evolutionary dynamics of seasonal influenza A viruses. AD plots for the major surface protein of seasonal influenza A (H3N2) and the 2009 swine-origin influenza A (H1N1) viruses show the succession of substitutions that became fixed in the evolution of the two viral populations. They also allow the early identification of those viral strains that later rise to predominance, which is important for the problem of vaccine strain selection. In summary, we describe a technique that reveals the evolutionary dynamics of a rapidly evolving population and allows us to identify alleles and associated genetic changes that might be under directional selection. The method can be applied for the study of influenza A viruses and other rapidly evolving species or viruses.

Bench-to-bedside review: bacterial pneumonia with influenza - pathogenesis and clinical implications

Seasonal and pandemic influenza are frequently complicated by bacterial infections, causing additional hospitalization and mortality. Secondary bacterial respiratory infection can be subdivided into combined viral/bacterial pneumonia and post-influenza pneumonia, which differ in their pathogenesis. During combined viral/bacterial infection, the virus, the bacterium and the host interact with each other. Post-influenza pneumonia may, at least in part, be due to resolution of inflammation caused by the primary viral infection. These mechanisms restore tissue homeostasis but greatly impair the host response against unrelated bacterial pathogens. In this review we summarize the underlying mechanisms leading to combined viral/bacterial infection or post-influenza pneumonia and highlight important considerations for effective treatment of bacterial pneumonia during and shortly after influenza.