Evolutionary graph pangenome of the order Poales from chloroplast genomes highlight phylogenetic inconsistencies

Evolutionary graph pangenome of the order Poales from chloroplast genomes highlight phylogenetic inconsistencies | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Evolutionary graph pangenome of the order Poales from chloroplast genomes highlight phylogenetic inconsistencies Rakan Haib, Sariel Hübner This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8786947/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 9 You are reading this latest preprint version Abstract With the increased accessibility and democratization of genomic sequencing, the number of fully sequenced genomes across all taxonomies is exploding. To benefit from the availability of complete genomes in evolutionary studies a complex frameworks such as graph-pangenomes are increasingly embraced. However, the high computational cost and precarious interpretation of graph-pangenomes at higher taxonomic levels hinder their broad implementation in evolutionary genomics. Here, we describe the development and application of an evolutionary graph pangenomes aproach which facilitates the analysis and interpretation of diversity across species and families. We applied our approach to 709 chloroplast genomes spanning the order Poales , overcoming key limitations of traditional phylogenetic approaches by uncovering structural variation and evolutionary signals. Our results show that despite overarching structural conservation, most genomic diversity arises from species-specific “cloud” variation. We also recover from the graph, a 77.7Kbp consensus core genome for Poales encoding essential cellular functions, while accessory regions capture adaptive traits. Graph-theory metrics were further implemented revealing distinct genus-level evolutionary signatures, with Triticum appearing homogeneous and highly connected, in contrast to the divergent, lineage-specific structure of Elymus . Collectively, we propose evolutionary graph pangenomes (evographs) as a powerful and nuanced framework for resolving structural and evolutionary complexity across species and higher taxonomies. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction The rapid maturation of DNA sequencing technologies have fundamentally reshaped how genomic data is generated, represented, and interpreted. The introduction of ultra-high-throughput sequencing enabled population-scale whole-genome sequencing, while subsequent advances in long-read platforms such as PacBio and Oxford Nanopore Technologies (ONT) dramatically improved contiguity, accuracy, and affordability of genome assemblies. Highly accurate long reads (e.g., PacBio HiFi) and ultra-long ONT reads now routinely resolve repetitive and structurally complex regions, enabling haplotype phasing, and support chromosome-scale telomere-to-telomere (T2T) assemblies (Nurk et al., 2022 ; Wenger et al., 2019 ). As a result, genome assembly has shifted from producing a single representative sequence to generating multiple high-quality assemblies across individuals, populations, and species, laying the foundation for comparative and evolutionary genomics at unprecedented scale (Montenegro et al., 2017 ; Jayakodi et al., 2020 ; Zhao et al., 2018 ; Liu et al., 2020 ; Minio et al., 2019 ; Walkowiak et al., 2020 ). Most population genomic analyses so far relied on a single linear reference genome although this paradigm inherently fails to capture the full spectrum of genetic diversity within a species. Sequences that differ substantially from the reference genome, particularly around structural variation (SVs) are often misrepresented, poorly aligned, or missed entirely (Alkan et al., 2011 ; Sherman & Salzberg, 2020 ). The introduced reference bias disproportionately favors sequences that are more similar to the chosen reference. This asertainment biased is specifically problematic in highly diverse or structured populations, leading to incomplete or distorted inferences of adaptation, gene flow, and evolutionary history. The availability of multiple high-quality assemblies made clear that no single reference can adequately represent the species sequence diversity, motivating the transition to pangenome frameworks (Tettelin et al., 2005 ; Eizenga et al., 2020 ). A pangenome represents the complete genomic repertoire of a species or clade, integrating sequences shared across individuals. Pangenomes are commonly partitioned into a core fraction representing genomic segments that are present in all or nearly all individuals, and sequences that are present in only a small (cloud) or moderate (shell) subset of individuals. Graph-based pangenomes further enhance this representation by encoding multiple genomes within a unified data structure, enabling more accurate read mapping, variant calling, and genotyping, particularly in regions affected by complex structural variation (Garrison et al., 2018 ; Paten et al., 2017 ). Graph pangenomes are typically formalized as sequence or variation graphs where nodes represent conserved DNA sequence fragments with no polymorphism, while edges encode observed adjacencies between nodes. Genetic variation is represented as branching paths or loops, and a full path through the graph corresponds to an observed or inferred haplotype. This structure generalizes the linear reference genome, allowing multiple alternative sequences to coexist in a single coordinated system and providing a natural framework for representing complex variation (Eizenga et al., 2020 ). As genomic sampling has expanded to encompass multiple species, pangenomes have been extended to what are often termed “super-pangenomes”. These are typically clade or genus-level graphs that integrate interspecific variation and was recently developed for Oryza , Vitis and Hordeum (Guo et al., 2025 ; Feng et al. 2025 ; Shang et al., 2022). However, the term “super-pangenome” emphasizes scale and complexity rather than the underlying purpose and can become terminologically confusing as graphs are extended to families or orders. A more appropriate term is evolutionary graph pangenome (or evograph in short), which explicitly reflects the goal of modeling genomic diversity across evolutionary time and taxonomic depth. Such graphs can augment phylogenetics trees to model evolution at the genome scale. Evographs offer a platform for evolutionary inference directly from the graph topology. For example, the shortest or most conserved path through the graph approximate the core genome, providing insight into indispensable genomic components, while alternative paths encode lineage- or species-specific polymorphism (Hübner 2022 ). Despite their promise, graph pangenomes face substantial challenges. Tool development remains limited, and many existing frameworks, such as the Pan-Genome Graph Builder (PGGB) and related variation-graph pipelines, require careful manual parameter tuning and substantial computational resources (Garrison et al., 2024 ). Evographs can contain millions of nodes and edges and are difficult to visualize, query, and interpret posing a major bottleneck. Despite the rich evolutionary information embeded in the graph, extracting biologically meaningful insights and presenting them in an intuitive manner remains challanging. Addressing these challenges including scalable visualization tools, and clearer conceptual frameworks are key for fully realizing the potential of evographs as a unifying paradigm in genomics and evolutionary biology. The order Poales is one of the most ecologically dominant clades comprising circa 20,000 species in 16 families and representing over one third of all monocots and 7% of angiosperms worldwide (APG IV, 2016; Linder & Rudall, 2005 ). Poales are also of exceptional economic importance, encompassing major crops including rice, wheat, maize, sugarcane, barley, bamboo and others. Ecologically, Poales are abundant along terrestrial and aquatic ecosystems, including grasslands, savannas, steppes, and wetlands. Owing to its size, ecological breadth, and functional innovation, Poales and particularly memebers of the Poaceae family have become a key model for studying macroevolution, ecological adaptation, and diversification under environmental change (Christin et al. 2014 ; Potapenko et al. 2025). The earliest-diverging lineages in the Poales include Bromeliaceae , Typhaceae , and Rapateaceae , followed by diversification into two major assemblages: the Cyperid clade and the Graminid clade, the latter giving rise to Poaceae and its major BOP/BEP and PACMAD lineages (Kellogg, 2015 ). Despite decades of study, reconstructing Poales evolution remains challenging with traditional phylogenetic approaches providing ambiguous support for key relationships, particularly within Poaceae (Saarela et al., 2018 ). Moreover, tree-based models occasionally fail to capture hybridization, introgression, and gene duplication that are pervasive in grasses. Evographs can potentially provide a powerful alternative by representing the full spectrum of genomic diversity, including structural variation and presence/absence polymorphisms, across species and lineages (Eizenga et al., 2020 ; Garrison et al., 2024 ). We developed MineGraph, a dedicated pipeline for constructing and analyzing evolutionary graph pangenomes (evographs) from genome assemblies. MineGraph incorporates multiple optimization steps designed to enhance graph quality, reduce complexity, and improve interpretability. We applied this pipeline to more than 700 chloroplast genome assemblies representing most of the order Poales , revealing patterns of genomic diversity that underlie evolutionary relationships among species, genera, and families within the order. Results Optimizing the evograph-genome construction Graph construction from genome assemblies relies on an intensive all-versus-all alignment process, where the parameters critically influence the resulting graph structure. To optimize these parameters, we implemented a data-driven approach in a pipeline called MineGraph and calculated genetic distances among samples, and the minimum segment length based on the longest repeat sequence for homology search. To evaluate this approach, evographs were constructed with PGGB for 17 mitochondrial and 17 chloroplast genomes representing the same set of species within the Poales order (Table S1 ). The optimized mitochondrial evograph with MineGraph comprised of 52,953 nodes and 73,670 edges, while the version produced with default parameters in PGGB yielded 10,668 nodes and 14,632 edges, a considerable reduction in polymorphism and connectivity (Fig. 1 a). The optimized mitochondrial evograph had much higher edge density and longer homology stretches; thus, fragmentation of shared sequences was reduced in 30% (Fig. 1 b). Calculating the pangenomic proportion of sequence representation in each evograph further highlighted the differences between versions. The evograph constructed with default parameters failed to detect core sequences (nodes) that are shared among more than 95% of the genomes that were included in the graph and identified only sparse shell and cloud nodes (defined by default as represented among less than 5% of the genomes). In contrast, the optimized evograph captured a more balanced representation of the pangenome with 2,912 core nodes, 32,623 shell nodes, and 16,678 cloud nodes (Fig. 1 c). For the chloroplast genomes which are smaller, structurally simpler, and more conserved than mitochondrial genomes, extensive homology and low polymorphism was observed when using default parameters. Applying the optimized protocol improved the sensitivity of the alignment, thus yielding a total of 3,076 edges (polymorphism) and 2,068 nodes, of which 1908 are core nodes. Together, these results demonstrate that parameter adjustment and optimization substantially improve the graph construction enabling the detection of polymorphism across species (Fig. 1 a–c, Figure S3-4). Elucidating the Ploales cholorplast evograph genome Approximately 1500 chloroplast genomes of Poales species are available from the NCBI database with clear overrepresentation of agronomic important species. To obtain a non-redundant dataset, we extracted all uniqe 709 chloroplast complete genomes with one representative accession per species (Table S3). A graph was constructed from all 709 chloroplast genomes, yielding a comprehensive evograph comprised of 413,416 nodes and 685,459 edges. The obtained graph is characterized with low sparsity indicated with an average node degree of 1.66 and a low density (0.008), thus reflecting the overall structural conservation of chloroplast genomes. A pangenome representation analysis revealed a dominance of species-specific sequences in the graph with 273,023 cloud nodes (66%) detected in less than 5% of the species, 135,931 nodes (33%) corespond to the shell fraction and 4,462 core nodes (1%) that were detected in more than 95% of the species, representing highly conserved sequences in the chloroplast genome (Fig. 2 a). This distribution highlights the variation at the order level with increasing conservation at the family and genus levels. Exploring the composition of genetic polymorphism represented in the graph indicated that the most common variants are SNPs (19,876), followed by indels (2,718), and larger structural variantion (1,461). Interestingly, most of the polymorphism was identified as multiallelic variats, reflecting the extensive diversity and divergence among Poales species. The average nucleotide diversity across all species (π = 0.0326) was fairly mild compared to the variation within families (π = 0.1092–0.4594), indicating that most of the variation is obtained at the genus and species levels, further supporting the stronger conservation of the core regions and divergence in the shell and cloud sequences (Figure S1 ). Collectively, these results indicate that the species-specific cloud genomic variation is the main driver of diversity in Poales while the core sequences represent deep and conserved sequences across familes and species. To further explore the conserved fraction among the Poales , a consensus chloroplast genome was assembled by traversing through nodes that are present across at least 80% of species (Fig. 2 b). The obtained consensus genome length was 77,746bp and provides a unified reference for comparative and evolutionary analyses across orders. The consensus genome containes 78 coding sequences (CDS), and 3 rRNAs. Functional annotation highlighted five genes that correspond to the Photosystem-I, 13 to Photosystem-II, 5 to the Cytochrome complex, 6 to ATP synthase, 4 are plastid-encoded RNA polymerase, and 29 correspond to ribosomal proteins (Table S4). The retention of 80% support for a node to be included in the consensus reflects that the cross-families core fraction of the chloroplast is dedicated to energy conversion, transcription, and translation. GO enrichment analysis highlighted energy conversion as the main category among consensus sequences, while non-consensus genes are associated mainly with auxiliary and regulatory functions including cytochrome complex assembly, chlorophyll binding, and light-response regulation (Fig. 2 c). The chloroplast evograph genome reframes the evolutionary history of Poales To further explore the phylogenetic relationship within families in the Poales , we curated a reduced dataset of 192 chloroplast genomes spanning all major tribes. This dataset better balances the broad phylogenetic coverage while reducing overrepresentation of species-rich subfamilies (e.g. Poaceae ), thus one representative species was selected from each tribe (Table S3). This representative dataset was used for evograph genome construction, consensus genome derivation, and graph-based phylogenetic analyses (Fig. 3 a). To investigate phylogenetic relationships among tribes and species, the evograph was transformed into a node presence/absence variation (PAV) matrix where each row represents a path (haplotype genome) and columes are nodes (topologicaly sorted) storing a binary value for presence and absence (0,1). To visualize the large PAV matrix across species, we grouped every 500 consecutive graph nodes into contiguous bins and calculated a bin-level PAV ratio (0–1) representing the proportion of nodes within each bin that were traversed by a given genome path. The obtained PAV binned matrix captures all levels of genomic variation, from SNPs to large SVs and was used for clusterization and phylogenetic analyses (Fig. 3 b). Clusterization was conducted with the t-SNE projection and highlighted a clear family-level divergence which is consistent with the known phylogenetic relationships within Poales (Fig. 3 d). The Poaceae family forms the most extensive and internally structured group, reflecting both its high tribal and species diversity and its complex pattern of chloroplast genome rearrangements. Within the Poaceae , specific tribes and subfamilies clusters were distinguished (e.g., Triticeae , Oryzeae , Bambusoideae , Andropogoneae ) indicating that the graph features capture the overall divergence between specific chloroplast lineages. The Bromeliaceae family was clustered separatly from other families, thus supporting its early divergence from other grasses lineages. In contrast, smaller families like Cyperaceae , Typhaceae , and Eriocaulaceae , were grouped together, reflecting shared evolutionary history in the chloroplast structure and sequence. These clustering patterns recapitulate the early crown-node divergence within Poales and highlight how PAV from graph features can retain meaningful evolutionary signals even across deep phylogenetic splits. Next, we compared the phylogenetic topology constructed from 111 universal genes, including tRNAs and rRNAs, to a topology ontained from the evograph (Figure S2 , Table S5). Comparison of traditional gene-based phylogeny and graph-derived topology revealed broad agreement at the family level with significant discrepancies within Poaceae , particularly among Panicoid taxa (e.g. Eriachne tenuiculmis , Eriachne sp.). The extent of topological displacement was calculated based on the absolute value of difference in the taxon leaf position in each tree and normalized based on the maximum difference found (ranges between 0 and 1). While the graph-based tree preserved the expected tribal topology the traditional phylogenetic tree blurred boundaries between Eriachneae and Isachneae , suggesting incomplete lineage sorting or introgression in early Panicoid evolution. Among species, Teisher 58, and Coelachne africana showed the strongest topological displacement (0.95–1.0), and moderate displacement signals were observed also among the hybrid-prone Andropogoneae species (e.g., Heteropogon , Hyparrhenia , Diheteropogon , Anatherum , Schizachyrium 0.45–0.60). Lower displacement was observed in Tragus mongolorum , Coix lacryma-jobi , Cyperus mutica , and Ananas comosus (0.25–0.35), further supporting recurrent reticulation across Poales (Fig. 3 d, Figure S12-13). These patterns were consistently reproduced in both SplitsTree and tanglegram displacement analyses (Fig. 3 c, Figure S3-4), demonstrating that the graph-based topology captures genuine reticulate evolutionary signals which is frequently obscured in traditional gene-based phylogenetic trees. To reconcile discrepancies between gene-based and evograph-derived phylogenies in Poales , we analyzed node-level features of the evograph beyond gene PAV, revealing that long, lineage-specific nodes strongly influence phylogenetic displacement by linking taxa through shared extended sequence homology. Across Poales families, node length distributions were highly right-skewed, with the majority of nodes shorter than 50bp (Figure S5). Families differed markedly in the relative abundance and contribution of longer nodes where Poaceae showed pervasive enrichment of short nodes, consistent with dense sequence turnover, whereas Joinvilleaceae and Eriocaulaceae displayed pronounced tail of long nodes, indicating retention of extended homologous sequence blocks. These long nodes, which vary in length distribution across families, amplify topological displacement in lineages such as Panicoideae and capture structural evolutionary signals that remain obscured in conventional gene-based trees. Comparison of evograph genomes at the family level To investigate divergence and connectivity at the genus level evographs, five representative Poales genera ( Avena , Oryza , Triticum , Elymus , and Bambusa ) were selected. We impelmented graph-derived metrics including graph density and cloud-node ratio (cloud/total nodes) to capture genomic variation within each genus. These two metrics provide complementary perspectives on the graph topology, where the cloud-node ratio reflects the proportion of lineage-specific polymorphism, and the graph density quantifies the extent of shared polymorphism and structural overlap among species. For each genus, ten species were randomly sampled for evograph construction over 30 independent iterations to ensure a balanced representation and reproducibility of the statistics (Fig. 4 , Table S6). The distribution of cloud-node ratios revealed significant differences in the level of divergence between genera. Elymus showed the highest divergence among species (median ratio = 0.28), while Triticum consistently exhibited low divergence ratios (median = 0.11), reflecting a highly homogeneous chloroplast genome. Oryza and Bambusa had intermediate levels of divergence (median = 0.18 and 0.17, respectively). Graph density further differentiated the connectivity patterns underlying these genera. Bambusa and Triticum displayed the highest graph density (median = 0.0019 and 0.0015, respectively), suggesting highly continuous conserved graph structure except for few edges of recurrent polymorphisms. In contrast, Elymus and Avena exhibited the lowest graph densities (media = 0.0006 and 0.0008, respectively), indicating a fragmented topology with few short conserved sequences. The Oryza had both low divergence and low graph density (0.0003) indicating a conserved topology which is dominated by a single main path with minimal branching or reticulate complexity (tree-like structure). This pattern is characteristic of groups with highly conserved genomes that reflects minimal structural complexity and polymorphism among species. Elymus represents a highly divergent and weakly connected genus, where variation is dominated by lineage-specific modifications rather than shared polymorphism. Triticum shows the opposite pattern, with strong graph connectivity and limited divergence, suggesting fine-scale shared variation within a conserved genome structure. Interestingly, Bambusa combines moderate divergence with high connectivity, reflecting extensive internal polymorphism, while Avena shows more heterogeneous outcomes with extensive differences between iterations. Discussion Advances in long-read sequencing have enabled the assembly of complete genomes and their integration into graph-based pangenomes derived from multiple individuals. This approach overcomes reference bias and uncovers previously hidden structural and non-reference variation across diverse taxa (Sherman and Salzberg 2020 ; Hübner 2022 ; Garrison et al. 2024 ). Despite their potential, graph pangenomes pose major computational challenges, as their complex networks of thousands to millions of interconnected nodes create dense, “hairball-like” structures that hinder intuitive visualization and complicate interpretation. Consequently, their application in evolutionary studies remains limited. We developed the MineGraph pipeline to address key challenges in graph construction, interpretability, and visualization by optimizing graph quality and sensitivity. The embedded optimization procedure provides a more biologically accurate and functionally meaningful representation of genetic diversity (Fig. 1 , Figures S1 –S6). The increased detection of shared nodes, core genome components, and polymorphisms demonstrates its suitability for both mitochondrial and chloroplast genomes (Fig. 1 ), with potential extension to more complex nuclear graph genomes. The emergence of super-pangenomes graph-based genomic references built at the genus or higher taxonomic level has extended the pangenome concept to broader evolutionary scales. The current term “super-pangenome” only partially captures this framework, therefore we propose the term “evograph” to describe a graph-genome structure designed to explore evolutionary relationships. Evographs help reveal functional accessory regions and agronomically important traits, as was recently demonstrated in Oryza, Vitis and Hordeum (Guo et al., 2025 ; Feng et al. 2025 ; Shang et al., 2022). We investigated evolutionary patterns in the order Poales using graph-based evolutionary pangenomes framework (evographs) that revealed structural, reticulate, and lineage-specific relationships obscured in a traditional gene-based phylogenetic tree (Figs. 3 – 4 , S6–S12). We also generated the first unified Poales consensus chloroplast genome (77,746 bp), which highlights a conserved core genes associated with energy and translation and variable dispensable genes at the order level enriched for regulatory functions (Fig. 2 b–c). Graph-to-matrix conversion and t‑SNE clustering clearly delineated family divergence, confirming early separation of Bromeliaceae and highlighting lineage-specific dynamics. Dispensable “Cloud” variation seems to have shaped most of the intraspecific diversity, while conserved “core” regions formed a functional backbone focused on energy metabolism and protein synthesis. Analysis of 709 chloroplast haplotypes resolved long-standing ambiguities, such as the blurred boundaries between Eriachneae and Isachneae within Poaceae , and uncovered strong reticulation in hybrid-prone Andropogoneae taxa like Coix and Cyperus . The evograph framework showed that species-specific “cloud” sequences drive most of the genomic diversity across families despite the conserved core chloroplast structure. Node length distributions across Poales were right-skewed, with most nodes shorter than 50bp, reflecting microvariation and fine-scale sequence fragmentation (Figure S5). However, long lineage-specific nodes are the major determinators of the phylogenetic signal, explaining much of the displacement between trees and graphs (Figures S3-4). Families differed in the prevalence of node lengths, where Poaceae was dominated by short nodes, Bromeliaceae and Typhaceae are depicted with broader distributions, and Joinvilleaceae and Eriocaulaceae retained long homologous sequences. Taxa showing strong topological shifts in the tanglegram and SplitsTree analyses were tied to lineage-specific long nodes rather than short shared ones. In Panicoideae , a taxa lacking plastid genes such as accD, ycf15, psaM, and ycf94, species were also enriched in long nodes, amplifying their separation in the evograph topology. At the genus level, we quantified inter-genera divergence using the cloud-nodes ratio index revealing that Elymus is the most divergent, and Triticum is the most homogeneous and conserved among the tested genera (Fig. 4 ). Moreover, the graph density enabled to highlight structural complexity of the graph indicating that Oryza genus has the simplest graph structure and Bambusa is the most complex. The concept of the evograph represents a shift from discrete reference genome structures to complex, interconnected models that capture genomic diversity across evolutionary timescales and taxonomic breadth. The continued development of graph-based metrics and algorithms capable of capturing genetic diversity, selective pressures, phylogenetic constraints, reticulate evolutionary events, and molecular clock dynamics will be crucial for establishing a comprehensive framework for evolutionary inference. Extending evograph principles to large, complex nuclear genomes will further necessitate advances in algorithms that can effectively handle highly repetitive and polyploid assemblies. Scalable visualization and browsing tools are also essential to enable intuitive exploration of large evographs and to deepen biological interpretability. Ultimately, evographs hold the potential to transform evolutionary and applied genomics by providing a multidimensional representation of genomic diversity that informs genomic selection, genome editing, and conservation strategies. Collectively, these directions position evographs as a powerful and unifying paradigm for studying genome evolution across the tree of life. Materials and Methods Evograph construction and optimization Exploring genomic diversity across multiple assemblies using a graph-based genome representation requires an efficient and reproducible framework. We developed a pipeline called MineGraph to provide a standardized and fully automated approach for constructing and optimizing evographs from multiple genomes. The pipeline is fully integrated into a Docker container to ensure compatibility and smooth adjustment to computing server environment ( https://github.com/hubner-lab/MineGraph ). The evograph construction proceedure starts with a renaming step following the PanSN specifications (Abel et al., 2020 ) to incorporate data of diferent sources and standartisize the different steps of the pipeline using the same samples nomenculture. After renaming, the assemblies fasta files from all samples are indexed using samtools faidx (Li et al., 2009 ), compressed with bgzip and integrated into a graph structure with the Pan-Genome Graph Builder (Garrison et al. 2024 ). The evograph construction is strongly affected by the alignment strategy and particularly the segment length and minimum mapping identify that are passed to the wfmash aligner (Guarracino, 2022). These parameters control the sensitivity and specificity of the mapping steps, thus determining the quality of the alignment and the integrity of the constructed graph. Conceptually, longer segment lengths and higher identity thresholds increase stringency and are suitable for closely related genomes or repeat-rich regions, while lower values improve alignment sensitivity among diverged sequences. Therefore, these parameters should be set in accordance with the studied system to ensure the graph captures meaningful biological variation without over-fragmenting or over-collapsing homologous genomic regions. We applied a data-driven approach for determining these parameters. The minimum mapping identity is first determined by calculating the maximum divergence among the genomes assemblies that are integrated in the graph. Based on the calculated divergence, the mapping identity (-p parameter in PGGB) is determined as p = (100 - (max_divergence * 100) – 2) . To correct for potential underestimation of sequence divergence due to local varioation, and to improve mapping sensitivity, we recommend using slightly lower mapping identity threshold, thus we subtract 2 from the result. The second parameter, homology segment length between genomes, tend to be confunded by long repeats (e.g. transposable elements, tabdem repeats). To optimize this parameter, all studied organelle genomes are concatenated into a single FASTA file using seqtk (Li, 2012 ) and 3Mbp sequence is randomly sampled for repetitive elements analysis. Sampled sequences are then analyzed in RepeatMasker v4.0.9 (Smit et al., 2013–2015) with default parameters (-no_is -s) to identify the longest repeat sequence. The segmet length parameter is determined as s = longest_segment * 1.2 to capture flanking sequence beyond the repeat in the homology search. The optimized parameters are saved to params.yaml file and used automatically for the graph construction stage. Evogaph formats and statistics The evograph is constructed in a graphical fragment assembly (GFA) format which describes graph components, namely the sequence segment (nodes), links (edges), and paths (sub-graphs). To increase functionality, we implemented automatic convertion to MAF/PAF alignments formats and variant calls in VCF and vg formats. Additionally, the graph is converted to a multiple sequence alignment (MSA) file and processed using RAxML v. 1.2.2 (Kozlov, 2019) to build a phylogenetic tree based on the evolutionary relationships captured in the graph. To infer biological insights directly from the evograph we apply a statistical workflow and extract key features and metrics. The graph is comprised of nodes and edges which are counted and sorted based on their frequency among samples in the graph. Nodes and edges are divided into three pangenomic fractions based on their frequency across haplotypes. Core nodes are the conserved sequences across samples and are defined by default as present among more than 95% of samples. Cloud nodes capture the rare fraction of sequences among sasmples and is defined by default as 5%. Shell nodes capture sequences that are shared among less than 95% of samples and in more than 5%. These parameters can be adjusted by the user although it is recommended to maintain a standard definition of the pangenome to enable comparison between datasets and studies (Glick and Mayrose 2023 ). To capture the sparsity and the rate of polymorphism in the graph, we calculate the average node degree which is defined as the average number of links (edges) between neighboring nodes (sequences): n edges /n nodes . In addition, the overall graph density is calculated from the ratio of observed versus theoretical edges in the graph: |E| / \(\:\left(\genfrac{}{}{0pt}{}{\left|\text{N}\right|\:}{2}\right)\) , where |E| is the absolute number of edges (regardless their direction) and |N| is the number of nodes. The graph density depicts the expected saturation of diversity in the graph, namely how much of the expected diversity of the pangenome is actually represented in the graph. To provide a standard population genetics metrics for comparison with other systems, we also calculate the average nucleotide diversity ( \(\:\pi\:\) ) based on biallelic SNPs called in the graph. Representation of polymorphism and consensus Genetic variants are called directly from the graph including single nucleotide polymorphisms (SNPs), small insertions and deletions (indels), and larger structural variations (SVs) including inversions, duplications, and translocations. These variants are summarized in a variant call format (VCF) to ensure compatibility with a broad range of external genomic analysis tools. Polymorphic sites are also converted to a presence/absence variation (PAV) matrix where the occurrence of each sequence (node) is indicated for each individual. The PAV matrix is then used to calculate distances between samples (paths) using Hamming distance or parsimony score and a phylogenetic tree is constructed. To generate a consensus sequence among samples, all nodes passing a minimum presence frequency threshold (configurable up to 100%) are concatenated into a contiuous path. The consensus sequence path is saved as a FASTA file and represents the shared path in the graph. To study the functional aspect of the obtained consensus genome across chloroplasts in the Poales , we annotated the sequence using InterProScan v5 (Jones et al., 2014 ) and assigned the protein domains and ontologies terms (GO) based on the coding sequences. GO enrichment analysis was carried with the GOATOOLS v1 (Klopfenstein et al., 2018) and topGO v2 (Alexa & Rahnenfuhrer, 2023) packages in R using the go-basic.obo ontology (release July 2025) for any subset of sequences. Fisher’s exact test was applied, and multiple testing was corrected using the Benjamini–Hochberg false discovery rate (FDR). Visualization and Interpretation We provide a variety of visual outputs to support intuitive exploration and interpretation of the generated evograph genome. The final GFA graph is converted into vg and FASTA formats, allowing compatibility with external visualization tools like SequenceTubeMap (vgteam, 2019 ). In addition, a dynamic HTML-based visualization is implemented using the Pyvis module (Pyvis developers, 2023) to interactively display nodes properties, connectivity, graph density, and the consensus subgraph. Other descriptive histograms and plots are generated automatically along the processing steps of the graph to facilitate further biological interpretation. To quantify local genomic diversity along the graph we implemented a sliding window quantification approach using a k-mer–based analysis (Moin & Seemann, 2023). Each genome path in the graph was divided into consecutive k-mer windows (default k = 1000), and the node metrics including degree, coverage, and length are aggregated into a single score. Low-frequency or structurally irregular windows in the graph indicate regions harboring private alleles, copy number variation, or a structural variant, whereas high-frequency windows indicate a conserved core segments in the graph. This unified framework captures both sequence-level and structural diversity across the pangenome. Poales genomic data To evaluate the performance of the pipeline, 17 representative mitochondrial and chloroplast genomes were randomly selected to construct evographs (Table S1 ). Genomic data were retrieved from the NCBI Nucleotide (GenBank) database for the Poales order. To study chloroplast diversity among the Poales , we retrieved all 1,541 complete circular chloroplast genomes available in NCBI. After filtering redundancy at the species level, 709 unique species remained (Table S2 ). To explore diversity among chloroplast genomes at the species, tribe and family level, presence and absence profiles of nodes in the evograph were analyzed. A principal component analysis (PCA) was performed with the PAV data using the package FactoMineR v2.0 (Lê et al., 2008 ). Samples were then clustered using the nonlinear dimensionality reduction methods t-SNE and UMAP as implemented in the Rtsne and uwot packages in R (Krijthe, 2015 , Melville, 2020 ). Hierarchical clustering was performed using Ward’s method on pairwise species distances, with 100 bootstraps to assess node support. Dendrograms were converted to phylo objects using ape v5 (Paradis & Schliep, 2019 ) for visualization. To compare graph-based and sequence-based phylogenies, RAxML trees were midpoint-rooted with phytools v1 (Revell, 2012 ) and compared using tanglegrams as implemented in R package dendextend v1.1 (Galili, 2015 ). Entanglement scores were quantified to evaluate topological agreement: for each taxon i , we define a displacement score as the absolute difference between its vertical positions in the two trees D i = | p i ⁽¹⁾ − p i ⁽²⁾ |, where p i ⁽¹⁾ is the position of taxon i in Tree 1, p i ⁽²⁾ is the position of taxon i in Tree 2, and D i is the displacement score. Because absolute displacement values depend on the total number of taxa and the layout of the tanglegram, we further computed a normalized displacement D̃ i = D i / maxⱼ Dⱼ where D̃ i is the normalized displacement (range 0–1) and maxⱼ Dⱼ is the maximum displacement observed across all taxa. species labels were harmonized across datasets using Jaro–Winkler string-distance matching (Winkler et al., 1990). Gene presence–absence variation (PAV) analysis Gene-level presence–absence variation (PAV) was quantified using a curated set of chloroplast genes annotated across all assemblies. For each taxon, gene presence was recorded as a binary state (1 = present, 0 = absent), generating a gene-by-taxon PAV matrix. Genes absent from all taxa or present in all taxa were excluded from downstream comparative analyses, as they do not contribute discriminatory phylogenetic signal. To explore the phylogenetic relationship among tribes and species, the graph was converted into a node PAV matrix using the ODGI path -H command (Guarracino et al., 2022 ). This matrix was used as the input for graph-based similarity analyses, enabling direct comparison between classical gene-content–based phylogenies and evograph-derived topologies. Because these graphs tend to scale to hunderds of thousands, millions, or more nodes, visualizing individual nodes is neither practical nor informatiove. Therfore, using the ODGI sorting order of the nodes, we clustered N nodes (defaul 500) to a bin representing a contiguous segemnt of the graph. Thus, the total number of bins = #nodes/N, for each path p and each bin b , and the bin-level PAV ratio is defined as: $$\:\text{P}\text{A}{V}_{p,b}=\frac{\text{n}umber\:of\:nodes\:in\:bin\:\:visited\:by\:path\:p}{total\:number\:of\:nodes\:in\:bin\:b}$$ This ratio quantifies the fraction of nodes within a given graph region that are traversed by a specific path and take values in the range [0,1]. A value of 0 indicates that the path is completely absent from the corresponding graph region, whereas a value of 1 indicates that the path traverses all nodes within that region. Intermediate values reflect partial traversal and capture regional structural divergence across paths. The overall ‘presence’ scores for an individual across all segments corresponds to its path in the graph which is equivalent to the full genome sequence. (Fig. 3 b, Figure S7), excluding non-polymorphic nodes. Graph node length computation and filtering Graph node lengths were computed from the final evograph by measuring the nucleotide sequence length associated with each node after graph construction and normalization. Node length was defined as the number of base pairs represented by the node, independent of path multiplicity. Nodes of length 1 bp typically arising from alignment boundaries or residual graph fragmentation were excluded from all analyses. To reduce the influence of short nodes while retaining biologically meaningful sequence blocks, nodes were demultiplexed by retaining only unique nodes with a minimum length of 50bp. This threshold was chosen because nodes shorter than 50bp are overwhelmingly abundant, widely shared across taxa, and primarily reflect local micro-variation or alignment noise, whereas longer nodes are more likely to represent extended homologous sequence segments. Node length distributions were summarized separately for each family using logarithmically scaled bins. Declarations Author Contribution SH conceived and supervised the research project. RH and SH planned and designed the research project. RH performed the bioinformatics and data analyses. SH and RH wrote the manuscript. Data Availability The MineGraph package is implemented with Python and is freely available at the Hübner lab Github repository: https://github.com/hubner-lab/MineGraph References Abel, H. J., Larson, D. E., Regier, A. A., Chiang, C., Das, I., Kanchi, K. L., et al. (2020). Mapping and characterization of structural variation in 17,795 human genomes. Nature, 583, 83–89. https://doi.org/10.1038/s41586-020-2371-0. Alexa, A., & Rahnenführer, J. (2023). topGO: Enrichment analysis for Gene Ontology. R package version 2.50.0. Alkan, C., Sajjadian, S., & Eichler, E. E. (2011). Limitations of next-generation genome sequence assembly. Nature Methods, 8, 61–65. https://doi.org/10.1038/nmeth.1527. Christin, P.-A., Spriggs, E., Osborne, C. P., Strömberg, C. A. E., Salamin, N., & Edwards, E. J. (2014). Molecular dating, evolutionary rates, and the age of the grasses. Systematic Biology, 63, 153–165. https://doi.org/10.1093/sysbio/syt072. Eizenga, J. M., Novak, A. M., Sibbesen, J. A., et al. (2020). Pangenome graphs. Annual Review of Genomics and Human Genetics, 21, 139–162. https://doi.org/10.1146/annurev-genom-120219-080406. Feng, J. W., Pidon, H., Cuacos, M., Lux, T., Himmelbach, A., Haghi, R., ... & Mascher, M. (2025). A haplotype-resolved pangenome of the barley wild relative Hordeum bulbosum. Nature, 645(8080), 429-438. Galili, T. (2015). dendextend: An R package for visualizing, adjusting, and comparing trees of hierarchical clustering. Bioinformatics, 31(22), 3718–3720. https://doi.org/10.1093/bioinformatics/btv428 Garrison, E., Sirén, J., Novak, A. M., Eizenga, J. M., Dawson, E. T., Jones, W., Garg, S., Markello, C., Paten, B., & Durbin, R. (2018). Variation graph toolkit improves read mapping by representing genetic variation in the reference. Nature Biotechnology, 36(9), 875–879. https://doi.org/10.1038/nbt.4227. Garrison, E., Guarracino, A., Heumos, S., et al. (2024). Variation graph pangenomes. Nature Methods, 21, 1–10. https://doi.org/10.1038/s41592-023-02180-8 Garrison, E., Guarracino, A., Heumos, S., Villani, F., Bao, Z., Tattini, L., ... & Prins, P. (2024). Building pangenome graphs. Nature Methods, 21(11), 2008-2012. Glick, L., & Mayrose, I. (2023). Partitioning and comparative analysis of pangenomes. Bioinformatics, 39(5), btad245. https://doi.org/10.1093/bioinformatics/btad245. Guarracino, A., Heumos, S., & Garrison, E. (2022). ODGI: Understanding pangenome graphs. Bioinformatics, 38(13), 3315–3317. https://doi.org/10.1093/bioinformatics/btac308. Guo, L., Wang, X., Ayhan, D. H., Rhaman, M. S., Yan, M., Jiang, J., ... & Ye, W. (2025). Super pangenome of Vitis empowers identification of downy mildew resistance genes for grapevine improvement. Nature Genetics, 1-13. Hubley, R., Finn, R. D., Clements, J., et al. (2016). The Dfam database of repetitive DNA families. Nucleic Acids Research, 44(D1), D81–D89. https://doi.org/10.1093/nar/gkv1272. Hübner, S. (2022). Are we there yet? Driving the road to evolutionary graph-pangenomics. Current Opinion in Plant Biology, 66, 102195. https://doi.org/10.1016/j.pbi.2022.102195. Jayakodi, M., et al. (2020). The barley pan-genome reveals the hidden legacy of mutation breeding. Nature, 588, 284–289. https://doi.org/10.1038/s41586-020-2947-8. Jones, P., et al. (2014). InterProScan 5: Genome-scale protein function classification. Bioinformatics, 30(9), 1236–1240. https://doi.org/10.1093/bioinformatics/btu031. Kellogg, E. A. (2015). Flowering Plants. Monocots. Poaceae. In K. Kubitzki (Ed.), The Families and Genera of Vascular Plants, Vol. XIII (pp. 1–416). Springer International Publishing. ISBN 978-3-319-15332-2. Klopfenstein, D. V., et al. (2018). GOATOOLS: A Python library for Gene Ontology analyses. Scientific Reports, 8, 10872. https://doi.org/10.1038/s41598-018-28948-z. Kozlov, A. M., Darriba, D., Flouri, T., Morel, B., & Stamatakis, A. (2019). RAxML-NG: A fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference. Bioinformatics, 35(21), 4453–4455. https://doi.org/10.1093/bioinformatics/btz305. Krijthe, J. H. (2015). Rtsne: T-Distributed Stochastic Neighbor Embedding using Barnes-Hut implementation. Li, H. (2012). seqtk: Toolkit for processing sequences in FASTA/Q formats. Li, H., et al. (2009). The Sequence Alignment/Map format and SAMtools. Bioinformatics, 25(16), 2078–2079. https://doi.org/10.1093/bioinformatics/btp352. Linder, H. P., & Rudall, P. J. (2005). Evolutionary history of Poales. Annual Review of Ecology, Evolution, and Systematics, 36, 107–124. https://doi.org/10.1146/annurev.ecolsys.36.102403.135635. Liu, Y., Du, H., Li, P., Shen, Y., Peng, H., Liu, Z., Shi, M., Zhou, G.-A., Zhang, H., Li, Y., Zhang, W., Wang, Z., Zhu, B., Han, B., Liang, C., & Tian, Z. (2020). Pan-Genome of Wild and Cultivated Soybeans. Cell, 182(1), 1–15. https://doi.org/10.1016/j.cell.2020.05.023. Lê, S., Josse, J., & Husson, F. (2008). FactoMineR: An R package for multivariate analysis. Journal of Statistical Software, 25(1), 1–18. Melville, J. (2020). uwot: The Uniform Manifold Approximation and Projection (UMAP) implementation for R. R package. Minh, B. Q., et al. (2020). IQ-TREE 2: New models and efficient methods for phylogenetic inference. Molecular Biology and Evolution, 37(5), 1530–1534. https://doi.org/10.1093/molbev/msaa015. Moin, M., Seemann, T., et al. (2023). gretl: Graph-based read error correction and variation discovery. Microbiology Resource Announcements, 12(6), e00243-24. https://doi.org/10.1128/mra.00243-24. Minio, A., Massonnet, M., Figueroa-Balderas, R., Castro, A., & Cantu, D. (2019). Diploid genome assembly of the wine grape Carménère. G3: Genes|Genomes|Genetics, 9(5), 1331–1337. https://doi.org/10.1534/g3.119.400030. Montenegro, J. D . , et al. (2017). The pangenome of hexaploid bread wheat . The Plant Journal, 90(5), 1007–1013. https://doi.org/10.1111/tpj.13615. Nurk, S., et al. (2022). The complete sequence of a human genome. Science, 376, 44–53. https://doi.org/10.1126/science.abj6987. Paradis, E., & Schliep, K. (2019). ape 5.0: An environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics, 35(3), 526–528. https://doi.org/10.1093/bioinformatics/bty633. Paten, B., et al. (2017). Genome graphs and the evolution of genome inference. Genome Research, 27, 665–676. https://doi.org/10.1101/gr.214155.116. Potapenko, E. V., Schwarts, D., Mandel, T., Ashkenazy, N., Fuerst, D., Atsmon, G., ... & Hübner, S. (2024). Not so selfish: transposable elements fuel adaptation to climate change. bioRxiv , 2024-09. Price, M. N., Dehal, P. S., & Arkin, A. P. (2010). FastTree 2 – Approximately maximum-likelihood trees for large alignments. PLOS ONE, 5(3), e9490. https://doi.org/10.1371/journal.pone.0009490. Revell, L. J. (2012). phytools: An R package for phylogenetic comparative biology. Methods in Ecology and Evolution, 3, 217–223. Saarela, J. M., Burke, S. V., Wysocki, W. P., Barrett, M. D., Clark, L. G., Craine, J. M., Peterson, P. M., Soreng, R. J., Vorontsova, M. S., & Duvall, M. R. (2018). A 250 plastome phylogeny of the grass family (Poaceae): Topological support under different data partitions. PeerJ, 6, e4299. https://doi.org/10.7717/peerj.4299. Sayers, E. W., Ciufo, S., Baker, J., O’Leary, N. A., Klimke, W., McVeigh, R., Pruitt, K. D., & Karsch-Mizrachi, I. (2023).Strategies for resolving taxonomic name inconsistencies in large comparative genomics datasets.Bioinformatics Advances, 3(1), vbad040. https://doi.org/10.1093/bioadv/vbad040. Shang, L., Li, X., He, H., Yuan, Q., Song, Y., Wei, Z., ... & Qian, Q. (2022). A super pan-genomic landscape of rice. Cell Research, 32(10), 878-896. Sherman, R. M., & Salzberg, S. L. (2020). Pan-genomics in the human genome era. Nature Reviews Genetics, 21(4), 243–254. https://doi.org/10.1038/s41576-020-0210-7. Simão, F. A., et al. (2015). BUSCO: Assessing genome assembly and annotation completeness. Bioinformatics, 31(19), 3210–3212. Smit, A. F. A., Hubley, R., & Green, P. (2013–2015). RepeatMasker Open-4.0. http://www.repeatmasker.org. Tettelin, H., et al. (2005). Genome analysis of multiple pathogenic isolates of Streptococcus agalactiae: Implications for the microbial pan-genome. Proceedings of the National Academy of Sciences USA, 102(39), 13950–13955. https://doi.org/10.1073/pnas.0506758102. The Angiosperm Phylogeny Group (APG IV). (2016). An update of the Angiosperm phylogeny group classification for the orders and families of flowering plants. Botanical Journal of the Linnean Society, 181, 1–20. https://doi.org/10.1111/boj.12385. Walkowiak, S., Gao, L., Monat, C., Haberer, G., Kassa, M. T., Brinton, J., Ramirez-Gonzalez, R. H., Kolodziej, M. C., Delorean, E., Thambugala, D., et al. (2020). Multiple wheat genomes reveal global variation in modern breeding. Nature, 588(7837), 277–283. https://doi.org/10.1038/s41586-020-2961-x. Wenger, A. M., et al. (2019). Accurate circular consensus long-read sequencing improves variant detection. Nature Biotechnology, 37, 1155–1162. https://doi.org/10.1038/s41587-019-0217-9. Winkler, W. E. (1990).String Comparator Metrics and Enhanced Decision Rules in the Fellegi–Sunter Model of Record Linkage.Proceedings of the Section on Survey Research Methods, American Statistical Association, pp. 354–359. Zhao, Q. , Feng, Q. , Lu, H. , Li, Y. , Wang, A. , Tian, Q. , Zhan, Q. , Lu, Y. , Zhang, L. , & Huang, T. (2018). Pan-genome analysis highlights the extent of genomic variation in cultivated and wild rice. Nature Genetics, 50 (2), 278–284. https://doi.org/10.1038/s41588-018-0041-z. vgteam. (2019). SequenceTubeMap. GitHub repository. https://github.com/vgteam/sequenceTubeMap. Additional Declarations No competing interests reported. Supplementary Files SupplementaryInformation.docx Supplementarytable.xlsx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 01 Apr, 2026 Reviews received at journal 24 Mar, 2026 Reviews received at journal 23 Mar, 2026 Reviewers agreed at journal 10 Mar, 2026 Reviewers agreed at journal 10 Mar, 2026 Reviewers invited by journal 06 Mar, 2026 Editor assigned by journal 11 Feb, 2026 Submission checks completed at journal 05 Feb, 2026 First submitted to journal 04 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8786947","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":604033945,"identity":"ab70e791-d168-4680-a9ed-a83c9fd4de73","order_by":0,"name":"Rakan Haib","email":"","orcid":"","institution":"Tel Hai University of Kiryat Shmona","correspondingAuthor":false,"prefix":"","firstName":"Rakan","middleName":"","lastName":"Haib","suffix":""},{"id":604033946,"identity":"7f1b490e-a03e-48df-8a85-057faa086f72","order_by":1,"name":"Sariel Hübner","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6klEQVRIiWNgGAWjYBACAwYeKIu9gYHhYQOUm1BAjBaewwwMiXAtBsRokUgGa4GL4wTmErnHHnzcUydvPvP9wQeJO+xkGNgPP2B4gEeL5Yy8dMMZzw4bzrmdzGyQeCaZh4EnzQC/w27kmEnzHDjAOEM6mU0isY0Z6M4cAn4BaflzoM5+huRh9h+JbfU8DPxviNDCcIA5cYYEMxtDYtthHgYJQraceZcm2XPgcPIMnmRjoMOO87BJPDM4gFfL8dxjEj8O1NnOYD/48MPHtmp7fv7khw9/VODWggnYgPgAKRpGwSgYBaNgFGACAGyITDskgZM0AAAAAElFTkSuQmCC","orcid":"","institution":"Tel Hai University of Kiryat Shmona","correspondingAuthor":true,"prefix":"","firstName":"Sariel","middleName":"","lastName":"Hübner","suffix":""}],"badges":[],"createdAt":"2026-02-04 13:10:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8786947/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8786947/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104442140,"identity":"389f4894-e59c-40b9-aafe-3d65db69f4ed","added_by":"auto","created_at":"2026-03-11 18:42:56","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":315571,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOptimized evograph genome construction for mitochondria and chloroplasts\u003c/strong\u003e. a) Comparison of graph metrics obtained for the optimized (green triangle) and the default (blue dot) evograph genomes. b) Homology search in sliding windows along the graph, where yellow bars indicate the most frequent segments among species and violet bars represent unique sequences for the optimized evograph (upper matrix) and the default evograph (lower matrix). c) Representation of the pangenomic fractions in the optimized (upper plot) and default evograph (bottom plot). Core sequences (nodes) that are found among more than 95% of the species included in the graph are marked in green, cloud nodes that represent sequences found in less than 5% of species are indicated in blue and shell nodes found in intermediate frequencies are marked in orange.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/92d29bbf0169003ec81ac32a.png"},{"id":104442137,"identity":"e5277632-da8d-4d87-89b9-092a15022b51","added_by":"auto","created_at":"2026-03-11 18:42:56","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":263993,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003ePloales\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003echloroplast evograph across 709 chloroplast haplotypes\u003c/strong\u003e. \u003cstrong\u003ea)\u003c/strong\u003e Histogram showing the frequency of nodes according to their presence across haplotypes, partitioning the graph into core nodes (present in 95% haplotypes; green), shell nodes (present in \u0026gt; 5% but fewer than 95 % of haplotypes; orange), and cloud nodes (restricted to less than 5% of the haplotypes; blue). The predominance of nodes highlights strong overall conservation of chloroplast genome structure, with a smaller fraction of cloud sequences contributing to lineage-specific diversity. \u003cstrong\u003eb)\u003c/strong\u003eConsensus sequence annotation derived from the graph, constructed by retaining nodes supported in ≥80% of paths. A total of 81 genes were annotated, spanning photosynthetic and plastid housekeeping functions. \u003cstrong\u003eb)\u003c/strong\u003e Gene Ontology (GO) terms of molecular function for consensus genes (upper plot) versus non-consensus gene set based on gene counts (x axis), Dot position along the x-axis indicates the number of study genes annotated with each GO term, and color represents enrichment significance (−log₁₀ p-value).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/810f15d70646b9e863ca0b86.png"},{"id":104442138,"identity":"3ea1f059-7f24-4637-bf40-6840a4845094","added_by":"auto","created_at":"2026-03-11 18:42:56","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":759077,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTribe-level EvoGraph of 192 Poales chloroplast genomes. a) \u003c/strong\u003eEvograph of 192 chloroplast genomes with each path representing a single tribe-level representative (Supplementary Table S5). Colored paths correspond to species-specific sequences, while nodes are represented as bars along the graph backbone. \u003cstrong\u003eb) \u003c/strong\u003eExcerpt of the graph-derived presence/absence variation (PAV) heatmap for a subgraph region, showing binary node retention across species. Red indicates node presence and white indicates absence, highlighting conserved versus variable graph regions \u003cstrong\u003ec)\u003c/strong\u003e SplitsTree tanglegram comparing the graph-derived distance tree with the SNP-based RAxML phylogeny. Tip labels are colored by family using the same palette as in panel (d), with branches shown in black and connector lines indicating taxon correspondence between trees. The highlighted region marks a subset of taxa exhibiting the strongest topological disagreement, characterized by extensive crossing of connector lines and large positional shifts between graph- and SNP-based trees. \u003cstrong\u003ed)\u003c/strong\u003e t-SNE embedding of graph-derived PAV profiles for 192 chloroplast genomes colored by family.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/0eff98974154c99cc7f6093c.png"},{"id":104442139,"identity":"46998902-8acc-4646-89e7-479fdd3a6e8b","added_by":"auto","created_at":"2026-03-11 18:42:56","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":381471,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGraph-derived metrics of divergence and connectivity across representative genera\u003c/strong\u003e.\u003cbr\u003e\n(a) Focused comparison of two informative could-node ratio (cloud/total nodes) and graph density-across five representative genera (\u003cem\u003eAvena, Oryza, Triticum, Elymus, and Bambusa\u003c/em\u003e). (b-d) Local 10-vs-10 sub-graph views generated for specific genomic region [60,000–70,000]. (b) \u003cem\u003eBambusa\u003c/em\u003e (upper 10 paths) versus \u003cem\u003eOryza\u003c/em\u003e. (c) \u003cem\u003eElymus\u003c/em\u003e (upper 10 paths) versus \u003cem\u003eTriticum\u003c/em\u003e. (d) \u003cem\u003eElymus\u003c/em\u003e (upper 10 paths) versus \u003cem\u003eOryza\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/1756a5e9318d835809026954.png"},{"id":104780270,"identity":"ce265f61-d3ee-4da4-b738-a3d1770e85f1","added_by":"auto","created_at":"2026-03-17 07:51:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2419957,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/3c7bdb3a-b966-41b2-82bb-d46fe00315b5.pdf"},{"id":104442141,"identity":"da35060d-f24e-4a82-aad8-ae63b0ebc93f","added_by":"auto","created_at":"2026-03-11 18:42:56","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":6870123,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/0168386f15f5f4af64e57709.docx"},{"id":104442142,"identity":"d57dd350-a566-4f76-af06-eba314f32d09","added_by":"auto","created_at":"2026-03-11 18:42:56","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":134651,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8786947/v1/beac42c2bcf705a062ea07d4.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Evolutionary graph pangenome of the order Poales from chloroplast genomes highlight phylogenetic inconsistencies","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe rapid maturation of DNA sequencing technologies have fundamentally reshaped how genomic data is generated, represented, and interpreted. The introduction of ultra-high-throughput sequencing enabled population-scale whole-genome sequencing, while subsequent advances in long-read platforms such as PacBio and Oxford Nanopore Technologies (ONT) dramatically improved contiguity, accuracy, and affordability of genome assemblies. Highly accurate long reads (e.g., PacBio HiFi) and ultra-long ONT reads now routinely resolve repetitive and structurally complex regions, enabling haplotype phasing, and support chromosome-scale telomere-to-telomere (T2T) assemblies (Nurk et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Wenger et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). As a result, genome assembly has shifted from producing a single representative sequence to generating multiple high-quality assemblies across individuals, populations, and species, laying the foundation for comparative and evolutionary genomics at unprecedented scale (Montenegro et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Jayakodi et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Liu et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Minio et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Walkowiak et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMost population genomic analyses so far relied on a single linear reference genome although this paradigm inherently fails to capture the full spectrum of genetic diversity within a species. Sequences that differ substantially from the reference genome, particularly around structural variation (SVs) are often misrepresented, poorly aligned, or missed entirely (Alkan et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Sherman \u0026amp; Salzberg, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The introduced reference bias disproportionately favors sequences that are more similar to the chosen reference. This asertainment biased is specifically problematic in highly diverse or structured populations, leading to incomplete or distorted inferences of adaptation, gene flow, and evolutionary history. The availability of multiple high-quality assemblies made clear that no single reference can adequately represent the species sequence diversity, motivating the transition to pangenome frameworks (Tettelin et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Eizenga et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA pangenome represents the complete genomic repertoire of a species or clade, integrating sequences shared across individuals. Pangenomes are commonly partitioned into a core fraction representing genomic segments that are present in all or nearly all individuals, and sequences that are present in only a small (cloud) or moderate (shell) subset of individuals. Graph-based pangenomes further enhance this representation by encoding multiple genomes within a unified data structure, enabling more accurate read mapping, variant calling, and genotyping, particularly in regions affected by complex structural variation (Garrison et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Paten et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Graph pangenomes are typically formalized as sequence or variation graphs where nodes represent conserved DNA sequence fragments with no polymorphism, while edges encode observed adjacencies between nodes. Genetic variation is represented as branching paths or loops, and a full path through the graph corresponds to an observed or inferred haplotype. This structure generalizes the linear reference genome, allowing multiple alternative sequences to coexist in a single coordinated system and providing a natural framework for representing complex variation (Eizenga et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAs genomic sampling has expanded to encompass multiple species, pangenomes have been extended to what are often termed \u0026ldquo;super-pangenomes\u0026rdquo;. These are typically clade or genus-level graphs that integrate interspecific variation and was recently developed for \u003cem\u003eOryza\u003c/em\u003e, \u003cem\u003eVitis\u003c/em\u003e and \u003cem\u003eHordeum\u003c/em\u003e (Guo et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Feng et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Shang et al., 2022). However, the term \u0026ldquo;super-pangenome\u0026rdquo; emphasizes scale and complexity rather than the underlying purpose and can become terminologically confusing as graphs are extended to families or orders. A more appropriate term is evolutionary graph pangenome (or evograph in short), which explicitly reflects the goal of modeling genomic diversity across evolutionary time and taxonomic depth. Such graphs can augment phylogenetics trees to model evolution at the genome scale.\u003c/p\u003e \u003cp\u003e Evographs offer a platform for evolutionary inference directly from the graph topology. For example, the shortest or most conserved path through the graph approximate the core genome, providing insight into indispensable genomic components, while alternative paths encode lineage- or species-specific polymorphism (H\u0026uuml;bner \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Despite their promise, graph pangenomes face substantial challenges. Tool development remains limited, and many existing frameworks, such as the Pan-Genome Graph Builder (PGGB) and related variation-graph pipelines, require careful manual parameter tuning and substantial computational resources (Garrison et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Evographs can contain millions of nodes and edges and are difficult to visualize, query, and interpret posing a major bottleneck. Despite the rich evolutionary information embeded in the graph, extracting biologically meaningful insights and presenting them in an intuitive manner remains challanging. Addressing these challenges including scalable visualization tools, and clearer conceptual frameworks are key for fully realizing the potential of evographs as a unifying paradigm in genomics and evolutionary biology.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe order \u003cem\u003ePoales\u003c/em\u003e is one of the most ecologically dominant clades comprising circa 20,000 species in 16 families and representing over one third of all monocots and 7% of angiosperms worldwide (APG IV, 2016; Linder \u0026amp; Rudall, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). \u003cem\u003ePoales\u003c/em\u003e are also of exceptional economic importance, encompassing major crops including rice, wheat, maize, sugarcane, barley, bamboo and others. Ecologically, \u003cem\u003ePoales\u003c/em\u003e are abundant along terrestrial and aquatic ecosystems, including grasslands, savannas, steppes, and wetlands. Owing to its size, ecological breadth, and functional innovation, \u003cem\u003ePoales\u003c/em\u003e and particularly memebers of the \u003cem\u003ePoaceae\u003c/em\u003e family have become a key model for studying macroevolution, ecological adaptation, and diversification under environmental change (Christin et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Potapenko et al. 2025). The earliest-diverging lineages in the \u003cem\u003ePoales\u003c/em\u003e include \u003cem\u003eBromeliaceae\u003c/em\u003e, \u003cem\u003eTyphaceae\u003c/em\u003e, and \u003cem\u003eRapateaceae\u003c/em\u003e, followed by diversification into two major assemblages: the \u003cem\u003eCyperid\u003c/em\u003e clade and the \u003cem\u003eGraminid\u003c/em\u003e clade, the latter giving rise to \u003cem\u003ePoaceae\u003c/em\u003e and its major BOP/BEP and PACMAD lineages (Kellogg, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Despite decades of study, reconstructing \u003cem\u003ePoales\u003c/em\u003e evolution remains challenging with traditional phylogenetic approaches providing ambiguous support for key relationships, particularly within \u003cem\u003ePoaceae\u003c/em\u003e (Saarela et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Moreover, tree-based models occasionally fail to capture hybridization, introgression, and gene duplication that are pervasive in grasses. Evographs can potentially provide a powerful alternative by representing the full spectrum of genomic diversity, including structural variation and presence/absence polymorphisms, across species and lineages (Eizenga et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Garrison et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWe developed MineGraph, a dedicated pipeline for constructing and analyzing evolutionary graph pangenomes (evographs) from genome assemblies. MineGraph incorporates multiple optimization steps designed to enhance graph quality, reduce complexity, and improve interpretability. We applied this pipeline to more than 700 chloroplast genome assemblies representing most of the order \u003cem\u003ePoales\u003c/em\u003e, revealing patterns of genomic diversity that underlie evolutionary relationships among species, genera, and families within the order.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eOptimizing the evograph-genome construction\u003c/h2\u003e \u003cp\u003eGraph construction from genome assemblies relies on an intensive all-versus-all alignment process, where the parameters critically influence the resulting graph structure. To optimize these parameters, we implemented a data-driven approach in a pipeline called MineGraph and calculated genetic distances among samples, and the minimum segment length based on the longest repeat sequence for homology search. To evaluate this approach, evographs were constructed with PGGB for 17 mitochondrial and 17 chloroplast genomes representing the same set of species within the \u003cem\u003ePoales\u003c/em\u003e order (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). The optimized mitochondrial evograph with MineGraph comprised of 52,953 nodes and 73,670 edges, while the version produced with default parameters in PGGB yielded 10,668 nodes and 14,632 edges, a considerable reduction in polymorphism and connectivity (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). The optimized mitochondrial evograph had much higher edge density and longer homology stretches; thus, fragmentation of shared sequences was reduced in 30% (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb). Calculating the pangenomic proportion of sequence representation in each evograph further highlighted the differences between versions. The evograph constructed with default parameters failed to detect core sequences (nodes) that are shared among more than 95% of the genomes that were included in the graph and identified only sparse shell and cloud nodes (defined by default as represented among less than 5% of the genomes). In contrast, the optimized evograph captured a more balanced representation of the pangenome with 2,912 core nodes, 32,623 shell nodes, and 16,678 cloud nodes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec).\u003c/p\u003e \u003cp\u003eFor the chloroplast genomes which are smaller, structurally simpler, and more conserved than mitochondrial genomes, extensive homology and low polymorphism was observed when using default parameters. Applying the optimized protocol improved the sensitivity of the alignment, thus yielding a total of 3,076 edges (polymorphism) and 2,068 nodes, of which 1908 are core nodes.\u003c/p\u003e \u003cp\u003eTogether, these results demonstrate that parameter adjustment and optimization substantially improve the graph construction enabling the detection of polymorphism across species (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea\u0026ndash;c, Figure S3-4).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eElucidating the\u003c/b\u003e \u003cb\u003ePloales\u003c/b\u003e \u003cb\u003echolorplast evograph genome\u003c/b\u003e\u003c/p\u003e \u003cp\u003eApproximately 1500 chloroplast genomes of \u003cem\u003ePoales\u003c/em\u003e species are available from the NCBI database with clear overrepresentation of agronomic important species. To obtain a non-redundant dataset, we extracted all uniqe 709 chloroplast complete genomes with one representative accession per species (Table S3). A graph was constructed from all 709 chloroplast genomes, yielding a comprehensive evograph comprised of 413,416 nodes and 685,459 edges. The obtained graph is characterized with low sparsity indicated with an average node degree of 1.66 and a low density (0.008), thus reflecting the overall structural conservation of chloroplast genomes. A pangenome representation analysis revealed a dominance of species-specific sequences in the graph with 273,023 cloud nodes (66%) detected in less than 5% of the species, 135,931 nodes (33%) corespond to the shell fraction and 4,462 core nodes (1%) that were detected in more than 95% of the species, representing highly conserved sequences in the chloroplast genome (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). This distribution highlights the variation at the order level with increasing conservation at the family and genus levels. Exploring the composition of genetic polymorphism represented in the graph indicated that the most common variants are SNPs (19,876), followed by indels (2,718), and larger structural variantion (1,461). Interestingly, most of the polymorphism was identified as multiallelic variats, reflecting the extensive diversity and divergence among \u003cem\u003ePoales\u003c/em\u003e species. The average nucleotide diversity across all species (π\u0026thinsp;=\u0026thinsp;0.0326) was fairly mild compared to the variation within families (π\u0026thinsp;=\u0026thinsp;0.1092\u0026ndash;0.4594), indicating that most of the variation is obtained at the genus and species levels, further supporting the stronger conservation of the core regions and divergence in the shell and cloud sequences (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Collectively, these results indicate that the species-specific cloud genomic variation is the main driver of diversity in \u003cem\u003ePoales\u003c/em\u003e while the core sequences represent deep and conserved sequences across familes and species.\u003c/p\u003e \u003cp\u003eTo further explore the conserved fraction among the \u003cem\u003ePoales\u003c/em\u003e, a consensus chloroplast genome was assembled by traversing through nodes that are present across at least 80% of species (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). The obtained consensus genome length was 77,746bp and provides a unified reference for comparative and evolutionary analyses across orders. The consensus genome containes 78 coding sequences (CDS), and 3 rRNAs. Functional annotation highlighted five genes that correspond to the Photosystem-I, 13 to Photosystem-II, 5 to the Cytochrome complex, 6 to ATP synthase, 4 are plastid-encoded RNA polymerase, and 29 correspond to ribosomal proteins (Table S4). The retention of 80% support for a node to be included in the consensus reflects that the cross-families core fraction of the chloroplast is dedicated to energy conversion, transcription, and translation. GO enrichment analysis highlighted energy conversion as the main category among consensus sequences, while non-consensus genes are associated mainly with auxiliary and regulatory functions including cytochrome complex assembly, chlorophyll binding, and light-response regulation (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eThe chloroplast evograph genome reframes the evolutionary history of\u003c/b\u003e \u003cb\u003ePoales\u003c/b\u003e\u003c/p\u003e \u003cp\u003eTo further explore the phylogenetic relationship within families in the \u003cem\u003ePoales\u003c/em\u003e, we curated a reduced dataset of 192 chloroplast genomes spanning all major tribes. This dataset better balances the broad phylogenetic coverage while reducing overrepresentation of species-rich subfamilies (e.g. \u003cem\u003ePoaceae\u003c/em\u003e), thus one representative species was selected from each tribe (Table S3). This representative dataset was used for evograph genome construction, consensus genome derivation, and graph-based phylogenetic analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003eTo investigate phylogenetic relationships among tribes and species, the evograph was transformed into a node presence/absence variation (PAV) matrix where each row represents a path (haplotype genome) and columes are nodes (topologicaly sorted) storing a binary value for presence and absence (0,1). To visualize the large PAV matrix across species, we grouped every 500 consecutive graph nodes into contiguous bins and calculated a bin-level PAV ratio (0\u0026ndash;1) representing the proportion of nodes within each bin that were traversed by a given genome path. The obtained PAV binned matrix captures all levels of genomic variation, from SNPs to large SVs and was used for clusterization and phylogenetic analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). Clusterization was conducted with the t-SNE projection and highlighted a clear family-level divergence which is consistent with the known phylogenetic relationships within \u003cem\u003ePoales\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). The \u003cem\u003ePoaceae\u003c/em\u003e family forms the most extensive and internally structured group, reflecting both its high tribal and species diversity and its complex pattern of chloroplast genome rearrangements. Within the \u003cem\u003ePoaceae\u003c/em\u003e, specific tribes and subfamilies clusters were distinguished (e.g., \u003cem\u003eTriticeae\u003c/em\u003e, \u003cem\u003eOryzeae\u003c/em\u003e, \u003cem\u003eBambusoideae\u003c/em\u003e, \u003cem\u003eAndropogoneae\u003c/em\u003e) indicating that the graph features capture the overall divergence between specific chloroplast lineages. The \u003cem\u003eBromeliaceae\u003c/em\u003e family was clustered separatly from other families, thus supporting its early divergence from other grasses lineages. In contrast, smaller families like \u003cem\u003eCyperaceae\u003c/em\u003e, \u003cem\u003eTyphaceae\u003c/em\u003e, and \u003cem\u003eEriocaulaceae\u003c/em\u003e, were grouped together, reflecting shared evolutionary history in the chloroplast structure and sequence. These clustering patterns recapitulate the early crown-node divergence within \u003cem\u003ePoales\u003c/em\u003e and highlight how PAV from graph features can retain meaningful evolutionary signals even across deep phylogenetic splits.\u003c/p\u003e \u003cp\u003eNext, we compared the phylogenetic topology constructed from 111 universal genes, including tRNAs and rRNAs, to a topology ontained from the evograph (Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e, Table S5). Comparison of traditional gene-based phylogeny and graph-derived topology revealed broad agreement at the family level with significant discrepancies within \u003cem\u003ePoaceae\u003c/em\u003e, particularly among \u003cem\u003ePanicoid\u003c/em\u003e taxa (e.g. \u003cem\u003eEriachne tenuiculmis\u003c/em\u003e, \u003cem\u003eEriachne\u003c/em\u003e sp.). The extent of topological displacement was calculated based on the absolute value of difference in the taxon leaf position in each tree and normalized based on the maximum difference found (ranges between 0 and 1). While the graph-based tree preserved the expected tribal topology the traditional phylogenetic tree blurred boundaries between \u003cem\u003eEriachneae\u003c/em\u003e and \u003cem\u003eIsachneae\u003c/em\u003e, suggesting incomplete lineage sorting or introgression in early \u003cem\u003ePanicoid\u003c/em\u003e evolution. Among species, Teisher 58, and \u003cem\u003eCoelachne africana\u003c/em\u003e showed the strongest topological displacement (0.95\u0026ndash;1.0), and moderate displacement signals were observed also among the hybrid-prone \u003cem\u003eAndropogoneae\u003c/em\u003e species (e.g., \u003cem\u003eHeteropogon\u003c/em\u003e, \u003cem\u003eHyparrhenia\u003c/em\u003e, \u003cem\u003eDiheteropogon\u003c/em\u003e, \u003cem\u003eAnatherum\u003c/em\u003e, \u003cem\u003eSchizachyrium\u003c/em\u003e 0.45\u0026ndash;0.60). Lower displacement was observed in \u003cem\u003eTragus mongolorum\u003c/em\u003e, \u003cem\u003eCoix lacryma-jobi\u003c/em\u003e, \u003cem\u003eCyperus mutica\u003c/em\u003e, and \u003cem\u003eAnanas comosus\u003c/em\u003e (0.25\u0026ndash;0.35), further supporting recurrent reticulation across \u003cem\u003ePoales\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed, Figure S12-13). These patterns were consistently reproduced in both SplitsTree and tanglegram displacement analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec, Figure S3-4), demonstrating that the graph-based topology captures genuine reticulate evolutionary signals which is frequently obscured in traditional gene-based phylogenetic trees.\u003c/p\u003e \u003cp\u003eTo reconcile discrepancies between gene-based and evograph-derived phylogenies in \u003cem\u003ePoales\u003c/em\u003e, we analyzed node-level features of the evograph beyond gene PAV, revealing that long, lineage-specific nodes strongly influence phylogenetic displacement by linking taxa through shared extended sequence homology. Across \u003cem\u003ePoales\u003c/em\u003e families, node length distributions were highly right-skewed, with the majority of nodes shorter than 50bp (Figure S5). Families differed markedly in the relative abundance and contribution of longer nodes where \u003cem\u003ePoaceae\u003c/em\u003e showed pervasive enrichment of short nodes, consistent with dense sequence turnover, whereas \u003cem\u003eJoinvilleaceae\u003c/em\u003e and \u003cem\u003eEriocaulaceae\u003c/em\u003e displayed pronounced tail of long nodes, indicating retention of extended homologous sequence blocks. These long nodes, which vary in length distribution across families, amplify topological displacement in lineages such as \u003cem\u003ePanicoideae\u003c/em\u003e and capture structural evolutionary signals that remain obscured in conventional gene-based trees.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eComparison of evograph genomes at the family level\u003c/h3\u003e\n\u003cp\u003eTo investigate divergence and connectivity at the genus level evographs, five representative \u003cem\u003ePoales\u003c/em\u003e genera (\u003cem\u003eAvena\u003c/em\u003e, \u003cem\u003eOryza\u003c/em\u003e, \u003cem\u003eTriticum\u003c/em\u003e, \u003cem\u003eElymus\u003c/em\u003e, and \u003cem\u003eBambusa\u003c/em\u003e) were selected. We impelmented graph-derived metrics including graph density and cloud-node ratio (cloud/total nodes) to capture genomic variation within each genus. These two metrics provide complementary perspectives on the graph topology, where the cloud-node ratio reflects the proportion of lineage-specific polymorphism, and the graph density quantifies the extent of shared polymorphism and structural overlap among species. For each genus, ten species were randomly sampled for evograph construction over 30 independent iterations to ensure a balanced representation and reproducibility of the statistics (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table S6). The distribution of cloud-node ratios revealed significant differences in the level of divergence between genera. \u003cem\u003eElymus\u003c/em\u003e showed the highest divergence among species (median ratio\u0026thinsp;=\u0026thinsp;0.28), while \u003cem\u003eTriticum\u003c/em\u003e consistently exhibited low divergence ratios (median\u0026thinsp;=\u0026thinsp;0.11), reflecting a highly homogeneous chloroplast genome. \u003cem\u003eOryza\u003c/em\u003e and \u003cem\u003eBambusa\u003c/em\u003e had intermediate levels of divergence (median\u0026thinsp;=\u0026thinsp;0.18 and 0.17, respectively).\u003c/p\u003e \u003cp\u003eGraph density further differentiated the connectivity patterns underlying these genera. \u003cem\u003eBambusa\u003c/em\u003e and \u003cem\u003eTriticum\u003c/em\u003e displayed the highest graph density (median\u0026thinsp;=\u0026thinsp;0.0019 and 0.0015, respectively), suggesting highly continuous conserved graph structure except for few edges of recurrent polymorphisms. In contrast, \u003cem\u003eElymus\u003c/em\u003e and \u003cem\u003eAvena\u003c/em\u003e exhibited the lowest graph densities (media\u0026thinsp;=\u0026thinsp;0.0006 and 0.0008, respectively), indicating a fragmented topology with few short conserved sequences. The \u003cem\u003eOryza\u003c/em\u003e had both low divergence and low graph density (0.0003) indicating a conserved topology which is dominated by a single main path with minimal branching or reticulate complexity (tree-like structure). This pattern is characteristic of groups with highly conserved genomes that reflects minimal structural complexity and polymorphism among species. \u003cem\u003eElymus\u003c/em\u003e represents a highly divergent and weakly connected genus, where variation is dominated by lineage-specific modifications rather than shared polymorphism. \u003cem\u003eTriticum\u003c/em\u003e shows the opposite pattern, with strong graph connectivity and limited divergence, suggesting fine-scale shared variation within a conserved genome structure. Interestingly, \u003cem\u003eBambusa\u003c/em\u003e combines moderate divergence with high connectivity, reflecting extensive internal polymorphism, while \u003cem\u003eAvena\u003c/em\u003e shows more heterogeneous outcomes with extensive differences between iterations.\u003c/p\u003e "},{"header":"Discussion","content":"\u003cp\u003eAdvances in long-read sequencing have enabled the assembly of complete genomes and their integration into graph-based pangenomes derived from multiple individuals. This approach overcomes reference bias and uncovers previously hidden structural and non-reference variation across diverse taxa (Sherman and Salzberg \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; H\u0026uuml;bner \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Garrison et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Despite their potential, graph pangenomes pose major computational challenges, as their complex networks of thousands to millions of interconnected nodes create dense, \u0026ldquo;hairball-like\u0026rdquo; structures that hinder intuitive visualization and complicate interpretation. Consequently, their application in evolutionary studies remains limited.\u003c/p\u003e \u003cp\u003eWe developed the MineGraph pipeline to address key challenges in graph construction, interpretability, and visualization by optimizing graph quality and sensitivity. The embedded optimization procedure provides a more biologically accurate and functionally meaningful representation of genetic diversity (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, Figures \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u0026ndash;S6). The increased detection of shared nodes, core genome components, and polymorphisms demonstrates its suitability for both mitochondrial and chloroplast genomes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), with potential extension to more complex nuclear graph genomes.\u003c/p\u003e \u003cp\u003eThe emergence of super-pangenomes graph-based genomic references built at the genus or higher taxonomic level has extended the pangenome concept to broader evolutionary scales. The current term \u0026ldquo;super-pangenome\u0026rdquo; only partially captures this framework, therefore we propose the term \u0026ldquo;evograph\u0026rdquo; to describe a graph-genome structure designed to explore evolutionary relationships. Evographs help reveal functional accessory regions and agronomically important traits, as was recently demonstrated in \u003cem\u003eOryza, Vitis\u003c/em\u003e and \u003cem\u003eHordeum\u003c/em\u003e (Guo et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Feng et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Shang et al., 2022).\u003c/p\u003e \u003cp\u003eWe investigated evolutionary patterns in the order \u003cem\u003ePoales\u003c/em\u003e using graph-based evolutionary pangenomes framework (evographs) that revealed structural, reticulate, and lineage-specific relationships obscured in a traditional gene-based phylogenetic tree (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, S6\u0026ndash;S12). We also generated the first unified \u003cem\u003ePoales\u003c/em\u003e consensus chloroplast genome (77,746 bp), which highlights a conserved core genes associated with energy and translation and variable dispensable genes at the order level enriched for regulatory functions (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb\u0026ndash;c). Graph-to-matrix conversion and t‑SNE clustering clearly delineated family divergence, confirming early separation of \u003cem\u003eBromeliaceae\u003c/em\u003e and highlighting lineage-specific dynamics. Dispensable \u0026ldquo;Cloud\u0026rdquo; variation seems to have shaped most of the intraspecific diversity, while conserved \u0026ldquo;core\u0026rdquo; regions formed a functional backbone focused on energy metabolism and protein synthesis. Analysis of 709 chloroplast haplotypes resolved long-standing ambiguities, such as the blurred boundaries between \u003cem\u003eEriachneae\u003c/em\u003e and \u003cem\u003eIsachneae\u003c/em\u003e within \u003cem\u003ePoaceae\u003c/em\u003e, and uncovered strong reticulation in hybrid-prone \u003cem\u003eAndropogoneae\u003c/em\u003e taxa like \u003cem\u003eCoix\u003c/em\u003e and \u003cem\u003eCyperus\u003c/em\u003e. The evograph framework showed that species-specific \u0026ldquo;cloud\u0026rdquo; sequences drive most of the genomic diversity across families despite the conserved core chloroplast structure. Node length distributions across \u003cem\u003ePoales\u003c/em\u003e were right-skewed, with most nodes shorter than 50bp, reflecting microvariation and fine-scale sequence fragmentation (Figure S5). However, long lineage-specific nodes are the major determinators of the phylogenetic signal, explaining much of the displacement between trees and graphs (Figures S3-4). Families differed in the prevalence of node lengths, where \u003cem\u003ePoaceae\u003c/em\u003e was dominated by short nodes, \u003cem\u003eBromeliaceae\u003c/em\u003e and \u003cem\u003eTyphaceae\u003c/em\u003e are depicted with broader distributions, and \u003cem\u003eJoinvilleaceae\u003c/em\u003e and \u003cem\u003eEriocaulaceae\u003c/em\u003e retained long homologous sequences. Taxa showing strong topological shifts in the tanglegram and SplitsTree analyses were tied to lineage-specific long nodes rather than short shared ones. In \u003cem\u003ePanicoideae\u003c/em\u003e, a taxa lacking plastid genes such as accD, ycf15, psaM, and ycf94, species were also enriched in long nodes, amplifying their separation in the evograph topology.\u003c/p\u003e \u003cp\u003eAt the genus level, we quantified inter-genera divergence using the cloud-nodes ratio index revealing that \u003cem\u003eElymus\u003c/em\u003e is the most divergent, and \u003cem\u003eTriticum\u003c/em\u003e is the most homogeneous and conserved among the tested genera (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Moreover, the graph density enabled to highlight structural complexity of the graph indicating that \u003cem\u003eOryza\u003c/em\u003e genus has the simplest graph structure and \u003cem\u003eBambusa\u003c/em\u003e is the most complex.\u003c/p\u003e \u003cp\u003eThe concept of the evograph represents a shift from discrete reference genome structures to complex, interconnected models that capture genomic diversity across evolutionary timescales and taxonomic breadth. The continued development of graph-based metrics and algorithms capable of capturing genetic diversity, selective pressures, phylogenetic constraints, reticulate evolutionary events, and molecular clock dynamics will be crucial for establishing a comprehensive framework for evolutionary inference. Extending evograph principles to large, complex nuclear genomes will further necessitate advances in algorithms that can effectively handle highly repetitive and polyploid assemblies. Scalable visualization and browsing tools are also essential to enable intuitive exploration of large evographs and to deepen biological interpretability. Ultimately, evographs hold the potential to transform evolutionary and applied genomics by providing a multidimensional representation of genomic diversity that informs genomic selection, genome editing, and conservation strategies. Collectively, these directions position evographs as a powerful and unifying paradigm for studying genome evolution across the tree of life.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eEvograph construction and optimization\u003c/h2\u003e \u003cp\u003eExploring genomic diversity across multiple assemblies using a graph-based genome representation requires an efficient and reproducible framework. We developed a pipeline called MineGraph to provide a standardized and fully automated approach for constructing and optimizing evographs from multiple genomes. The pipeline is fully integrated into a Docker container to ensure compatibility and smooth adjustment to computing server environment (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/hubner-lab/MineGraph\u003c/span\u003e\u003cspan address=\"https://github.com/hubner-lab/MineGraph\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The evograph construction proceedure starts with a renaming step following the PanSN specifications (Abel et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) to incorporate data of diferent sources and standartisize the different steps of the pipeline using the same samples nomenculture. After renaming, the assemblies fasta files from all samples are indexed using \u003cem\u003esamtools faidx\u003c/em\u003e (Li et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), compressed with \u003cem\u003ebgzip\u003c/em\u003e and integrated into a graph structure with the Pan-Genome Graph Builder (Garrison et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The evograph construction is strongly affected by the alignment strategy and particularly the segment length and minimum mapping identify that are passed to the wfmash aligner (Guarracino, 2022). These parameters control the sensitivity and specificity of the mapping steps, thus determining the quality of the alignment and the integrity of the constructed graph. Conceptually, longer segment lengths and higher identity thresholds increase stringency and are suitable for closely related genomes or repeat-rich regions, while lower values improve alignment sensitivity among diverged sequences. Therefore, these parameters should be set in accordance with the studied system to ensure the graph captures meaningful biological variation without over-fragmenting or over-collapsing homologous genomic regions. We applied a data-driven approach for determining these parameters. The minimum mapping identity is first determined by calculating the maximum divergence among the genomes assemblies that are integrated in the graph. Based on the calculated divergence, the mapping identity (-p parameter in PGGB) is determined as \u003cem\u003ep = (100 - (max_divergence * 100) \u0026ndash; 2)\u003c/em\u003e. To correct for potential underestimation of sequence divergence due to local varioation, and to improve mapping sensitivity, we recommend using slightly lower mapping identity threshold, thus we subtract 2 from the result. The second parameter, homology segment length between genomes, tend to be confunded by long repeats (e.g. transposable elements, tabdem repeats). To optimize this parameter, all studied organelle genomes are concatenated into a single FASTA file using \u003cem\u003eseqtk\u003c/em\u003e (Li, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and 3Mbp sequence is randomly sampled for repetitive elements analysis. Sampled sequences are then analyzed in RepeatMasker v4.0.9 (Smit et al., 2013\u0026ndash;2015) with default parameters (-no_is -s) to identify the longest repeat sequence. The segmet length parameter is determined as \u003cem\u003es\u0026thinsp;=\u0026thinsp;longest_segment * 1.2\u003c/em\u003e to capture flanking sequence beyond the repeat in the homology search. The optimized parameters are saved to \u003cem\u003eparams.yaml\u003c/em\u003e file and used automatically for the graph construction stage.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eEvogaph formats and statistics\u003c/h2\u003e \u003cp\u003eThe evograph is constructed in a graphical fragment assembly (GFA) format which describes graph components, namely the sequence segment (nodes), links (edges), and paths (sub-graphs). To increase functionality, we implemented automatic convertion to MAF/PAF alignments formats and variant calls in VCF and vg formats. Additionally, the graph is converted to a multiple sequence alignment (MSA) file and processed using RAxML v. 1.2.2 (Kozlov, 2019) to build a phylogenetic tree based on the evolutionary relationships captured in the graph.\u003c/p\u003e \u003cp\u003eTo infer biological insights directly from the evograph we apply a statistical workflow and extract key features and metrics. The graph is comprised of nodes and edges which are counted and sorted based on their frequency among samples in the graph. Nodes and edges are divided into three pangenomic fractions based on their frequency across haplotypes. Core nodes are the conserved sequences across samples and are defined by default as present among more than 95% of samples. Cloud nodes capture the rare fraction of sequences among sasmples and is defined by default as 5%. Shell nodes capture sequences that are shared among less than 95% of samples and in more than 5%. These parameters can be adjusted by the user although it is recommended to maintain a standard definition of the pangenome to enable comparison between datasets and studies (Glick and Mayrose \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). To capture the sparsity and the rate of polymorphism in the graph, we calculate the average node degree which is defined as the average number of links (edges) between neighboring nodes (sequences): n\u003csub\u003eedges\u003c/sub\u003e/n\u003csub\u003enodes\u003c/sub\u003e. In addition, the overall graph density is calculated from the ratio of observed versus theoretical edges in the graph: |E| / \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\left(\\genfrac{}{}{0pt}{}{\\left|\\text{N}\\right|\\:}{2}\\right)\\)\u003c/span\u003e\u003c/span\u003e, where |E| is the absolute number of edges (regardless their direction) and |N| is the number of nodes. The graph density depicts the expected saturation of diversity in the graph, namely how much of the expected diversity of the pangenome is actually represented in the graph. To provide a standard population genetics metrics for comparison with other systems, we also calculate the average nucleotide diversity (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\pi\\:\\)\u003c/span\u003e\u003c/span\u003e) based on biallelic SNPs called in the graph.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eRepresentation of polymorphism and consensus\u003c/h3\u003e\n\u003cp\u003eGenetic variants are called directly from the graph including single nucleotide polymorphisms (SNPs), small insertions and deletions (indels), and larger structural variations (SVs) including inversions, duplications, and translocations. These variants are summarized in a variant call format (VCF) to ensure compatibility with a broad range of external genomic analysis tools. Polymorphic sites are also converted to a presence/absence variation (PAV) matrix where the occurrence of each sequence (node) is indicated for each individual. The PAV matrix is then used to calculate distances between samples (paths) using Hamming distance or parsimony score and a phylogenetic tree is constructed.\u003c/p\u003e \u003cp\u003eTo generate a consensus sequence among samples, all nodes passing a minimum presence frequency threshold (configurable up to 100%) are concatenated into a contiuous path. The consensus sequence path is saved as a FASTA file and represents the shared path in the graph. To study the functional aspect of the obtained consensus genome across chloroplasts in the \u003cem\u003ePoales\u003c/em\u003e, we annotated the sequence using InterProScan v5 (Jones et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) and assigned the protein domains and ontologies terms (GO) based on the coding sequences. GO enrichment analysis was carried with the GOATOOLS v1 (Klopfenstein et al., 2018) and topGO v2 (Alexa \u0026amp; Rahnenfuhrer, 2023) packages in R using the go-basic.obo ontology (release July 2025) for any subset of sequences. Fisher\u0026rsquo;s exact test was applied, and multiple testing was corrected using the Benjamini\u0026ndash;Hochberg false discovery rate (FDR).\u003c/p\u003e\n\u003ch3\u003eVisualization and Interpretation\u003c/h3\u003e\n\u003cp\u003eWe provide a variety of visual outputs to support intuitive exploration and interpretation of the generated evograph genome. The final GFA graph is converted into vg and FASTA formats, allowing compatibility with external visualization tools like SequenceTubeMap (vgteam, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In addition, a dynamic HTML-based visualization is implemented using the Pyvis module (Pyvis developers, 2023) to interactively display nodes properties, connectivity, graph density, and the consensus subgraph. Other descriptive histograms and plots are generated automatically along the processing steps of the graph to facilitate further biological interpretation.\u003c/p\u003e \u003cp\u003eTo quantify local genomic diversity along the graph we implemented a sliding window quantification approach using a k-mer\u0026ndash;based analysis (Moin \u0026amp; Seemann, 2023). Each genome path in the graph was divided into consecutive k-mer windows (default k\u0026thinsp;=\u0026thinsp;1000), and the node metrics including degree, coverage, and length are aggregated into a single score. Low-frequency or structurally irregular windows in the graph indicate regions harboring private alleles, copy number variation, or a structural variant, whereas high-frequency windows indicate a conserved core segments in the graph. This unified framework captures both sequence-level and structural diversity across the pangenome.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePoales\u003c/b\u003e \u003cb\u003egenomic data\u003c/b\u003e\u003c/p\u003e \u003cp\u003eTo evaluate the performance of the pipeline, 17 representative mitochondrial and chloroplast genomes were randomly selected to construct evographs (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Genomic data were retrieved from the NCBI Nucleotide (GenBank) database for the \u003cem\u003ePoales\u003c/em\u003e order.\u003c/p\u003e \u003cp\u003eTo study chloroplast diversity among the \u003cem\u003ePoales\u003c/em\u003e, we retrieved all 1,541 complete circular chloroplast genomes available in NCBI. After filtering redundancy at the species level, 709 unique species remained (Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). To explore diversity among chloroplast genomes at the species, tribe and family level, presence and absence profiles of nodes in the evograph were analyzed. A principal component analysis (PCA) was performed with the PAV data using the package FactoMineR v2.0 (L\u0026ecirc; et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Samples were then clustered using the nonlinear dimensionality reduction methods t-SNE and UMAP as implemented in the Rtsne and uwot packages in R (Krijthe, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, Melville, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Hierarchical clustering was performed using Ward\u0026rsquo;s method on pairwise species distances, with 100 bootstraps to assess node support. Dendrograms were converted to phylo objects using ape v5 (Paradis \u0026amp; Schliep, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) for visualization. To compare graph-based and sequence-based phylogenies, RAxML trees were midpoint-rooted with phytools v1 (Revell, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and compared using tanglegrams as implemented in R package dendextend v1.1 (Galili, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Entanglement scores were quantified to evaluate topological agreement: for each taxon \u003cem\u003ei\u003c/em\u003e, we define a displacement score as the absolute difference between its vertical positions in the two trees D\u003csub\u003ei\u003c/sub\u003e = | p\u003csub\u003ei\u003c/sub\u003e⁽\u0026sup1;⁾ \u0026minus; p\u003csub\u003ei\u003c/sub\u003e⁽\u0026sup2;⁾ |, where p\u003csub\u003ei\u003c/sub\u003e⁽\u0026sup1;⁾ is the position of taxon \u003cem\u003ei\u003c/em\u003e in Tree 1, p\u003csub\u003ei\u003c/sub\u003e⁽\u0026sup2;⁾ is the position of taxon \u003cem\u003ei\u003c/em\u003e in Tree 2, and D\u003csub\u003ei\u003c/sub\u003e is the displacement score. Because absolute displacement values depend on the total number of taxa and the layout of the tanglegram, we further computed a normalized displacement D̃\u003csub\u003ei\u003c/sub\u003e = D\u003csub\u003ei\u003c/sub\u003e / maxⱼ Dⱼ where D̃\u003csub\u003ei\u003c/sub\u003e is the normalized displacement (range 0\u0026ndash;1) and maxⱼ Dⱼ is the maximum displacement observed across all taxa. species labels were harmonized across datasets using Jaro\u0026ndash;Winkler string-distance matching (Winkler et al., 1990).\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eGene presence\u0026ndash;absence variation (PAV) analysis\u003c/h2\u003e \u003cp\u003eGene-level presence\u0026ndash;absence variation (PAV) was quantified using a curated set of chloroplast genes annotated across all assemblies. For each taxon, gene presence was recorded as a binary state (1\u0026thinsp;=\u0026thinsp;present, 0\u0026thinsp;=\u0026thinsp;absent), generating a gene-by-taxon PAV matrix. Genes absent from all taxa or present in all taxa were excluded from downstream comparative analyses, as they do not contribute discriminatory phylogenetic signal. To explore the phylogenetic relationship among tribes and species, the graph was converted into a node PAV matrix using the ODGI path -H command (Guarracino et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This matrix was used as the input for graph-based similarity analyses, enabling direct comparison between classical gene-content\u0026ndash;based phylogenies and evograph-derived topologies. Because these graphs tend to scale to hunderds of thousands, millions, or more nodes, visualizing individual nodes is neither practical nor informatiove. Therfore, using the ODGI sorting order of the nodes, we clustered \u003cem\u003eN\u003c/em\u003e nodes (defaul 500) to a bin representing a contiguous segemnt of the graph. Thus, the total number of bins = #nodes/N, for each path \u003cem\u003ep\u003c/em\u003e and each bin \u003cem\u003eb\u003c/em\u003e, and the bin-level PAV ratio is defined as:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\text{P}\\text{A}{V}_{p,b}=\\frac{\\text{n}umber\\:of\\:nodes\\:in\\:bin\\:\\:visited\\:by\\:path\\:p}{total\\:number\\:of\\:nodes\\:in\\:bin\\:b}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThis ratio quantifies the fraction of nodes within a given graph region that are traversed by a specific path and take values in the range [0,1]. A value of 0 indicates that the path is completely absent from the corresponding graph region, whereas a value of 1 indicates that the path traverses all nodes within that region. Intermediate values reflect partial traversal and capture regional structural divergence across paths. The overall \u0026lsquo;presence\u0026rsquo; scores for an individual across all segments corresponds to its path in the graph which is equivalent to the full genome sequence. (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb, Figure S7), excluding non-polymorphic nodes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eGraph node length computation and filtering\u003c/h2\u003e \u003cp\u003eGraph node lengths were computed from the final evograph by measuring the nucleotide sequence length associated with each node after graph construction and normalization. Node length was defined as the number of base pairs represented by the node, independent of path multiplicity. Nodes of length 1 bp typically arising from alignment boundaries or residual graph fragmentation were excluded from all analyses.\u003c/p\u003e \u003cp\u003eTo reduce the influence of short nodes while retaining biologically meaningful sequence blocks, nodes were demultiplexed by retaining only unique nodes with a minimum length of 50bp. This threshold was chosen because nodes shorter than 50bp are overwhelmingly abundant, widely shared across taxa, and primarily reflect local micro-variation or alignment noise, whereas longer nodes are more likely to represent extended homologous sequence segments. Node length distributions were summarized separately for each family using logarithmically scaled bins.\u003c/p\u003e \u003c/div\u003e "},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eSH conceived and supervised the research project. RH and SH planned and designed the research project. RH performed the bioinformatics and data analyses. SH and RH wrote the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe MineGraph package is implemented with Python and is freely available at the H\u0026uuml;bner lab Github repository: https://github.com/hubner-lab/MineGraph\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbel, H. J., Larson, D. E., Regier, A. A., Chiang, C., Das, I., Kanchi, K. L., et al. (2020). Mapping and characterization of structural variation in 17,795 human genomes. Nature, 583, 83\u0026ndash;89. https://doi.org/10.1038/s41586-020-2371-0.\u003c/li\u003e\n\u003cli\u003eAlexa, A., \u0026amp; Rahnenf\u0026uuml;hrer, J. (2023). topGO: Enrichment analysis for Gene Ontology. R package version 2.50.0.\u003c/li\u003e\n\u003cli\u003eAlkan, C., Sajjadian, S., \u0026amp; Eichler, E. E. (2011). Limitations of next-generation genome sequence assembly. Nature Methods, 8, 61\u0026ndash;65. https://doi.org/10.1038/nmeth.1527.\u003c/li\u003e\n\u003cli\u003eChristin, P.-A., Spriggs, E., Osborne, C. P., Str\u0026ouml;mberg, C. A. E., Salamin, N., \u0026amp; Edwards, E. J. (2014). Molecular dating, evolutionary rates, and the age of the grasses. Systematic Biology, 63, 153\u0026ndash;165. https://doi.org/10.1093/sysbio/syt072.\u003c/li\u003e\n\u003cli\u003eEizenga, J. M., Novak, A. M., Sibbesen, J. A., et al. (2020). Pangenome graphs. Annual Review of Genomics and Human Genetics, 21, 139\u0026ndash;162. https://doi.org/10.1146/annurev-genom-120219-080406.\u003c/li\u003e\n\u003cli\u003eFeng, J. W., Pidon, H., Cuacos, M., Lux, T., Himmelbach, A., Haghi, R., ... \u0026amp; Mascher, M. (2025). A haplotype-resolved pangenome of the barley wild relative Hordeum bulbosum. Nature, 645(8080), 429-438.\u003c/li\u003e\n\u003cli\u003eGalili, T. (2015). dendextend: An R package for visualizing, adjusting, and comparing trees of hierarchical clustering. Bioinformatics, 31(22), 3718\u0026ndash;3720. https://doi.org/10.1093/bioinformatics/btv428\u003c/li\u003e\n\u003cli\u003eGarrison, E., Sir\u0026eacute;n, J., Novak, A. M., Eizenga, J. M., Dawson, E. T., Jones, W., Garg, S., Markello, C., Paten, B., \u0026amp; Durbin, R. (2018). Variation graph toolkit improves read mapping by representing genetic variation in the reference. Nature Biotechnology, 36(9), 875\u0026ndash;879. https://doi.org/10.1038/nbt.4227.\u003c/li\u003e\n\u003cli\u003eGarrison, E., Guarracino, A., Heumos, S., et al. (2024). Variation graph pangenomes. Nature Methods, 21, 1\u0026ndash;10. https://doi.org/10.1038/s41592-023-02180-8\u003c/li\u003e\n\u003cli\u003eGarrison, E., Guarracino, A., Heumos, S., Villani, F., Bao, Z., Tattini, L., ... \u0026amp; Prins, P. (2024). Building pangenome graphs. Nature Methods, 21(11), 2008-2012.\u003c/li\u003e\n\u003cli\u003eGlick, L., \u0026amp; Mayrose, I. (2023). Partitioning and comparative analysis of pangenomes. Bioinformatics, 39(5), btad245. https://doi.org/10.1093/bioinformatics/btad245.\u003c/li\u003e\n\u003cli\u003eGuarracino, A., Heumos, S., \u0026amp; Garrison, E. (2022). ODGI: Understanding pangenome graphs. Bioinformatics, 38(13), 3315\u0026ndash;3317. https://doi.org/10.1093/bioinformatics/btac308.\u003c/li\u003e\n\u003cli\u003eGuo, L., Wang, X., Ayhan, D. H., Rhaman, M. S., Yan, M., Jiang, J., ... \u0026amp; Ye, W. (2025). Super pangenome of Vitis empowers identification of downy mildew resistance genes for grapevine improvement. Nature Genetics, 1-13.\u003c/li\u003e\n\u003cli\u003eHubley, R., Finn, R. D., Clements, J., et al. (2016). The Dfam database of repetitive DNA families. Nucleic Acids Research, 44(D1), D81\u0026ndash;D89. https://doi.org/10.1093/nar/gkv1272.\u003c/li\u003e\n\u003cli\u003eH\u0026uuml;bner, S. (2022). Are we there yet? Driving the road to evolutionary graph-pangenomics. Current Opinion in Plant Biology, 66, 102195. https://doi.org/10.1016/j.pbi.2022.102195.\u003c/li\u003e\n\u003cli\u003eJayakodi, M., et al. (2020). The barley pan-genome reveals the hidden legacy of mutation breeding. Nature, 588, 284\u0026ndash;289. https://doi.org/10.1038/s41586-020-2947-8.\u003c/li\u003e\n\u003cli\u003eJones, P., et al. (2014). InterProScan 5: Genome-scale protein function classification. Bioinformatics, 30(9), 1236\u0026ndash;1240. https://doi.org/10.1093/bioinformatics/btu031.\u003c/li\u003e\n\u003cli\u003eKellogg, E. A. (2015). Flowering Plants. Monocots. Poaceae. In K. Kubitzki (Ed.), The Families and Genera of Vascular Plants, Vol. XIII (pp. 1\u0026ndash;416). Springer International Publishing. ISBN 978-3-319-15332-2.\u003c/li\u003e\n\u003cli\u003eKlopfenstein, D. V., et al. (2018). GOATOOLS: A Python library for Gene Ontology analyses. Scientific Reports, 8, 10872. https://doi.org/10.1038/s41598-018-28948-z.\u003c/li\u003e\n\u003cli\u003eKozlov, A. M., Darriba, D., Flouri, T., Morel, B., \u0026amp; Stamatakis, A. (2019). RAxML-NG: A fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference. Bioinformatics, 35(21), 4453\u0026ndash;4455. https://doi.org/10.1093/bioinformatics/btz305.\u003c/li\u003e\n\u003cli\u003eKrijthe, J. H. (2015). Rtsne: T-Distributed Stochastic Neighbor Embedding using Barnes-Hut implementation.\u003c/li\u003e\n\u003cli\u003eLi, H. (2012). seqtk: Toolkit for processing sequences in FASTA/Q formats.\u003c/li\u003e\n\u003cli\u003eLi, H., et al. (2009). The Sequence Alignment/Map format and SAMtools. Bioinformatics, 25(16), 2078\u0026ndash;2079. https://doi.org/10.1093/bioinformatics/btp352.\u003c/li\u003e\n\u003cli\u003eLinder, H. P., \u0026amp; Rudall, P. J. (2005). Evolutionary history of Poales. Annual Review of Ecology, Evolution, and Systematics, 36, 107\u0026ndash;124. https://doi.org/10.1146/annurev.ecolsys.36.102403.135635.\u003c/li\u003e\n\u003cli\u003eLiu, Y., Du, H., Li, P., Shen, Y., Peng, H., Liu, Z., Shi, M., Zhou, G.-A., Zhang, H., Li, Y., Zhang, W., Wang, Z., Zhu, B., Han, B., Liang, C., \u0026amp; Tian, Z. (2020). Pan-Genome of Wild and Cultivated Soybeans. Cell, 182(1), 1\u0026ndash;15. https://doi.org/10.1016/j.cell.2020.05.023.\u003c/li\u003e\n\u003cli\u003eL\u0026ecirc;, S., Josse, J., \u0026amp; Husson, F. (2008). FactoMineR: An R package for multivariate analysis. Journal of Statistical Software, 25(1), 1\u0026ndash;18.\u003c/li\u003e\n\u003cli\u003eMelville, J. (2020). uwot: The Uniform Manifold Approximation and Projection (UMAP) implementation for R. R package.\u003c/li\u003e\n\u003cli\u003eMinh, B. Q., et al. (2020). IQ-TREE 2: New models and efficient methods for phylogenetic inference. Molecular Biology and Evolution, 37(5), 1530\u0026ndash;1534. https://doi.org/10.1093/molbev/msaa015.\u003c/li\u003e\n\u003cli\u003eMoin, M., Seemann, T., et al. (2023). gretl: Graph-based read error correction and variation discovery. Microbiology Resource Announcements, 12(6), e00243-24. https://doi.org/10.1128/mra.00243-24.\u003c/li\u003e\n\u003cli\u003eMinio, A., Massonnet, M., Figueroa-Balderas, R., Castro, A., \u0026amp; Cantu, D. (2019). Diploid genome assembly of the wine grape Carm\u0026eacute;n\u0026egrave;re. G3: Genes|Genomes|Genetics, 9(5), 1331\u0026ndash;1337. https://doi.org/10.1534/g3.119.400030.\u003c/li\u003e\n\u003cli\u003eMontenegro, J. D\u003cstrong\u003e.\u003c/strong\u003e, \u003cem\u003eet al.\u003c/em\u003e (2017). The pangenome of hexaploid bread wheat\u003cem\u003e.\u003c/em\u003e The Plant Journal, 90(5), 1007\u0026ndash;1013. https://doi.org/10.1111/tpj.13615.\u003c/li\u003e\n\u003cli\u003eNurk, S., et al. (2022). The complete sequence of a human genome. Science, 376, 44\u0026ndash;53. https://doi.org/10.1126/science.abj6987.\u003c/li\u003e\n\u003cli\u003eParadis, E., \u0026amp; Schliep, K. (2019). ape 5.0: An environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics, 35(3), 526\u0026ndash;528. https://doi.org/10.1093/bioinformatics/bty633.\u003c/li\u003e\n\u003cli\u003ePaten, B., et al. (2017). Genome graphs and the evolution of genome inference. Genome Research, 27, 665\u0026ndash;676. https://doi.org/10.1101/gr.214155.116.\u003c/li\u003e\n\u003cli\u003ePotapenko, E. V., Schwarts, D., Mandel, T., Ashkenazy, N., Fuerst, D., Atsmon, G., ... \u0026amp; H\u0026uuml;bner, S. (2024). Not so selfish: transposable elements fuel adaptation to climate change. \u003cem\u003ebioRxiv\u003c/em\u003e, 2024-09.\u003c/li\u003e\n\u003cli\u003ePrice, M. N., Dehal, P. S., \u0026amp; Arkin, A. P. (2010). FastTree 2 \u0026ndash; Approximately maximum-likelihood trees for large alignments. PLOS ONE, 5(3), e9490. https://doi.org/10.1371/journal.pone.0009490.\u003c/li\u003e\n\u003cli\u003eRevell, L. J. (2012). phytools: An R package for phylogenetic comparative biology. Methods in Ecology and Evolution, 3, 217\u0026ndash;223.\u003c/li\u003e\n\u003cli\u003eSaarela, J. M., Burke, S. V., Wysocki, W. P., Barrett, M. D., Clark, L. G., Craine, J. M., Peterson, P. M., Soreng, R. J., Vorontsova, M. S., \u0026amp; Duvall, M. R. (2018). A 250 plastome phylogeny of the grass family (Poaceae): Topological support under different data partitions. PeerJ, 6, e4299. https://doi.org/10.7717/peerj.4299.\u003c/li\u003e\n\u003cli\u003eSayers, E. W., Ciufo, S., Baker, J., O\u0026rsquo;Leary, N. A., Klimke, W., McVeigh, R., Pruitt, K. D., \u0026amp; Karsch-Mizrachi, I. (2023).Strategies for resolving taxonomic name inconsistencies in large comparative genomics datasets.Bioinformatics Advances, 3(1), vbad040. https://doi.org/10.1093/bioadv/vbad040.\u003c/li\u003e\n\u003cli\u003eShang, L., Li, X., He, H., Yuan, Q., Song, Y., Wei, Z., ... \u0026amp; Qian, Q. (2022). A super pan-genomic landscape of rice. Cell Research, 32(10), 878-896.\u003c/li\u003e\n\u003cli\u003eSherman, R. M., \u0026amp; Salzberg, S. L. (2020). Pan-genomics in the human genome era. Nature Reviews Genetics, 21(4), 243\u0026ndash;254. https://doi.org/10.1038/s41576-020-0210-7.\u003c/li\u003e\n\u003cli\u003eSim\u0026atilde;o, F. A., et al. (2015). BUSCO: Assessing genome assembly and annotation completeness. Bioinformatics, 31(19), 3210\u0026ndash;3212.\u003c/li\u003e\n\u003cli\u003eSmit, A. F. A., Hubley, R., \u0026amp; Green, P. (2013\u0026ndash;2015). RepeatMasker Open-4.0. http://www.repeatmasker.org.\u003c/li\u003e\n\u003cli\u003eTettelin, H., et al. (2005). Genome analysis of multiple pathogenic isolates of Streptococcus agalactiae: Implications for the microbial pan-genome. Proceedings of the National Academy of Sciences USA, 102(39), 13950\u0026ndash;13955. https://doi.org/10.1073/pnas.0506758102.\u003c/li\u003e\n\u003cli\u003eThe Angiosperm Phylogeny Group (APG IV). (2016). An update of the Angiosperm phylogeny group classification for the orders and families of flowering plants. Botanical Journal of the Linnean Society, 181, 1\u0026ndash;20. https://doi.org/10.1111/boj.12385.\u003c/li\u003e\n\u003cli\u003eWalkowiak, S., Gao, L., Monat, C., Haberer, G., Kassa, M. T., Brinton, J., Ramirez-Gonzalez, R. H., Kolodziej, M. C., Delorean, E., Thambugala, D., et al. (2020). Multiple wheat genomes reveal global variation in modern breeding. Nature, 588(7837), 277\u0026ndash;283. https://doi.org/10.1038/s41586-020-2961-x.\u003c/li\u003e\n\u003cli\u003eWenger, A. M., et al. (2019). Accurate circular consensus long-read sequencing improves variant detection. Nature Biotechnology, 37, 1155\u0026ndash;1162. https://doi.org/10.1038/s41587-019-0217-9.\u003c/li\u003e\n\u003cli\u003eWinkler, W. E. (1990).String Comparator Metrics and Enhanced Decision Rules in the Fellegi\u0026ndash;Sunter Model of Record Linkage.Proceedings of the Section on Survey Research Methods, American Statistical Association, pp. 354\u0026ndash;359.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eZhao, Q.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eFeng, Q.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eLu, H.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eLi, Y.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eWang, A.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eTian, Q.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eZhan, Q.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eLu, Y.\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e\u003cstrong\u003eZhang, L.\u003c/strong\u003e\u003cstrong\u003e, \u003c/strong\u003e\u0026amp;\u003cstrong\u003eHuang, T.\u003c/strong\u003e (2018). \u003cem\u003ePan-genome analysis highlights the extent of genomic variation in cultivated and wild rice.\u003c/em\u003e\u003cstrong\u003eNature Genetics, 50\u003c/strong\u003e(2), 278\u0026ndash;284. https://doi.org/10.1038/s41588-018-0041-z.\u003c/li\u003e\n\u003cli\u003evgteam. (2019). SequenceTubeMap. GitHub repository. https://github.com/vgteam/sequenceTubeMap.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"genome-biology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"gbio","sideBox":"Learn more about [Genome Biology](https://genomebiology.biomedcentral.com/)","snPcode":"13059","submissionUrl":"https://submission.springernature.com/new-submission/13059/3","title":"Genome Biology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8786947/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8786947/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWith the increased accessibility and democratization of genomic sequencing, the number of fully sequenced genomes across all taxonomies is exploding. To benefit from the availability of complete genomes in evolutionary studies a complex frameworks such as graph-pangenomes are increasingly embraced. However, the high computational cost and precarious interpretation of graph-pangenomes at higher taxonomic levels hinder their broad implementation in evolutionary genomics. Here, we describe the development and application of an evolutionary graph pangenomes aproach which facilitates the analysis and interpretation of diversity across species and families. We applied our approach to 709 chloroplast genomes spanning the order \u003cem\u003ePoales\u003c/em\u003e, overcoming key limitations of traditional phylogenetic approaches by uncovering structural variation and evolutionary signals. Our results show that despite overarching structural conservation, most genomic diversity arises from species-specific \u0026ldquo;cloud\u0026rdquo; variation. We also recover from the graph, a 77.7Kbp consensus core genome for \u003cem\u003ePoales\u003c/em\u003e encoding essential cellular functions, while accessory regions capture adaptive traits. Graph-theory metrics were further implemented revealing distinct genus-level evolutionary signatures, with \u003cem\u003eTriticum\u003c/em\u003e appearing homogeneous and highly connected, in contrast to the divergent, lineage-specific structure of \u003cem\u003eElymus\u003c/em\u003e. Collectively, we propose evolutionary graph pangenomes (evographs) as a powerful and nuanced framework for resolving structural and evolutionary complexity across species and higher taxonomies.\u003c/p\u003e","manuscriptTitle":"Evolutionary graph pangenome of the order Poales from chloroplast genomes highlight phylogenetic inconsistencies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-11 18:42:51","doi":"10.21203/rs.3.rs-8786947/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-01T04:47:43+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-24T08:00:34+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-23T07:06:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"66602311671981095580855255612390399565","date":"2026-03-11T03:58:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"126455223231822853727946314071710730722","date":"2026-03-11T03:53:58+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-06T08:16:59+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-11T11:38:24+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-05T06:53:16+00:00","index":"","fulltext":""},{"type":"submitted","content":"Genome Biology","date":"2026-02-04T12:49:53+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"genome-biology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"gbio","sideBox":"Learn more about [Genome Biology](https://genomebiology.biomedcentral.com/)","snPcode":"13059","submissionUrl":"https://submission.springernature.com/new-submission/13059/3","title":"Genome Biology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d5c6c4e0-7fd6-4eb3-bf9b-5461879da284","owner":[],"postedDate":"March 11th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T04:55:30+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-11 18:42:51","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8786947","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8786947","identity":"rs-8786947","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}