Performance Profiles of Short DNA Barcode Segments for Family Level Detection of Asteraceae Within Asterales Authors: Shuai Jiang, Shunxing Ye, Yian Wang, Hongyi Zhu, Yangxu Wu, Zeqi Li, Junnan Hang, Xiaoping Li, Yan Yin, Qinghua Lu, Xinhong Guo Published: 2026 ABSTRACT Short DNA barcodes may facilitate sequence recovery from degraded material, but their ability to retain target-family identity while excluding related taxa varies among genomic regions. We computationally evaluated 16 nuclear, plastid, and mitochondrial marker regions from 11 Asterales families using 279,956 NCBI locus–record matches and an accession-disjoint discovery/test design. Thirty-one candidate segments of 50–200 bp (mean, 98.55 bp) were screened in discovery data and evaluated for within-Asteraceae sequence recall, differentiation from non-Asteraceae Asterales, in silico primer behavior, phylogenetic placement, and exploratory matching across 808 metadata-defined metagenomic samples. Conserved regions such as matR and rbcL showed high within-Asteraceae identity, whereas ITS1, ITS, and trnH-psbA showed larger differences from related-family backgrounds; ITS2 and ycf1 showed intermediate profiles. Candidate segments were placed within or immediately adjacent to Asteraceae reference branches in segment-specific maximum-likelihood analyses, although support and topology varied among regions. Metadata-defined target-containing groups had higher mean query coverage and identity than background groups; because target presence was not independently verified and no classifier was fitted, these comparisons were descriptive and did not estimate diagnostic accuracy. Definitionally linked sequence statistics were interpreted as structural associations rather than evidence of causal evolutionary mechanisms. These results provide a family-level computational comparison of candidate short segments for Asteraceae detection within Asterales. Species identification, operational marker combinations, threshold robustness, and laboratory performance require validation using taxonomically dense, voucher-linked, and experimentally characterized datasets. 1. Introduction Accurate, rapid, and reproducible species identification was fundamental to plant-resource regulation, ecological monitoring, analysis of complex environmental samples, and traceability of medicinal and edible plants [1,2,3,4]. However, plant DNA barcoding had long faced a fundamental conflict: loci that could be stably amplified and recovered across broad taxonomic groups were often not those with the highest species-level resolution, whereas highly variable regions rich in taxonomic information were frequently constrained by primer universality, segment length, and lineage specificity. Accordingly, the CBOL Plant Working Group designated rbcL and matK as core barcodes for land plants while also acknowledging that no single locus could simultaneously satisfy the ideal criteria of universal amplification, high recovery, and high resolution [5]. Subsequent large-scale studies promoted the inclusion of ITS and ITS2 in the core plant barcode system [6,7], and ycf1 was proposed as a chloroplast candidate locus with higher resolution [8]. Together, these advances indicated that plant barcoding had never been a fully resolved standardization problem, but rather had remained a problem of how to achieve an optimal compromise among different objectives. This conflict was particularly pronounced in plants because the performance of candidate regions depended not only on the number of variable sites but also on genomic compartment, mode of inheritance, and lineage history. Chloroplast regions generally provided favorable sequence recoverability and compatibility with reference databases, but their relatively conservative evolutionary rates and potential plastid introgression could reduce resolution among closely related species [9,10]. Ribosomal DNA spacer regions often accumulated more interspecific differences but could be affected by incomplete homogenization among copies, hybridization, and lineage admixture [11]. Therefore, evaluation of candidate regions should not have been limited to the degree of variation but should also have considered stability within the target taxon, the ability to exclude closely related backgrounds, and practical detectability. Asteraceae and the order Asterales provided a challenging system for evaluating family-level target recall and related-background exclusion. Asteraceae is species rich and phylogenetically complex, with cytonuclear discordance, rapid diversification, hybridization, and introgression reported in nuclear and plastid studies [12,13,14,15]. Consequently, a segment that distinguishes Asteraceae from other Asterales families cannot, without taxonomically dense congeneric testing, be assumed to identify species within Asteraceae. Plant-identification applications increasingly include degraded DNA, mixed backgrounds, low-abundance targets, and environmental metagenomic samples. Shorter markers can improve recovery from fragmented templates but may contain less taxonomic information [16,17,18,19]. Existing work has often emphasized a small number of standard loci or idealized reference matching. Here, public sequences and metadata-defined metagenomes were used to compare several aspects of candidate-segment behavior. Because neither wet-laboratory amplification nor independently verified presence/absence was available, these analyses were treated as computational screening and exploratory external matching rather than experimental or diagnostic validation. Accordingly, this study evaluated 16 candidate regions across 11 families of Asterales and screened short DNA barcode segments of 50–200 bp. The analyses asked two bounded questions: (i) how conservation, within-Asteraceae recall, and exclusion of non-Asteraceae Asterales backgrounds varied among regions; and (ii) whether the resulting multidimensional profiles could generate hypotheses for task-specific marker selection. The study did not test species identification within Asteraceae and did not establish operational screening/confirmation rules, clinical-style thresholds, sensitivity, or specificity. 2. Results 2.1. Patterns of Genetic Variation Across Candidate Regions in Asteraceae and Asterales For most candidate regions, expanding the comparison level from Asteraceae (Figure 1A) to Asterales (Figure 1B) markedly increased the proportions of variable and parsimony-informative sites (Pi sites), whereas GC content changed relatively little. In the non-coding sequences (Non-CDSs), ITS1 and ITS2, the proportions of Pi sites reached 81.0% and 82.4%, respectively, in Asterales, exceeding the corresponding values of 31.0% and 53.0% in Asteraceae; trnH-psbA likewise showed high information content. Among the coding sequences (CDSs), matK, ycf1, and rpoC1 showed relatively high variation, whereas matR and COI remained comparatively conserved at both taxonomic levels. The GC content of matR was approximately 52%, but its proportion of Pi sites was low, indicating that GC composition did not have a simple correspondence with sequence polymorphism. The numbers of transitions and transversions differed markedly among candidate regions (Figure 2). Overall, Non-CDSs accumulated more substitutions than most CDSs. Within Asteraceae (Figure 2A), ITS, ITS1, and ycf1 showed relatively high substitution levels, whereas COI and psbA showed relatively low levels. Most candidate regions showed higher substitution counts at the Asterales level (Figure 2B); however, transition and transversion counts alone were insufficient to determine substitution saturation and were therefore treated as descriptive evidence of differences in evolutionary rates among regions. The family-level GEDI matrices showed that the among-family GEDI of ITS2, trnH-psbA, and ycf1 were generally high and usually exceeded their corresponding within-family GEDI; ITS also showed pronounced among-family differentiation (Figure 3). In contrast, rbcL, psbA, and matR showed relatively small GEDI differences between among-family and within-family GEDI. These results indicated that highly variable ribosomal spacer regions and some highly variable chloroplast regions contained strong family-level discriminatory signals, whereas relatively conserved CDSs were more likely to maintain within-taxon consistency. The descriptive diversity statistics were consistent with these patterns. The mean proportion of segregating sites (Ps), Watterson’s Θ, and nucleotide diversity (π) values of the Non-CDSs in Asteraceae were 0.352, 0.0528, and 0.0349, respectively, exceeding the corresponding values of 0.281, 0.0395, and 0.0202 for the CDSs (Table 1). The corresponding values for the Non-CDSs in Asterales were 0.553, 0.0730, and 0.0588, respectively, again exceeding those of 0.448, 0.0635, and 0.0329 for the CDSs (Table 2). The π values of ITS1 were 0.0834 and 0.2214 in Asteraceae and Asterales, respectively, whereas those of COI were 0.0041 and 0.0050, respectively. 2.2. Candidate Short DNA Barcode Segments and Their Database-Matching Performance Under the prespecified operational threshold of 0.80, 31 candidate barcode segments (a mean length of 98.55 bp) were retained from the 16 genomic regions (the nucleotide sequences, conservation scores, and screening scores of the barcode segments were provided in Table S1). These candidates represent the output of the specified screening configuration rather than a threshold-independent set. The stability of candidate number, boundaries, and ranking across alternative conservation thresholds was not assessed. The support vector machine (SVM) internal stability scores of all segments exceeded 0.95, and most Mean_cons values ranged from 0.96 to 0.99. Segments from CDSs were generally more conserved; matR, rbcL, rpoC1, and psbA retained high identity even under stringent thresholds. Greater variation occurred among Non-CDSs: some segments from ITS1, ITS2, and psbK-psbI combined high conservation, whereas trnH-psbA and trnL-trnF showed comparatively lower conservation. Under the specified search conditions, the best-match query coverage and sequence identity reached 100% for all segments except ITS(1), trnH-psbA(1), and trnL-trnF(1), and all E-values were below 0.001. The reference-sequence locations of the candidate segments were consistent with the expected regions (Figure 4A–C; Table S2). Scanning tests showed that the one-dimensional and two-dimensional barcodes fully presented the segment sequences and their associated information (Figure 5). 2.3. Comprehensive In Silico Primer-Amplification Results All 31 primer pairs generated the expected Asteraceae amplicons in the custom target panel (Table S3). In the examined panels, no primer pair produced a non-Asteraceae Asterales amplicon meeting the predefined high-risk off-target rule (0/31). This panel- and parameter-specific result does not establish universal specificity; per-primer databases, target coverage, mismatches, and all predicted off-target products are reported in Table S3. For CDS-derived segments, the mean predicted amplicon length was approximately 106 bp. Mean forward- and reverse-primer Tm values were 50.8 and 49.9 °C, and the mean absolute within-pair Tm difference was 3.38 °C. For non-CDS-derived segments, the mean predicted length was approximately 90.7 bp and the mean absolute Tm difference was 7.2 °C. Primer pairs for atpF-atpH, psbK-psbI, and ITS2 showed greater GC-content and Tm variation. 2.4. Descriptive Comparison of Barcode-Matching Signals Across Metadata-Defined Metagenomic Sample Groups Among the 808 metagenomic samples, group A was assigned to the metadata-supported high-confidence target-containing group, group B to the low-abundance or complex-background target-containing group, group C to the non-Asteraceae Asterales background group, and group D to the distantly related background group (Table 3; the basic information and accession number of the metagenomic dataset are detailed in Table S4). The descriptive matching summaries showed a pronounced gradient among the four groups. Mean query coverage and percent identity were 93.17% and 99.06%, respectively, in group A and 90.86% and 98.41%, respectively, in group B. The corresponding values decreased to 40.45% and 45.59% in group C and to 10.86% and 13.02% in group D. These results indicated stronger barcode-matching signals in the metadata-supported target-containing groups than in the two background groups. Because no independent calibration set, sample-level classification threshold, or final positive/negative decision rule was used, these results represented descriptive between-group separation rather than estimates of diagnostic sensitivity, specificity, or accuracy. 2.5. Target-Taxon Recall and Discrimination Against Closely Related Backgrounds After quality control and cluster-level partitioning, 221,113 sequences were assigned to the discovery dataset and 58,843 to the held-out test dataset. No accession identifier or identical/near-duplicate cluster was shared across sets. Candidate discovery used only the discovery dataset, whereas target-family recall and related-family differentiation used only the held-out test dataset. The accession and cluster audit is provided with the reproducibility files (Table S5). The target-taxon recall analysis showed pronounced differences among segments under stringent identity thresholds (Figure 6). matR reached 100.00% for mean nucleotide identity, recall at the >95%, >98%, and >99% identity thresholds, and the proportion of full-length exact matches, thereby showing the highest within-Asteraceae stability (Figure 6A,B). At the >99% identity threshold, the recall rates of ITS2, ycf1, and rbcL were 93.36%, 93.07%, and 92.56%, respectively, and their proportions of full-length exact matches were 78.67%, 77.81%, and 80.82%, respectively. The recall rates of ITS1, rpoB, psbA, rpoC1, matK, and trnL at the >99% identity threshold ranged from 70.44% to 83.23%, and their proportions of full-length exact matches ranged from 46.65% to 82.81%. atpF-atpH, psbK-psbI, trnH-psbA, and trnL-trnF showed relatively weak recall, and substantial variation occurred among the different psbK-psbI subsegments. ITS showed a recall rate of 90.28% at the >99% identity threshold, but its proportion of full-length exact matches was 0, indicating that it was suitable for high-similarity matching but not for use of complete identity as the sole decision criterion. The analysis of differences from closely related backgrounds showed that the mean nucleotide differences in ITS1 and ITS were 48.12% and 43.33%, respectively, and that 100.00% of sequences reached a difference of at least 20% (Figure 6C,D). The mean nucleotide differences in trnH-psbA and ycf1 were 26.57% and 21.59%, respectively, with corresponding proportions of 96.51% and 80.56%. The mean nucleotide differences in matR and rbcL were only 4.42% and 2.40%, respectively, and the proportions reaching the 20% difference threshold were both 0. Joint consideration of within-Asteraceae recall and related-family differentiation produced distinct empirical profiles. ITS2 and ycf1 combined relatively high recall with moderate differentiation; ITS1 showed stronger differentiation but lower exact matching; matR and rbcL showed high recall but limited differentiation; and ITS and trnH-psbA showed larger background differences. These are descriptive family-level profiles, not predefined screening or confirmatory roles. 2.6. Structural Associations Among Conservation, Diversity, and Family Level Differentiation Under the mean within-taxon nucleotide-identity framework, average nucleotide-level identity (ANI, %) showed a significant positive correlation only with base content. The proportion of conserved sites (C sites) was negatively correlated with among-family GEDI and the proportion of Pi sites, whereas among-family GEDI was positively correlated with π, and the proportion of Pi sites was positively correlated with Watterson’s Θ (Figure 7). Gapped sequence length was positively correlated with absolute substitution counts, indicating that some count-based metrics were affected by sequence length in cross-region comparisons. Under the mean among-taxon nucleotide-difference framework, average nucleotide-level difference (AND, %) was significantly positively correlated with overall GEDI (OGEDI), whereas π, Watterson’s Θ, the proportion of variable sites (V sites), and the proportion of Pi sites showed coordinated increases. The minimum recombination events (Rm) were also highly correlated with its density (Figure 8). In addition, stratified analysis showed that the negative “conservation–informativeness” association was more stable in CDSs: as the proportion of C sites increased, the proportion of Pi sites, Watterson’s Θ, and the degree of among-family differentiation generally decreased. In Non-CDS regions, the correlation structure was less stable because fewer segments were available and the range of variation was greater. 2.7. Family Level Phylogenetic Placement Signals of Different Short Segments To evaluate whether the candidate barcode segments retained sufficient phylogenetic information for family-level placement, neighbor-joining (NJ) reference trees were first constructed from the complete sequences of each of the 16 marker regions, excluding the candidate segment queries (Figure S1). Each of the 31 candidate segments was subsequently incorporated into its corresponding complete-region reference dataset, and separate maximum-likelihood (ML) trees were constructed to assess its phylogenetic placement (Figure 9 and Figure 10). Across the segment-specific ML analyses, all candidate barcode segments were placed within or immediately adjacent to branches containing Asteraceae reference sequences, and none was unambiguously embedded within another major family of Asterales. These results indicated that the short segments generally retained phylogenetic signals supporting their assignment to Asteraceae relative to the sampled non-Asteraceae families of Asterales. However, this placement should be interpreted as evidence of family-level detection rather than species-level discrimination within Asteraceae. Placement patterns were not fully consistent among candidate segments from the same marker region. Some were nested within Asteraceae-associated branches, whereas others occurred near branch bases or showed weak separation from related families. Segments from psbA, rpoC1, matR, and other conserved regions frequently showed short internal branches, star-like patterns, or weak support. These trees assess family-level placement, not species-level identification or a complete Asterales phylogeny. 3. Discussion This study identified a descriptive gradient between within-Asteraceae sequence stability and differentiation from other Asterales families. Conserved regions more often retained high within-target identity, whereas variable regions more often showed larger differences from related family-level backgrounds. Part of this pattern is mathematically expected because conserved, variable, segregating, and parsimony-informative sites are not independent quantities. The results therefore support comparative profiling of candidate regions but do not establish that evolutionary trade-offs causally shaped barcode performance. The study integrated short-segment screening, family-level target recall, exclusion of non-Asteraceae Asterales backgrounds, exploratory metagenomic matching, and family-level phylogenetic placement. Its contribution is a multidimensional comparison rather than a claim that one locus is universally best. Because marker roles were assigned after observing the same performance profiles, the data do not yet justify an operational hierarchy of screening and confirmatory markers. This study extends earlier comparisons by integrating short-segment screening, within-Asteraceae sequence recall, exclusion of non-Asteraceae Asterales backgrounds, exploratory metagenomic matching, and family-level phylogenetic placement [17,20]. The resulting profiles can guide hypotheses for later assay development, but they do not establish operational screening or confirmatory roles because marker categories and application weights were not predefined. The differences among ITS1, ITS2, and ycf1 illustrate the observed profiles. ITS1 showed the largest mean difference from non-Asteraceae Asterales backgrounds but lower exact matching than ITS2. ITS2 and ycf1 combined relatively high within-Asteraceae recall with moderate background differentiation. These statements are restricted to the family-level comparison used here. In addition, full ITS, ITS1, and ITS2 overlap physically and are not independent loci; retaining all three as alternative candidate definitions increases representation of the ribosomal cistron and requires sensitivity analysis [7,8,11]. High within-Asteraceae identity in matR and rbcL identifies them as candidates for broad target recovery, but the present data do not support stand-alone confirmatory use. Formal marker-role assignment should use a prespecified loss function or Pareto-front analysis in discovery data and independent testing. Decision-curve analysis would become appropriate only after a validated binary endpoint and operational prediction model are available [20,21]. The public metagenomic comparison provided an external-context plausibility check rather than diagnostic validation. Public metadata did not independently establish the presence or absence of Asteraceae DNA [16,17,18,19,22]. A latent-class model was not fitted because the dataset lacked multiple conditionally independent assays required for reliable identification [23]. Future validation should use voucher-linked morphology, whole-genome, plastome evidence, spike-ins, or orthogonal assays. The current ranking is post hoc. Before assigning operational screening or confirmatory categories, future studies should define objective weights, calculate a Pareto frontier using held-out recall, nontarget exclusion, primer feasibility, and uncertainty, and report bootstrap stability [23,24]. The phylogenetic comparisons were not designed to infer the complete evolutionary history of Asterales. Mechanistic explanations would require locus-appropriate analyses. For CDSs, this could include orthology-checked, codon-aware dN/dS models stratified by branch or site; for Non-CDSs, conserved-element annotation and substitution-rate models would be more appropriate. Recombination summaries calculated across mixed species are descriptive and cannot be interpreted as population recombination rates. These analyses are beyond the current descriptive scope [25,26]. From an applied perspective, the screened segments should be regarded as candidates for assay development. Any future use in medicinal-plant identification, food traceability, or environmental DNA monitoring requires comprehensive primer-specificity assessment and experimental testing of amplification efficiency, limit of detection, repeatability, mixture performance, and cross-region robustness [27]. The following workflow is proposed for future experimental validation and was not performed in the present study (Figure 11). Representative candidate segments should first be selected using prespecified criteria and an independent computational test set. Primer pairs should then undergo comprehensive genomic off-target assessment before laboratory testing [3,4,16]. Voucher-linked Asteraceae tissues, closely related non-Asteraceae Asterales, degraded templates, and complex environmental samples should be tested. PCR products should be evaluated by gel electrophoresis and confirmed by Sanger or amplicon sequencing. Taxonomically dense congeneric sampling is required before any species-level claim can be made [5,6,10,15]. Artificial mixtures and dilution series should quantify the limit of detection, background interference, replicate consistency, and false-positive amplification. Morphological vouchers, whole-genome or plastome evidence, or orthogonal assays should establish target presence and absence independently of the candidate barcodes [17,18,19]. Candidate segments may subsequently be developed for qPCR or probe-based detection, with amplification efficiency, limit of detection, limit of quantification, and repeatability reported according to applicable assay guidelines [22,27]. Operational screening and confirmation roles should be assigned only after prospective validation and prespecified decision analysis. A principal limitation of this study was that the performance of the candidate barcode segments was evaluated mainly using public sequence databases and computational analyses; therefore, their amplification efficiency, analytical sensitivity, specificity, and ecological applicability still require validation across a broader range of real-sample scenarios. In addition, candidate-segment discovery relied on a single operational conservation threshold of 0.80, corresponding to normalized Shannon entropy ≤ 0.20. Although this threshold was applied uniformly to restrict the screening to relatively low-entropy alignment positions, it was not a biologically validated or universally optimal cutoff. Because alternative threshold settings were not evaluated, the number, coordinates, length distribution, and performance ranking of the 31 candidate segments should be interpreted as conditional on the specified screening configuration rather than as threshold-independent or universally robust results. Moreover, although a systematic multi-metric evaluation framework was established, its generalizability requires further assessment using independent, accession-separated external datasets and cross-regional validation. Future studies should therefore systematically evaluate alternative conservation thresholds, repeat both candidate discovery and downstream performance assessment using independent datasets, and validate representative segments experimentally. Combining such validation with continuously updated sequence resources and data-driven models could ultimately enable more precise, robust, and dynamically optimized barcode-design strategies [20,22]. 4. Materials and Methods 4.1. Overview of the Study Design and Workflow A systematic framework was used to compare candidate short DNA barcode segments for Asteraceae detection within Asterales (Figure 12). Public sequences were quality controlled and aligned; segments were screened using conservation and length criteria; and descriptive variation, internal matching, within-Asteraceae recall, non-Asteraceae Asterales exclusion, exploratory metagenomic matching, and family-level phylogenetic placement were compared. The workflow did not evaluate species identification within Asteraceae. 4.2. Sequence Retrieval, Quality Control, and Multiple-Sequence Alignment Sixteen candidate DNA barcode regions were retrieved for plants from 11 families of Asterales (Alseuosmiaceae, Argophyllaceae, Asteraceae, Calyceraceae, Campanulaceae, Goodeniaceae, Menyanthaceae, Pentaphragmataceae, Phellinaceae, Rousseaceae, and Stylidiaceae) from the Nucleotide and Gene databases of the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov/, accessed on 10 January 2026) [28,29]. The CDSs included COI/COI-5P, matK, matR, psbA, rbcL, rpoB, rpoC1, and ycf1, whereas the noncoding sequences (Non-CDSs) included atpF-atpH, ITS, ITS1, ITS2, psbK-psbI, trnH-psbA, trnL, and trnL-trnF. As of 20 January 2026, the 176 NCBI Nucleotide queries yielded 279,956 locus–record matches. After deduplication by NCBI UID across queries and loci, 135,942 unique sequence records remained. Taxonomic validation retained 134,372 records assigned to the intended 11 families of Asterales. These records represented 14,886 unique named species-level taxa belonging to 1706 genera, including 14,745 non-hybrid species and 141 named hybrid taxa. An additional 3442 records lacked an unambiguous species-level assignment and were therefore excluded from the species-level taxon counts. Exact queries, dates, accession lists, raw and deduplicated records, quality-controlled sequences, species-level taxa, genera, hybrids, and ambiguous records for every locus were reported in Table S5. After sequence retrieval, standardized quality control was performed using SeqKit v2.13 [30]. For CDSs, clearly incomplete or abnormally extended sequences were removed. For ITS and intergenic spacer regions, sequence annotations were used to determine whether abnormal lengths resulted from genuine insertion–deletion variation, and only incomplete, misannotated, or clearly anomalous records were removed. Multiple-sequence alignments were then performed separately for each candidate region using MAFFT v7 [31] with default parameters. Before candidate-segment discovery, the quality-controlled sequences for each locus were divided into nonoverlapping discovery and test datasets. The discovery dataset was used exclusively for multiple-sequence alignment, site-conservation estimation, SVM-assisted screening, and determination of candidate-segment coordinates. The test dataset was withheld from all candidate-discovery procedures and was used only for the subsequent evaluation of target-group recall and differentiation from non-Asteraceae Asterales backgrounds. To prevent information leakage, identical and near-duplicate sequences were identified across the combined dataset using SPS v1.0 software [27] (https://github.com/JS-HNU/SPS, accessed on 10 January 2026) with thresholds of ≥80% sequence identity and ≥80% alignment coverage (Refer to the research conducted by Jiang et al.) [17,21,27]. All sequences belonging to the same duplicate or near-duplicate cluster were assigned entirely to either the discovery dataset or the test dataset. Dataset assignments were performed at the cluster level rather than at the individual-record level. Accession numbers and cluster identifiers were subsequently cross-checked, confirming that no accession or near-duplicate sequence cluster occurred in both datasets. 4.3. Characterization of Genetic Variation and Sequence Diversity Descriptive analyses of genetic variation were performed for the candidate regions at the Asteraceae and Asterales levels using MEGA v12 and DnaSP v5 [32,33], respectively. The statistical metrics included sequence length, nucleotide composition, C sites, V sites, Pi sites, segregating sites (S sites), the number of identical pairs (II), transitionsal pairs (SI) and transversional pairs (SV), nucleotide diversity (π), Ps, Watterson’s Θ, Tajima’s D, and the minimum number of Rm [34,35,36]. Because the data spanned multiple species and taxonomic units, Tajima’s D and Rm were used only as descriptive indicators of sequence-variation structure and were not used to directly infer population-level selection or actual recombination processes. To compare within-family and among-family differentiation across candidate regions, family-level GEDI and OGEDI matrices were calculated [37]. For locus l, let dl(xi,xj) be the pairwise nucleotide distance after alignment filtering, and let na and nβ be retained sequence counts for families a and b.GEDIl(a,a)=2na(na−1)∑1≤i