Metagenomic Characterisation of Putatively Novel Cereal Crop Virus Candidates in Ukraine and Azerbaijan: Implications for Phytosanitary Risk Assessment

The study was conducted in agrocenoses of major cereal crops (winter wheat, cultivar Podolianka; barley, cultivar Deviatyi Val; and maize, cultivar Bereka) in Ukraine and Azerbaijan during 2020-2023. The selection of Ukraine and Azerbaijan was justified by the need to compare the viral complexes of cereals within the Black Sea-Caucasus macroregion under conditions of similar study objects (leading cereal crops and comparable agrotechnological approaches) but contrasting climatic and agricultural regimes, including gradients of moisture availability and aridity, temperature background, frequency of irrigation, and technological intensity. This design made it possible to disentangle the contributions of agroclimatic factors and technological practices to virome formation and to identify viral taxa with potentially increased phytosanitary risk under different cultivation conditions.

The field stage was carried out during the growing seasons from April 2020 to September 2023. Sampling was carried out during periods of maximum representativeness of phytoviral load in the above-ground plant biomass and peak activity of vectors: for winter wheat, May-June; for barley, May-June; and for maize, July-August (with an additional control sampling at the end of the season in September, if required). The research was conducted at the Institute of Plant Protection of the National Academy of Agrarian Sciences of Ukraine (Kyiv) and at the Research Institute of Plant Protection and Technical Crops of the Ministry of Agriculture of the Republic of Azerbaijan (Ganja).

The selection of agroclimatic zones was based on the principle of covering key ecological gradients that determine the structure of viral communities and the circulation of vectors, including humidity, temperature regime, seasonal contrast, the need for irrigation, and the intensity of agrotechnological practices. In Ukraine, 24 composite samples were formed from three zones: the Southern Steppe (Odesa and Mykolaiv regions) as a model of warmer and drier conditions; the Forest-Steppe (Cherkasy and Vinnytsia regions) as a model of moderate conditions with a high proportion of intensive cultivation; and Polissia (Zhytomyr and Chernihiv regions) as a model of more humid conditions. In Azerbaijan, 20 composite samples were formed from the main cereal-producing regions: the Southern region (Nakhchivan) as a model of arid conditions; the Shamakhi Valley as a model of a temperate climate; and the Ganja Highlands as a model of stable cereal production conditions with a distinct altitudinal and climatic profile.

To ensure the comparability of the agrotechnological context, each field was described using a unified observation sheet that included the type of farming system (intensive monoculture systems or traditional/less intensive fields), the duration of monoculture cultivation, crop rotation schemes, the application of glyphosate-based herbicides, the irrigation system (none, surface irrigation, or other systems), as well as basic soil and climatic characteristics. The areas of the fields included in the sampling were as follows: in Ukraine, 10-120 ha (Southern Steppe: 30-120 ha; Forest-Steppe: 20-90 ha; Polissia: 10-70 ha); in Azerbaijan, 5-40 ha (Nakhchivan: 5-30 ha; Shamakhi Valley: 6-35 ha; Ganja Highlands: 8-40 ha). This approach made it possible to include both intensive production areas and less intensive fields without altering the sampling logic.

The number and distribution of crops and cultivars across the zones were as follows (in all cases, the same cultivars were used: winter wheat Podolianka, barley Deviatyi Val, and maize Bereka). In Ukraine, a total of 24 composite samples were collected: winter wheat, n = 8; barley, n = 8; maize, n = 8. In the Southern Steppe, n = 8 samples were formed (winter wheat n = 3; barley n = 2; maize n = 3); in the Forest-Steppe, n = 8 (winter wheat n = 3; barley n = 3; maize n = 2); and in Polissia, n = 8 (winter wheat n = 2; barley n = 3; maize n = 3). In Azerbaijan, a total of 20 composite samples were collected: winter wheat, n = 7; barley, n = 7; maize, n = 6. In Nakhchivan, n = 7 samples were formed (winter wheat n = 3; barley n = 2; maize n = 2); in the Shamakhi Valley, n = 7 (winter wheat n = 2; barley n = 3; maize n = 2); and in the Ganja Highlands, n = 6 (winter wheat n = 2; barley n = 2; maize n = 2). Such a distribution ensured a balance of crops within each country and adequate representation of each agroclimatic zone/region for comparative statistical procedures and the interpretation of metagenomic profiles.

Composite samples were formed by combining above-ground parts of cereal crops, small accompanying plants, plant residues, and small invertebrates present in the above-soil layer at the time of sampling. Fungal structures associated with plant residues and, where present, nematode-associated material from the near-surface plant matrix were also retained as part of the composite sample, since these organisms may contribute to soil- or residue-mediated viral circulation. To ensure the comparability of the material and to prevent systematic bias in metagenomic profiles, exclusion criteria were established. A sample was not included in further analysis if: (1) signs of external contamination were recorded during sampling or transport (contact with non-sterile surfaces or instruments outside the established procedure, substantial soil contamination, or mixing of material between fields); (2) the sample had insufficient mass or volume of biomaterial for standardised ribonucleic acid (RNA) extraction and library preparation, or was formed with a substantial disproportion of components that violated the unified principle of “compositeness”; (3) storage conditions were violated (lack of cooling or freezing, thawing during transport, or prolonged exposure to room temperature); (4) laboratory quality control revealed RNA of inadequate quality for metagenomic sequencing (pronounced degradation, low concentration or yield), which made library preparation impossible; or (5) the sample contained a disproportionately high contribution of a single dominant non-target biological component (for example, large fragments of one accompanying plant species or a high abundance of invertebrates), rendering it unrepresentative of the above-soil layer of the field. In cases of exclusion, repeat sampling was carried out within the same field and crop at the nearest available time to maintain a balanced distribution of samples across countries, zones, and crops.

Within each site, 25-point samples were collected and combined into a single composite sample. The 25-point scheme was selected to capture within-field microheterogeneity in plant condition, residue distribution, and potential vector occurrence while maintaining a comparable sampling unit across countries, crops, and agroclimatic zones. Points were distributed across each field to avoid edge effects and localised disease foci, thereby reducing the probability that the metagenomic profile would be dominated by a single infected plant or a single microhabitat. The material was frozen in liquid nitrogen at approximately -196 ℃ and transported to the laboratory under a maintained cold chain at a temperature not exceeding -20 ℃. Prior to RNA extraction, the samples were homogenised in liquid nitrogen to obtain a uniform mass, and equal subsample weights were taken from the homogenate for RNA extraction, which ensured cellular integrity and prevented nucleic acid degradation.

Total RNA was extracted from 100-200 mg of homogenised material using a commercial kit for plant tissues with additional purification steps to remove inhibitors. RNA quality and quantity were assessed by spectrophotometry using a NanoDrop 2000 instrument and by capillary electrophoresis using an Agilent 2100 BioAnalyzer. Only samples with an RNA integrity number of at least 7.0 were accepted for further analysis.

This threshold was used as the main RNA-quality criterion before rRNA depletion and library preparation. In addition to RIN, RNA concentration, A260/280 and A260/230 absorbance ratios, electrophoretic profile, and the absence of pronounced degradation were checked for each sample. Samples that did not meet these quality requirements or provided insufficient RNA yield were excluded from sequencing or re-extracted before library construction. The retained libraries subsequently showed high sequencing quality, as confirmed by Q30 values exceeding 90%, supporting the reliability of downstream taxonomic and genome-reconstruction analyses.

To enrich the viral and non-coding fraction, protocols for the depletion of plant and bacterial ribosomal RNA were applied using the commercial QIAseq FastSelect – rRNA Plant kit (QIAGEN) [11]. Complementary deoxyribonucleic acid (DNA) libraries were prepared from the purified RNA and sequenced on the Illumina NovaSeq 6000 platform, generating paired-end reads of approximately 150 bp per sample. The target sequencing depth, defined as “hundreds of millions of paired-end reads per library”, was considered sufficient to achieve two objectives simultaneously: sensitive detection of low-abundance viruses and de novo reconstruction of complete or near-complete viral genomes. In the composite matrices of agrocenoses, the predominant proportion of nucleic acids is usually derived from the host plants, accompanying vegetation, microbiota, and the eukaryotic fauna of the above-soil layer. Accordingly, the viral fraction may be represented either abundantly (for dominant taxa) or at very low levels (for rare or transient viruses). Under such conditions, an increase in the total number of sequencing reads enhances the probability of detecting low-abundance taxa after the stages of filtering, trimming, and removal of host-derived sequences and also provides sufficient coverage for the assembly of longer contigs and the reconstruction of genomes without gaps. Thus, the selected data volume was consistent with the subsequent virus identification strategy, which relied not only on the taxonomic classification of reads but also on the assembly and analysis of virus-like contigs and genomes, a process that requires adequate sequencing depth. In total, more than 1.2 gigabases of paired-end reads were generated for the entire study, and the quality metrics of the libraries (in particular, the proportion of Q30 bases) confirmed the suitability of the dataset for sensitive taxonomic analysis and subsequent genome reconstruction.

Raw sequencing data were cleaned of adaptor sequences and low-quality nucleotides using Trimmomatic (version 0.39) and FastQC (version 0.11.9); reads with quality scores below the defined threshold and excessively short fragments were removed. To eliminate host-derived sequences, the reads were mapped to the reference genomes of the main cereal crops, as well as to available reference genomes of accompanying species, using Bowtie2 (version 2.4.4) and BWA (version 0.7.17). To assess library quality, Q20 and Q30 metrics, average read length, and guanine-cytosine (GC) content were calculated using FastQC and GATK (Genome Analysis Toolkit, version 4.2.6.1). Filtered reads were then assembled de novo using metagenomic algorithms, notably SPAdes (version 3.14.0), and the resulting contigs were used as the basis for subsequent viral annotation. The taxonomic assignment of reads and contigs was performed by aligning them against reference databases of viruses, bacteria, and eukaryotes using Diamond (version 0.9.32) and BLAST+ (version 2.9.0). Virus-like contigs underwent refined annotation, and candidate genomes were selected if they covered the majority of the predicted genome length and had sufficient coverage.

To reduce the risk of false assignment caused by reagent contamination, cross-sample contamination, or misannotation of host-derived sequences, candidate viral contigs were interpreted conservatively. Contigs were retained only when they showed consistent viral-domain matches, sufficient read coverage, coherent open reading frame (ORF) organisation, and no stronger assignment to plant, fungal, bacterial, or other non-viral sequences in comparative database searches. Since external negative controls from certified virus-free plant material or sterile growth-chamber samples were not included in the sequencing run, the identified sequences were classified as putatively novel viral genome candidates rather than as experimentally confirmed novel viral species.

Viral genomes were considered complete or near-complete if their length was consistent with expectations and they contained a full complement of predicted ORFs, as determined using GeneMarkS (version 4.34) and Prodigal (version 2.6.3).

Virus assignments were not based on a single alignment output alone. The results obtained with BLAST+ and Diamond were compared for consistency at the family and genus levels, and ambiguous contigs were retained only when viral attribution was additionally supported by coherent genome organisation, predicted ORF structure, conserved protein domains, and phylogenetic placement. The <90% amino acid identity threshold was therefore used as an initial criterion for recognising putatively novel viral genome candidates, rather than as an isolated taxonomic decision. Where genome architecture allowed, additional structural signals, including tRNA-like elements or promoter-associated motifs reported for related viral groups, were considered as supportive evidence; however, these features were not treated as universal criteria because they are not present or diagnostically informative across all plant virus families.

For each genome, the genomic organisation, ORF composition, and the presence of atypical structural or functional motifs were further analysed using InterProScan (version 5.49-82.0) and Pfam (version 33). For selected virus groups, multiple alignments of the amino acid sequences of key proteins were constructed using MAFFT (version 7.490) and Clustal Omega (version 1.2.4). Optimal evolutionary models were determined with ModelTest-NG (version 0.1.7), and phylogenetic trees with bootstrap support were reconstructed using RAxML (version 8.2.12) and IQ-TREE (version 2.1.2). The selection of evolutionary models was performed separately for each alignment of key viral proteins, as different taxonomic groups and different proteins (for example, replicase domains and capsid proteins (CP)) can vary substantially in amino acid substitution frequencies, degree of conservation, and saturation levels. For this reason, a “data→model” approach was applied: based on multiple alignments of amino acid sequences, the optimal substitution model was determined in ModelTest-NG by comparing alternative candidate models and selecting the best-fitting model according to information criteria (minimising Bayesian Information Criterion (BIC)/Akaike Information Criterion (AIC)), accounting for among-site rate heterogeneity (γ-distribution) and, where necessary, additional parameters such as empirical amino acid frequencies. The selected model (or its equivalent implementation in a given programme) was then applied in both RAxML and IQ-TREE to avoid biases in topology and bootstrap support caused by model misspecification, which is critical for viral genomes with high evolutionary rates and uneven substitution patterns across sites. To detect recombination events, complete or near-complete genomes were analysed using multiple algorithms, including RDP4 (version 4.97), SimPlot (version 3.5.1), GENECONV (version 1.81), and BootScan. Recombination signals were considered only if supported statistically and consistent across methods.

For the risk-scoring procedure, recombination evidence was considered significant only when the same event was detected by at least two independent algorithms and when the inferred breakpoint positions were broadly concordant across methods. The algorithms were not treated as equally weighted single votes; instead, priority was given to events supported by complementary approaches, such as similarity-plot shifts, phylogenetic incongruence, and statistically significant recombination signals.

Alpha diversity of viral communities was assessed using the abundance of viral taxa normalised by sequencing depth, and indices of species richness and evenness were calculated using QIIME2 (version 2020.8) and the Vegan R package. Comparisons of metrics between countries, agrocenosis types, and crops were conducted using non-parametric statistical tests in R (version 4.0.5) with the Vegan package. The taxonomic structure of viromes was described at the family, genus, and species levels, highlighting a shared core as well as taxa unique to each country. For each sample, structural characteristics of the virome were compared with metadata on farming systems. The frequency of detection of individual viruses and viral families was assessed in monocultures and crop rotations, on irrigated and non-irrigated plots, and under different herbicide application regimes. Statistical associations between viral metrics and agrotechnological factors were evaluated through correlation analyses and comparison of group distributions.

To strengthen the interpretation of these associations, an observational causal-inference framework was additionally applied. Irrigation type, farming system, monoculture duration, and herbicide regime were treated as explanatory agricultural factors, while Shannon and Chao1 indices were treated as response variables reflecting virome evenness and richness. Country, crop species, agroclimatic zone, and sequencing depth were considered potential confounding variables. This framework was used to distinguish stable associations from simple bivariate correlations; however, because the study did not involve experimental manipulation of agricultural practices, causal conclusions were interpreted cautiously.

Prior to selecting parametric or non-parametric procedures, the normality of quantitative variables was assessed using the Shapiro-Wilk test to identify deviations and outliers. For variables exhibiting an approximately normal distribution (p > 0.05 in the Shapiro-Wilk test) without pronounced skewness, correlations between continuous variables were calculated using Pearson’s method. In cases of non-normality (p ≤ 0.05), the presence of ordinal variables or marked non-linearity in relationships, Spearman’s rank correlation was applied. To compare viral metrics between two independent groups (for example, intensive versus traditional farming, presence versus absence of irrigation, or glyphosate application versus no application), a parametric t-test was used when assumptions of normality and homogeneity of variances were met, or the Mann-Whitney U test was applied when these assumptions were violated. For comparisons involving three or more groups, analysis of variance (ANOVA) or the Kruskal-Wallis test was employed, respectively.

To identify potential vectors of novel viruses, the eukaryotic fraction of the metagenomic data was analysed using MetaPhlAn (version 3.0). Reads assigned to insects, mites, and other invertebrates were classified to the genus or species level using Kraken2 (version 2.1.2) and Centrifuge (version 1.0.4). For key novel viruses, the frequency of co-detection with individual taxa of potential vectors was calculated within the same samples. Such “virus-potential vector” pairs were treated as hypotheses regarding transmission routes for further validation.

For a systematic assessment of the risk associated with newly discovered viruses, an aggregated scoring system was developed based on four key criteria. Criterion (A), “transboundary potential”, was scored as 1 if the same novel virus (or a closely related lineage within a single monophyletic cluster in the phylogenetic trees of key proteins) was detected in at least one sample from Ukraine and at least one sample from Azerbaijan. If the virus was not independently detected in both countries, a score of 0 was assigned. This approach aligned with the interpretation of “detected in both countries in similar, but not identical, variants” for priority isolates, where the risk was defined by the presence of potential transboundary spread rather than complete sequence identity. Criterion (B), “relatedness to economically important pathogens”, was determined by comparing the amino acid sequences of key proteins or polyproteins of novel viruses with reference sequences of known phytopathogens that have an established economic impact (noted as causal agents of major diseases in cereals or related crops). The criterion was considered met (1 point) if the best match in amino acid identity exceeded 85% over the aligned regions, interpreted as a high degree of evolutionary relatedness to already known pathogenic lineages and, therefore, an increased likelihood of similar biological properties, including host range and epidemic potential. The >85% threshold was used solely as a “trigger” for prioritisation, not as a taxonomic criterion for species demarcation. Criterion (C), “presence of a vector”, was formalised based on the taxonomic composition of the eukaryotic fraction of the metagenomic data, with a focus on insect and mite taxa known to act as vectors of plant viruses. The criterion was considered confirmed (1 point) if co-detection was observed between a virus and a taxon with known or likely vector status. The frequency of co-detection was calculated as the percentage of virus-positive samples in which sequences of the corresponding potential vector were simultaneously detected. In the absence of co-detection with taxa relevant to vector transmission, or where no biologically plausible vector status existed, a score of 0 was assigned. Criterion (D), “evidence of recombination”, was assessed using complete or near-complete genomes with multiple algorithms (RDP4, SimPlot, GENECONV, BootScan), and recombination signals were considered only if statistically supported and reproducible/consistent across methods. Thus, 1 point was assigned only when recombination was confirmed as a genuine signal, rather than an artefact of a single method, whereas 0 points were assigned in the absence of statistically supported evidence. The detection of recombination events in 14 genomes resulted specifically from the application of this multi-approach procedure. Each criterion was scored either 0 or 1, with fulfilment of the criterion receiving 1 point and non-fulfilment receiving 0 points. After evaluating all criteria, the total score was used to calculate an integrated risk index, allowing the prioritisation of viruses for subsequent phytosanitary measures. Viruses with a risk index of 4 were classified as high risk and required immediate inclusion in monitoring systems and the implementation of appropriate measures. Viruses scoring 3 were considered medium risk and required further study, while those scoring 2 were regarded as low risk but still warranted observation.

Comparisons of diversity indices between groups were conducted using the non-parametric Mann-Whitney U test in R (version 4.x). Before selecting parametric or non-parametric procedures, the normality of quantitative variables (including diversity/evenness indices and other viral metrics) was assessed using the Shapiro-Wilk test, with deviations additionally checked visually via Q-Q plots. Since deviations from normality were observed for diversity indices and some related metrics, differences between groups were evaluated using the non-parametric Mann-Whitney U test (for two independent groups). Statistical associations between viral metrics and agrotechnological factors were assessed through correlation analysis. Spearman’s rank correlation was used as the primary measure of association, given that agrotechnological variables included ordinal or categorical characteristics and/or did not conform to a normal distribution. In cases where both variables were approximately normally distributed (p > 0.05 in the Shapiro-Wilk test) and the relationship was linear, Pearson’s correlation could be applied; however, for consistency across analyses, Spearman’s approach was prioritised. The threshold for statistical significance was set at p < 0.05.

Comparing the results of this study with other publications, it is evident that the data obtained on viral communities in Ukraine and Azerbaijan reflect trends similar to those reported in other countries. For instance, Gupta [29] highlighted the role of metagenomics in the detection of novel plant viruses, significantly improving diagnostic capabilities and the monitoring of viral threats. The findings of the present study, particularly the identification of new viruses within the Potyviridae and Virgaviridae families, corroborate the existence of comparable viral threats in cereal agroecosystems observed in other regions worldwide. Specifically, Yousuf et al. [30] emphasised the importance of metagenomic approaches for investigating agricultural microbiomes and viral communities, noting that advanced sequencing technologies provide new opportunities for studying such threats. These observations are consistent with the results presented here, which demonstrate that metagenomic methods not only facilitate virus identification but also reveal their potential impacts on agricultural ecosystems. This approach has enhanced understanding of viral evolution and allowed assessment of their influence on the sustainable development of agroecosystems. The application of virus-associated nucleic acids in metagenomic analyses to gain deeper insight into viral communities, as explored by Moubset et al. [31], is similarly reflected in this study, where a comparable methodology was employed for the selection and analysis of viral sequences. The detection of novel viruses in cereal agroecosystems in Ukraine and Azerbaijan aligns with patterns described in studies employing metagenomic sequencing methods for virus detection and classification.

The detection of 14 recombinant viral genomes also has important evolutionary implications. Recombination may arise when genetically related viruses co-infect the same host plant or circulate within the same agroecosystem through shared vectors, crop residues, or soil-associated reservoirs. In this study, the concentration of recombination signals among potyviruses suggests that mixed infections and vector-mediated circulation may create favourable conditions for genome exchange. The localisation of recombination points near functional regions of the polyprotein may also indicate that recombination is not merely a random event, but a possible mechanism for generating variants with altered replication efficiency, vector compatibility, or host adaptation [32, 33]. However, these mechanisms remain inferential because host-range assays and experimental infection studies were not conducted. Therefore, the recombinant genomes identified here should be interpreted as evolutionary candidates that require functional validation to determine whether recombination affects pathogenicity, host specificity, or ecological fitness.

In the study by Piombo et al. [34], the importance of metagenomic approaches for identifying emerging pathogens was also emphasised, particularly under conditions of climate change and intensive agriculture, which is directly relevant to the present research. Such approaches enable not only the identification of known viruses but also the detection of new viral pathogens that may arise through evolutionary changes in existing viral populations. Investigations of viral communities in weeds and wild host plants, conducted by Hasiów-Jaroszewska et al. [35], further demonstrated the efficacy of metagenomics in characterising the diversity and ecology of viruses in uncultivated reservoirs, which can be crucial for understanding infection sources in agroecosystems. A review by Kreuze et al. [36] specifically highlighted that those new technologies, including metagenomics, provide innovative opportunities to enhance understanding of major viral diseases threatening global food security, which fully supports the practical conclusions of the present study regarding the need to integrate such methods into national phytosanitary monitoring systems.

It is also noteworthy that the findings of this study, which identified novel viral species within the Potyviridae and Virgaviridae families, correspond with the results reported by Wamaitha et al. [37], who conducted a metagenomic analysis of viruses associated with maize lethal necrosis in Kenya, providing a relevant comparison for evaluating viral threats across different regions. These findings underscore the importance of monitoring viral infections across different climatic zones, enabling the timely detection of emerging threats. The study by Mohsin et al. [38] highlighted the potential of viral metagenomics as a monitoring tool for the early identification of novel pathogens. This reinforces the significance of the metagenomic sequencing methods employed in the present study for the rapid detection of new viruses that may pose a risk to agricultural crops. Attention to emerging viral threats in rice was emphasised in the research of Ding et al. [39], where the authors reviewed decades of discoveries in the field and their implications for crop protection. Comparing these findings with those of the current study suggests that similar challenges exist regarding novel viruses in cereal crops, as these pathogens can cause substantial damage to wheat and maize in Ukraine and Azerbaijan. A common feature across these studies is the emphasis on early detection of new viruses using metagenomic technologies, which allows for forecasting and mitigating their potential impact. In their review, Sharuddin et al. [40] highlighted the role of metatranscriptomics in agriculture, particularly in the context of sustainable farming and the improvement of crop productivity. The results of this study, which also applied metagenomic approaches, demonstrated the effectiveness of these methods for identifying emerging viral threats. Such approaches can be applied to expand current knowledge and improve control of viruses that adversely affect cereal crop yields in Ukraine and Azerbaijan.

The application of metagenomic sequencing for the surveillance of viral diseases associated with food and water, which is relevant to the study of viral infections affecting agricultural crops, was discussed by Nieuwenhuijse and Koopmans [41]. This study reinforces the necessity of employing metagenomics for virus detection, as it enables the identification of novel pathogens that may be transmitted via the environment or other pathways, similar to the viruses identified in cereal crop agroecosystems in the present research. Giolai et al. [42] focused on measuring airborne metagenomic diversity in agricultural ecosystems, providing insights into ecological conditions that facilitate the spread of viral infections. Their findings are pertinent to this research, as they highlight the significance of environmental factors in viral dissemination and offer a foundation for further investigation into the influence of agroclimatic conditions on viral communities in cereal crops. The detection of a novel virus transmitted by aphids and infecting rice, also identified through metagenomic analysis, was reported by Yan et al. [43]. The discovery of new viruses and their transmission routes in plant systems parallels the findings of this study, in which novel viral species were detected in cereal crop agroecosystems and are likely transmitted via vectors.

Islam [44] emphasised the role of genomic monitoring in combating emerging plant diseases, including wheat viruses, which is directly relevant to the analysis of novel viral threats in cereal crops. The study provided recommendations for managing viral diseases that could be adapted to the specific conditions of Ukraine and Azerbaijan, particularly regarding the establishment of monitoring systems and the detection of new viruses in cereal agroecosystems. In the research of Smadi et al. [45], metagenomic analysis was employed to detect viruses in mixed fruit orchards, assessing viral diversity through samples collected by honeybees (Apis mellifera). This study demonstrated the application of metagenomics for investigating viral communities not only in agroecosystems but also in natural ecosystems, highlighting the importance of virus monitoring across different ecological levels. Similar to the present research, the detection of vector-transmitted viruses underlines the role of biological agents in the spread of viral diseases [46, 47]. Viral threats to wheat and innovative approaches to their diagnosis and management were reviewed by Tanu et al. [48], which is also relevant to this study, as the focus here is on viruses of cereal crops. The authors emphasised the importance of integrated diagnostic approaches that combine metagenomic technologies with novel methods for rapid detection of viral diseases, aligning closely with the findings of the current research. The study by Muruu [49] focused on the detection of taro viruses in Kenya using next-generation sequencing, demonstrating the effectiveness of modern approaches for identifying novel plant pathogens under variable climatic conditions. This methodology could be adapted for similar research in Ukraine and Azerbaijan, where local climate and specific pathogens play a key role in shaping the diversity of viral species in cereal agroecosystems. Scott et al. [50] highlighted the impact of plant diseases on food security in the twenty-first century, emphasising the growing threat posed by viral infections to global food systems. This aligns with the findings of the present study, which demonstrated that viral pathogens can significantly compromise the stability of food systems by affecting the yield and quality of cereal crops.

Thus, a comparison of the results of this study with existing scientific literature confirms the value of a metagenomic approach for investigating viral communities in cereal crops and demonstrates the effectiveness of contemporary methods for identifying emerging viral threats. The data obtained from Ukraine and Azerbaijan correspond with findings from other studies that underscore the high diversity of viruses in cereal agroecosystems and the critical importance of monitoring these threats to safeguard food security. Given the identification of new viral species and potentially hazardous pathogens, it is evident that a systematic metagenomic approach is essential for the effective management of viral threats in agriculture. These results provide a basis for outlining subsequent steps in the development of strategies for monitoring and managing viral infections.