{"paper_id":"5c54a8bc-4ef2-4fcc-aa74-29fc541744bf","body_text":"Title: Protists Exhibit Stronger and More Recent Oceanic Genetic \nStructure than Archaeplastida and Metazoa \nAuthors: Rubén González-Miguéns1*, Alex Gàlvez-Morante1, Guifré Torruella1, \nCédric Berney1, Elena Casacuberta1, Iñaki Ruiz-Trillo1,2* \nAffiliations:  \n1Institut de Biologia Evolutiva (CSIC-Universitat Pompeu Fabra), 08003, Barcelona, \nSpain \n2ICREA, Barcelona, Spain \n* Rubén González-Miguéns. Email: ruben.miguens@ibe.upf-csic.es \n* Iñaki Ruiz-Trillo. Email: inaki.ruiz@ibe.upf-csic.es \n \nAbstract: The global distribution of biodiversity is shaped by a complex interplay of \nevolutionary history and ecological processes. While biogeographic patterns are well \ndefined for animals and plants, the global distributions of protists remain unclear. A key \nquestion is whether protists follow the same broad biogeographic principles as \nmacroscopic life. To address this, we compiled 88 marine COI metabarcoding studies \nperforming population‑genetic analyses across ocean basins. Our results reveal that most \nprotist phyla exhibit pronounced genetic structure among oceans, a pattern exceeding that \nreported for Archaeplastida and Metazoa. This likely reflects  recent, and potentially \nhuman-mediated, introductions , influencing protist dispersal and contemporary \ncommunity assembly.  By demonstrating that protist distributions are not historically \ncosmopolitan, our study supports the existence of common eukaryotic biogeographic \npatterns that transcend organismal size. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nMain Text:  \nUnicellular eukaryotes (protists), which represent most of the diversity of eukaryotes \nfrom which multicellular groups evolved, have classically been viewed as fundamentally \ndifferent from multicellular plants and animals with regard to their biogeographic patterns \n(1). Early theories postulated that protists have cosmopolitan distributions and virtually \nunlimited dispersal capacity, giving rise to the famous concept “everything is everywhere, \nbut the environment selects”  (2). This view, however, was based largely on limited \nmorphological data. The advent of large -scale molecular techniques, such as \nmetabarcoding, now permits better characterization of protist distributions (3).  \nSome recent marine metabarcoding studies have begun to reveal that protists exhibit \nbiogeographic patterns more similar to those of multicellular eukaryotes than previously \nassumed (4–6). Nevertheless, these investigations have predominantly focused on  \npresence-absence matrices, beta diversity patterns, and relative abundances inferred from \nread depth. A critical knowledge gap persists regarding population-level genetic diversity \nand the  historical biogeographic processes shaping these distributions  (7). This gap \nhinders our ability to determine whether the biogeographic principles observed in \nmulticellular animals and plants are truly universal across all eukaryotes , or if smaller \ntaxa follow distinct rules in marine environments. To resolve this metabarcoding offers a \npromising approach (11), as the large amount of molecular and geographic data from \ndifferent organisms enables the integration of geographical and population genetic \ntheories within a unified biogeographic framework  (8–10). However, its effectiveness \ndepends on meeting certain conditions: i) the use of molecular markers with intraspecific \nresolution (for example , the mitochondrial cytochrome oxidase subunit I gene, COI, \nwhich has proven invaluable in animal biogeography  (12)); ii) the availability of a \ncomprehensive global database for such markers that includes data from both protists and \nmulticellular eukaryotes (13); iii) the use of standardized taxonomic units that mitigate  \nthe inherent biases of metabarcoding  (14); and iv) the integration of population genetic \nmethods and theory into metabarcoding analyses.  \n  \nConstruction of a database with informative OTUs to test biogeographic patterns \nTo investigate global eukaryotic biogeographic patterns at the population level, a critical \nfirst step is the creation of a comprehensive metabarcoding database amenable to \npopulation-genetic analyses. So, we here compiled data from 88 independent \nmetabarcoding studies into an environmental -COI (eKOI metabarcoding) database, \ncomprising 976,865 amplicon -sequence variants (ASVs) with no abundance filter ( Fig. \nS1) [see supplementary data (15)]; and 302,809,791 reads from 4,102 samples collected \nat 957 unique sites that span all oceans (Fig. 1a). We then taxonomically annotated them \nat phylum level, clustering them into 51,558 operational taxonomic units (OTUs).   \nTo ensure robust population -genetic analyses and minimize potential artifacts, we \nstringently filtered the data to retain only informative OTUs, defined as those i) \ncontaining at least four sequences from a minimum of two distinct locations , and ii) \ncontaining at least one pair of sequences that differ by  ≥ 1% (uncorrected p‑distance). \nThis rigorous filtering resulted in a final dataset of 5,970 informative OTUs, comprising \n276,303 ASVs and 101,898,078 reads distributed across 41 phyla (Fig. 1b, 2a). Crucially, \ninter-phylum comparisons of nucleotide diversity (π) and Tajima’s D revealed minimal \nsignificant differences (0.001% and 6.9% of pairwise comparisons, respectively; Fig. S2; \nSupplementary Data), confirming broadly comparable molecular-evolutionary pressures \nacross phyla. Consequently, this constructed eKOI metabarcoding resource provides \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nstandardized taxonomic units, thereby establishing a solid foundation for rigorous, \npopulation-level investigations of marine biogeographic hypotheses. \n \nProtists have larger geographic ranges than animals and plants \nA long -standing debate in eukaryote biogeography concerns whether protists possess \nlarger geographic ranges than macro -organisms (1). Testing this hypothesis  has been \nhindered by the sheer volume of molecular and geographic species-level data required at \na global scale  (16). Our eKOI database allows for a direct assessment of this question. \nWe analyzed the distributional patterns of all eukaryotic phyla by comparing the range \nsize of every informative OTU within each eukaryotic phylum, quantifying i) the mean \npairwise geographic distance among sampling sites , and ii) the maximum distance \nbetween any two sites. The results show that while Metazoa and Archaeplastida show \nmean inter-sample distances around 1,000 km, most protist phyla average over 5,000 km, \nwith some individual OTUs exceeding 10,000 km ( Fig. S3). This demonstrates that \nprotists, at the OTU level defined by COI,  possess considerably larger geographic ranges \nthan plants and animals. \n  \nMolecular and geographic expansions in protists \nTo understand why protists exhibit these distinctive geographic distances patterns, \nadditional data, particularly molecular data, are essential, since they enable the detection \nof within-OTU genetic patterns linked to geographic range (9). Therefore, we compared \ngenetic versus geographic distances within each informative OTU to test whether the \nobserved spatial patterns were non -random and consistent across different phyla. Our \nanalyses using both linear and nonlinear regressions revealed that most OTUs displayed \nsignificant non -random spatial patterns. Specifically, when compared to null models, \n81.5% of OTUs in the linear models and 92.6% in the nonlinear models had R² values \nthat were statistically significant (p < 0.05; Supplementary Data). Furthermore, linear and \nnonlinear models (Fig. S4), as well as Mantel tests (Fig. S5), consistently showed positive \nrelationships between genetic and geographic distances across all eukaryotic phyla \nexamined. In other words, genetic divergence tended to increase with greater geographic \nseparation. The rate of genetic divergence per unit distance, quantified by the slope of the \nlinear regression and by the parameter “a” of the nonlinear (exponential) model, was \nlargely uniform across phyla. Significant differences among phyla in these rates were \nrare, occurring in only 5.7% of pairwise comparisons for the linear slopes and 3.6% for \nthe nonlinear “a” parameters. Moreover, the geographic extent of an OTU (mean inter -\nsample distance) was only very weakly co rrelated with its genetic divergence rate, \nexplaining merely 3.4% of the variance in slope and 4.0% in the “a” parameter. Together, \nthese results indicate that geographic distance alone does not strongly limit genetic \ndifferentiation in protists, even across vast oceanic distances. We propose that the primary \nexplanation for this weak isolation-by-distance pattern is the recent introduction and rapid \ngeographic expansion of many protist species  (17, 18). Such rapid expansion over short \nevolutionary timescales would limit the accumulation of substantial genetic divergence, \nthus maintaining high genetic homogeneity across geographically distant populations.  \nNevertheless, this interpretation should be treated with caution, as multiple introductions \ncan sometimes reduce genetic founder effects (19), yet molecular diversity often remains \nlower compared to that observed in the centers of origin. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nMetazoa exhibit higher richness of haplotypes among eukaryotes \nTo follow up on those rapid geographic expansions, we then inquired how and when \ncurrent species distributions were established. To achieve this, we performed explicit \npopulation‑genetic analyses to determine how molecular diversity is structured within \neach informative OTU. Our approach involved examining the genetic structure of these \nOTUs across different oceanic regions, treating each region as a distinct population. This \nallows for inferences regarding both contemporary and historical processes shaping  \ncurrent species distributions. To ensure unbiased analysis, population -level units within \neach OTU were defined irrespective of sequencing depth variations. We observed that \nhaplotype richness per informative OTU had only a weak correlation with sequencin g \ndepth (R² = 18.2%), therefore being a robust metric for inferring population -genetic \npatterns in metabarcoding studies. Using the haplotypes identified within each OTU, we \nconstructed haplotype networks for each informative OTU, without imposing a strict ly \nbifurcating tree topology  (20). This network -based approach allowed us to infer intra -\nOTU genealogies while avoiding the assumption of purely bifurcating relationships \ninherent to traditional phylogenies, which may not accurately capture genetic structure at \nthe population level (21). \nAs a first step, we characterized haplotype diversity distributions across major phyla \nusing three complementary metrics: haplotype diversity (Hd) (22), branch diversity (Bd), \nand a combined haplotype -branch diversity index (H bd) (23). These metrics were \ncalculated based on the number of ASVs (unique haplotypes), the total read abundance \nper haplotype, and log -transformed read counts. B d and H bd values showed relative \nconsistency across metazoan phyla, with few significant inter -phylum differences ( Fig. \nS6). In contrast, H d exhibited greater variation across phyla, primarily driven by \ncomparisons involving Metazoa, which consistently displayed near -maximal Hd values. \nThis indicates a significantly higher richness of unique haplotypes and, consequently, \ngreater intra-OTU genetic diversity within metazoan OTUs compared to those of other \neukaryotic phyla. \n \nGenetic Structuring Among Oceanic Regions suggest recent OTU introduction \nevents in protist phyla \nFollowing the characterization of within -phylum haplotype diversity, we next assessed \ngenetic structuring across oceanic regions. For each informative OTU, we analysed \nhaplotype frequency and distribution patterns across sampled oceans. We quantified \npopulation-level differentiation using four standard metrics of population genetic \nstructure: total genetic diversity (H t), within -ocean genetic diversity (H s), the fixation \nindex (Fst) (a measure of among-ocean differentiation) (24), and the effective number of \nmigrants (Nm) (an indirect estimate of gene flow) (25). These metrics revealed significant \ndifferences among phyla in their degree of genetic structuring, with 20% of inter -phyla \ncomparisons showing a significant difference in H t, 9.7% in H s, 21.9% in F st, and 9.3% \nin Nm (Fig. S7). The most pronounced disparities were observed in comparisons involving \nMetazoa and Archaeplastida versus other phyla. Specifically, Metazoa and \nArchaeplastida showed higher H s values, reflecting greater average haplotypic diversity \nwithin individual oceans. This inter -oceanic divers ity was coupled with very low F st \nvalues in both clades (often approaching zero and never exceeding 0.5), signifying \nminimal genetic differentiation among oceanic regions. So, Metazoa and Archaeplastida \nhaplotypes tend to be geographically restricted. In contrast, protist phyla exhibited much \nstronger genetic structuring, with consistently lower Hs and high F st values (≥ 0.5). This \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\npattern indicates that although lineages in those phyla occur across multiple oceans, their \npopulations remain genetically isolated between regions. We propose that this strong \nstructure in protists may reflect recent geographic expansions or introductions rather than \nlong-term historical isolation. \nTo address the limitations of H s, in capturing haplotype distribution across oceans, we \ndeveloped the Population-level Haplotype Diversity (PHd) metric, which integrates intra-\noceanic diversity and the evenness of haplotype distribution across oceans  [see \nsupplementary materials and methods (15)]. PHd penalizes uneven distributions, and the \ndifference between PHd and Hs (PHd - Hs) indicates the degree to which an OTUs genetic \ndiversity is structured between oceans. PH d analysis revealed significant inter -phylum \ndifferences, 20.6% of comparisons, largely mirroring H s and F st trends. Metazoa, \nArchaeplastida, and Ochrophyta showed near -zero PH d values and negative PH d - Hs \ndifferences, suggesting concentrated haplotypic diversity in limited oceanic regions (Fig \n2a and b). However, some metazoan group s such as Arthropoda (PH d - Hs = 0), \nGastrotricha (-0.17), Placozoa (0.09), and Rotifera (-0.21), exhibited more even haplotype \ndistributions. Protist phyla generally showed slightly higher PH d values and PH d - Hs \ndifferences close to zero, indicating moderate population genetic structuration with \ntraceable dominant oceans. Overall, PH d analysis supports the hypothesis of recent \ngeographic expansions or introductions shaping genetic patterns in these marine \neukaryotes. \nFurther analysis using the PH d metric to identify dominant source oceans, harboring the \nmajority of that OTU’s haplotypic diversity, and introduced (recipient) oceans revealed \nclear differences among taxonomic groups. Metazoa, Archaeplastida, and Ochrophyta \nshowed minimal evidence of multi-ocean distributions (Fig. 2C, S8). In contrast, certain \nmetazoan and protist phyla frequently displayed one or more introduced oceans per OTU, \nindicating recent inter -ocean colonization. For example, Arthropoda, Gastrotri cha, \nPorifera, Rotifera, and Xenacoelomorpha, often had broader distributions, averaging \naround one introduced ocean per OTU (Fig. S8). Similarly, protists frequently displayed \none or more introduced oceans per OTU, indicating a strong signature of recent inter -\nocean introduction, with low haplotypic diversity, in these phyla. However, caution is \nrequired when characterizing species as native or introduced, as the evolutionary history \nof some species remains unclear with the current data (i.e, cryptogenic species) (26). To \nminimize misclassifications, we only considered OTUs with clear patterns of intra -OTU \ndiversity across oceans [see supplementary materials and methods (15)]. \nTo validate the capacity of our approach to identify genuine introduction events, we \nexamined specific OTUs within the phylum Chordata, where introduction histories are \noften well -documented (27).  Our analysis successfully highlighted several OTUs \ncorresponding to species known for human -mediated introductions or widespread \ndispersal, such as Acipenser gueldenstaedtii, Botryllus schlosseri, Salmo trutta, Siganus \nrivulatus, and Styela plicata [see supplementary data (15)]. This reinforced the reliability \nof our approach in identifying recent introductions from metabarcoding data.  \nFurthermore, the inferred geographic scope of these colonization events differed \nmarkedly between groups. In Metazoa and Archaeplastida, most inferred source -sink \nconnections occurred between geographically adjacent oceans and within the same \nhemisphere, suggesting predominantly regional spread (Fig. S9). In contrast, protist phyla \nOTUs often showed long -distance dispersal between non -adjacent oceans and different \nhemispheres, implying that recent introductions of these microorganisms frequently span \ndistant oceanic regions,  far exceeding the regional spread seen in larger eukaryotes (28).  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nFinally, by aggregating dominant and introduced ocean assignments across all phyla, we \nuncovered broad biogeographic trends that align with known global patterns of species \nintroductions [see supplementary materials and methods  (15)]. The highest numbers of \nintroduced informative OTUs were detected in Northern Hemisphere oceans, particularly \nthe Atlantic and North Pacific  (Fig. 3), consistent with these regions’ historically high \nrates of species introductions (27, 29, 30). In contrast, the main source regions (dominant \noceans) across taxa were the Mediterranean Sea and Indian Ocean, both noted for their \nhigh levels of endemic diversity  (31, 32). Overall, these findings underscore that recent \ncolonization, most likely facilitated by anthropogenic activities, have left a distinct \nimprint on global marine genetic diversity, with identifiable source areas and introduction \nhotspots explaining the distribution patterns we observe. \n  \nConclusion \nOur global COI analysis, leveraging informative OTUs and the innovative PHd metric \nfor assessing population -level genetic diversity, definitively reveals fundamental \ndistinctions in the contemporary biogeography of marine eukaryotes. Metazoa and \nArchaeplastida generally exhibit significantly restricted ranges and weaker ocean -basin \ngenetic structure compared to protist phyla, underscoring divergent molecular population \ndynamics across major lineages.  \nThese patterns likely reflect the profound impact of contemporary events, notably \nanthropogenic dispersal  (33). While global shipping is a known vector for metazoan \nintroductions (34), our data robustly indicate that protists and smaller metazoans \n(Arthropoda, Rotifera, Gastrotricha, Placozoa) show heightened susceptibility to long -\ndistance dispersal, evidenced by elevated PHd values with respect other metazoan phyla. \nMoreover, the lower PH d in these smaller metazoans compared to protist s phyla likely \nstems from protists rapid generation times  (35), or multiple introductions  (19), could \nfacilitate the genetic homogenization in newly colonized oceans , reducing the founder \neffect. Conversely, restricted distributions and weak structure in some small and \nunicellular Archaeplastida suggest lower transoceanic transport survival, potentially due \nto light and resource limitations, corroborated by the differential recovery of taxa in \nballast water studies. Consistent with this explanation, studies of ballast water frequently \nrecover resilient photosynthetic taxa such as Bacillariophyta and  dinoflagellates, which \nare capable of producing resistant forms  (36), whereas groups like Chlorophyta appear \nunderrepresented (37, 38). This suggests that ballast water may be a less effective vector \nfor long-distance dispersal in certain Archaeplastida lineages. \nUltimately, our global study establishes a unified framework for interpreting broad \nbiogeographic and population genetic patterns across diverse eukaryotic phyla. The \ncongruent patterns of marine geographic distribution and genetic structure within each \ngroup, strongly suggest shared historical processes, such as recent introductions, shaping \ntheir contemporary ranges. These findings challenge the traditional view of protists as \nubiquitously dispersed entities (1), at the population genetic level, and reveal potentially \nuniversal principles governing biogeographic dynamics and community assembly , that \ntranscend organismal size and complexity, particularly in the context of the \nAnthropocene. \n \nMaterials and Methods \neKOI metabarcoding database \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nTo generate the eKOI metabarcoding database we selected 88 marine metabarcoding \nstudies, with publicly accessible raw sequencing data, that had been generated on an \nIllumina paired-end platform using the cytochrome oxidase subunit I (COI) molecular \nmarker, and employed the following primers: HCO \n(TAAACTTCAGGGTGACCAAAAAATCA) (39), LCO \n(GGTCAACAAATCATAAAGATATTGG) (39), jgHCO2198 \n(TAIACYTCIGGRTGICCRAARAAYCA) (40), mlCOIintF \n(GGWACWGGWTGAACWGTWTAYCCYCC) (41), and mlCOIintF -XT \n(GGWACWRGWTGRACWITITAYCCYCC) (42). After obtaining the raw data for \neach study, we processed each dataset separately according to the protocol outlined in \n(43). As a first step, primer trimming and demultiplexing were performed with Cutadapt \n(version 2.8) (44), when necessary. Next, the resulting reads for each sample in each \nmetabarcoding study were processed independently using the DADA2 R package (45), \nwith minor modifications applied depending on the specific metabarcoding study. \nChimeric sequences were subsequently removed using the DADA2 \nremoveBimeraDenovo function. After inferring the amplicon sequence variants (ASVs), \nwe stored them in a FASTA file using the following header format: >sequence_ID \nmerged_sample={ 'LOCALITY_ID1': reads, 'LOCALITY_ID2': reads, ... }. A separate \nmetadata file contained information associated with each locality ID (see metadata.csv \nin the supplementary data). For each sample in the metadata, we then assigned the \ncorresponding ocean or sea. To achieve this, we downloaded shapefile datasets from the \n“Global Oceans and Seas” database (version 1, accessed 05/11/2024) and “Natural Earth” \n(accessed 05/11/2024). Using each locality’s latitude and longitude coordinates, we \ndetermined and recorded the respective ocean or sea for that sample. For subsequent \nbiogeographic analyses, we utilised the broad ocean regions defined in the “Global \nOceans and Seas” database, which groups the world into 11 major seas and oceans. This \napproach was chosen to avoid excessive subdivision of oceanic regions. \nAll FASTA files generated by each metabarcoding study, in the above format, were then \nmerged into a single file ( eKOI_metabarcoding_database.fasta in the supplementary \ndata) using a custom Python script (1_fastas_combine.py). This script utilises the \nBiopython library (46) to identify identical ASVs across studies and combine their \nlocality information. Each unique ASV was assigned an identifier in the format eKOIX \n(where X is a unique number for each ASV). Once the final merged FASTA file was \nassembled, we checked it for c himeric sequences using the VSEARCH uchime_denovo \ncommand (version 2.14.1) (47). \nTaxonomic assignment of the ASVs was performed using the eKOI taxonomy database \n(13). For this purpose, we used the script (5_taxonomic_assignation.py) develop in that \nstudy. This script created a separate folder for each FASTA file in the working directory. \nWithin each folder, an Excel file was generated containing the taxonomic assignme nt \ninformation for each ASV \n(eKOI_metabarcoding_database_taxonomic_annotations.xlsx in the supplementary \ndata), obtained via the VSEARCH usearch_global command. ASVs with less than 84% \nsimilarity to any reference sequence were not considered further. Finally, we generated \nseparate FASTA files for each taxonomic level of interest; in this case, we chose the \nphylum level. The resulting ASV sequences for ea ch phylum can be downloaded from \nthe “Phyla_data” folder in the supplementary data as eKOI_database.fasta file per \nphylum. \n \nSequence Data Processing and Analysis \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nOTU Clustering \nOnce FASTA files were generated for each phylum, they were analysed independently. \nFirst, all metadata for each ASV was extracted using the script \n2_generate_abundances_unique.py. Biopython was used to extract each ASV’s ID and \nmatch it with the metadata f ile, producing a DataFrame containing the sequence ID, \ngeographic coordinates, reads and ocean for each record (abundances_unique.csv in the \nsupplementary data per phylum). Because some ASVs were found in multiple localities, \nduplicate entries were created for each locality where an ASV occurred by appending a \nsuffix “_dupN” (where N is the duplicate number) to the ASV’s ID, yielding a unique \nidentifier for each locality occurrence of that ASV. To avoid biases from multiple \nsequencing replicates per locality (common in metabarcoding studies), only one ASV \nentry per locality was retained (duplicate entries with identical ASV ID and coordi nates \nwere removed). After this filtering, a comprehensive FASTA file was compiled, including \nall ASV sequences (with each locality -specific duplicate as a separate entry). This file \nwas aligned using MAFFT version 7.490 (48), with optimised parameters (--auto, --ep, -\n-op, --maxiterate, --large) (aligned_sequences_mafft.fasta in the supplementary data per \nphylum). Operational Taxonomic Units (OTUs) were then delineated from the alignment \nusing the script 3_generate_OTU.py. In particular, VSEARCH clustering at 97% identity \nwas employed to group sequences into OTUs. We set the clustering threshold at 97 % \nidentity (i.e., a 3 % divergence cut -off) because COI barcoding studies of Amoebozoa \n(43), Cercozoa (49) and diverse animal taxa report similar values (12). Although each \nlineage can display its own diversification rate, published barcoding gaps generally fall \nbetween 2 % and 3 %, so we applied a uniform 3 % threshold across all phyla. \nRepresentative ASV (centroids) for each OTU were saved, and a mapping of each ASV \nto its OTU was generated as a .uc file ( otus.uc in the supplementary data per phylum). \nThe .uc file was subsequently processed to produce a tabular mapping file \n(otus_mapping.txt in the supplementary data per phylum) that records the assignment of \neach ASV ID to its corresponding OTU.  \n \nMolecular and geographic distance Matrix Construction and OTU Filtering \nAfter obtaining the OTUs, the relationship between genetic and geographical distances \namong ASVs was analysed using the R script \n4_1_generate_molecular_and_geographic_distances.R. We used the Biostrings R \npackage to convert the phylum-level ASVs alignments into DNAbin objects, suitable for \ndistance calculations. Genetic distance matrices were computed under the K80 model \n(Kimura 2 -parameter) (50), with gaps (insertions/deletions) excluded using the \npairwise.deletion method. In parallel, a matrix of geographical distances between \nsampling localities was calculated from the coordinate data using the \ndistVincentyEllipsoid function of the geosphere R package. The genetic and geographical \ndistance matrices were both converted to long format and merged into a single data frame, \nso that each row contained a pairwise genetic distance, the corresponding geographical \ndistance, an d the associated OTU (and phylum). To ensure robust analysis, we first \nfiltered the data to include only OTUs represented by at least four independent ASVs. \nNext, we further filtered these OTUs, called the informative OTUs, to retain only those \nin which at least one pair of sequences had a genetic distance ≥ 0.01 and at least one pair \nhad a geographical distance ≥ 1 m. The ratio of the number of total OTUs and filtered \ninformative OTUs per phylum can be downloaded from \nratio_total_OTUs_informative_OTUs_by_phylum.csv in the supplementary data, and \nthe ID of the informative OTUs in the file informative_OTUs.txt of each phylum. This \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\ntwo-step filtering removed OTUs with very few sequences or with negligible variation, \nwhich could otherwise introduce noise. Metabarcoding studies that uses general primers \nlack targeted taxonomic resolution and often retrieve fragmented sequences or sequences \nfrom multiple species, potentially biasing interpretations (14). Filtering out OTUs with \nuniformly low molecular divergence could mitigates these issues. While this approach \nmight also exclude some genuine low genetic diversity  cases (for example, populations \nthat have undergone bottlenecks or are in decline), it pri marily reduces spurious OTUs \narising from methodological biases, improving the reliability of biogeographic patterns \ninterpretations.  \nThe results of each analyses of each informative OTUs were compiled into the file \ninformative_OTUs_results.csv, in the supplementary data. In this file, each row \ncorresponds to a informative OTU and each column the results of the analysis or metrics, \nsuch as: Phylum (the taxonomic phylum of that OTU), OTU (the OTU identifier, unique \nwithin each phylum), Max_Distance (the maximum geographical distance, in metres, \nbetween any two localities), Mean_Distance (the mean geographical distance among all \nlocality pairs), Std_Dev_Distance (standard deviation of geographical distances among \nlocalities) and Count (number of pairwise ASV comparisons, genetic/geographical \ndistance pairs). \n \nWithin-OTU Diversity and Tajima’s D \nTo quantify genetic diversity within each informative OTU, we calculated the nucleotide \ndiversity (π). This was done using the previously generated alignments per phylum: the \nnuc.div function (51) was applied with pairwise.deletion = TRUE to compute π for each \nOTU using the script 9_calculate_nucleotide_diversity_by_otu.R. The resulting values \nwere recorded in the Nucleotide_Diversity_Pi column of \ninformative_OTUs_results.csv. We also calculated Tajima’s D for each informative \nOTU to investigate potential signals of selection or demographic history (e.g. recent \nexpansion, bottleneck) using the script 10_calculate_tajimasD_by_otu.R. The tajima.test \nfunction (51) was applied to the alignment data for each informative OTU. In theory, a \nnegative Tajima’s D indicates an excess of rare variants consistent with recent population \nexpansion or purifying selection, whereas a positive Tajima’s D suggests a deficiency of \nrare variants, which may indicate a population bottleneck or balancing selection (52). \nGiven our filtering criteria (which ensured multiple sequences per OTU and some level \nof genetic divergence), we expected Tajima’s D values to trend negative. Each OTU’s \nTajima’s D value was recorded in a Tajima_s_D column, with its associated p -values \nrecorded in P_Value_Normal (normal approximation) and P_Value_Beta (beta \ndistribution approximation) columns. Subsequently, we examined whether nucleotide \ndiversity and Tajima’s D varied significantly among phyla. (For Tajima’s D, this \ncomparison was restricted to OTUs with a Beta p -value < 0.05) We first checked the \ndistribution of each variable: the Shapiro -Wilk test indicated that the data were not \nnormally distributed. Therefore, a non -parametric Kruskal -Wallis test (using the \nkruskal.test function in R) was employed to test for differences among phyla. A \nsignificant Kruskal-Wallis p-value was taken as evidence of variation among phyla for \nthe given metric. In cases of significance, post-hoc pairwise comparisons were performed \nusing Dunn’s test (dunnTest function in R) with a Bonferroni adjustment for multiple \ntesting (53). The results of the pairwise comparisons for each variable were saved to \nsignificant_group_comparisons_Dunn_test.csv (available in the Supplementary Data). \nThe file contains the columns Z (Z-Statistic), P_unadj (Uncorrected p-value) and P_adj \n(Adjusted p-value) for every variable in the phyla pairwise comparison. Phyla with an \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\ninsufficient number of informative OTUs were excluded from these comparative analyses \n(Filasterea, Hemichordata, Nibbleridia, Prasinodermophyta and Ustilaginomycotina). \nAdditionally, for infer differences between and within groups, phyla were grouped into \nthree broad categories (Archaeplastida, Metazoa, and all other phyla) to compare the \nprevalence of significant differences between and within these groups \n(significant_group_comparations_Dunn_test.csv in the supplementary data). The \nsame statistical approach was applied in later analyses for other variables and metrics of \ninterest, as described below. \n \nRead Count Filtering Impact \nIt is common in metabarcoding studies to discard ASVs with very low read counts (below \na certain threshold) to avoid spurious ASVs (54). To evaluate how such filtering might \naffect the recovery of low -abundance haplotypes and OTUs, we characterised the \noccurrence of rare haplotypes within each OTU. First, for each informative OTU, the total \nnumber of ASVs and the total number of reads was recorded in the Num_Sequences and  \nTotal_Reads columns, respectively in the informative_OTUs_results.csv file. In \naddition, separate columns captured the number of reads contributed by each ocean \nnamed <ocean>_reads. We then evaluate the potential loss of data and in reads counts \nfiltering by counting the number of haplotypes with fewer than 10 reads for each \ninformative OTU, recorded in the haplotypes_lt10 column. We also determined how \nmany of these rare haplotypes were supported by at least two independent ASVs (each \nwith <10 reads); this number was recorded in haplotypes_recovered_multiple column. \nBy requiring multiple independent ASVs, we can identify rare haplotypes that are \nconsistently observed, thereby reducing the likelihood of counting sequencing artifacts as \ntrue haplotypes. We also assessed whether the low -read filter could inadver tently affect \nmore abundant haplotypes. For each OTU, we counted the number of haplotypes with \ntotal reads > 10 that nevertheless contained at least one constituent ASV with <10 reads. \nThis count was recorded in haplotypes_gt10_with_lt10_seq column. Finally, to examine \nthe biogeographic distribution of low -abundance ASVs, we counted how many oceans \nhad a total read count < 10 for each OTU. This value (the number of distinct oceans in \nwhich the OTU is represented by fewer than 10 reads) was re corded in the global_lt10 \ncolumn. Given that metabarcoding studies are typically constrained to specific \nregions/oceans, any potential “tag jumping” across oceans (a form of cross -sample \ncontamination) can be ruled out at this stage. Overall, these metrics allowed us to \nunderstand the extent to which stringent read filtering might exclude genuine but low -\nfrequency haplotypes and the veracity of ASVs with low number of reads. \n \nGeographic and genetic distances \nLinear regression and non -linear analyses were conducted for each informative OTU to \nevaluate the relationship, strength and statistical significance of the correlation between \ngenetic and geographic distances using the R scripts \n5_calculate_linear_models_by_otu.R and 6_calculate_nls_models_by_otu.R, \nrespectively. The pa rameters of the linear model (lm) were calculated for each \ninformative OTU using the dplyr R package, and the results were saved in the file \ninformative_OTUs_results.csv, with the Columns: Intercept_lm (intercept), Slope_lm \n(slope), R2_lm (coefficient of determination), P_value _lm (p-value of the slope), \nAIC_lm (Akaike Information Criterion), and BIC_lm (Bayesian Information Criterion). \nAdditionally, a non -linear regression analysis (nls) was performed using the model: \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n“Genetic Distance” = a * log(“Geographical Distance” + 1) + b. Parameters a and b were \nestimated from predefined initial values. The output included the a_nls (slope parameter), \nb_nls (vertical offset parameter), RSS_nls (residual sum of squares), R2_nls (coefficient \nof determination), AIC_nls (Akaike Information Criterion), and BIC_nls (Bayesian \nInformation Criterion). To interpret the differences in AIC and BIC between these two \nmodels, categories were established based on the magnitude of the differences (Δ): (1) no \nevidence to prefer either model (|Δ| ≤ 2), indicating similar fit; (2) moderate evidence (2 \n< |Δ| ≤ 7), indicating moderate support for one model; and (3) strong evidence (| Δ| > 7), \nindicating strong support for one model ( models_comparison_summary.csv file in \nsupplementary data). Based on these results we continued with the non-linear models for \nthe graphical representations, as it better explained the genetic and molecular distance \ncorrelations per phylum. \nNext, the linear and non-linear models were compared with  null models to assess whether \nthe relationship between genetic and geographic distances was consistent with a \nstochastic pattern. The null models were generated by simulating random distributions of \ngenetic and geographic distances using the Python script \n7_calculate_null_models_by_otu.py, fitting linear models with the scikit -learn library \nand optimising non-linear models with SciPy’s curve_fit function. For each OTU, 1,000 \nsimulations were performed. In each simulation, the number of data points equalled the \nnumber of pairwise comparisons (count column in informative_OTUs_results.csv file) \nfor that OTU. Geographical distances were sampled uniformly between 0 and the \nmaximum recorded distance ( max_distance column), and genetic distances were \nsampled uniformly between 0 and 0.03 (the specified OTU cutoff). From these \nsimulations, the mean, median and standard deviation of R^2, AIC and BIC were \ncalculated for comparison with the observed models. Results were saved in \ninformative_OTUs_results.csv file with the following columns: Mean_R2_lm_null, \nMedian_R2_lm_null, Std_R2_lm_null, Mean_AIC_lm_null, Median_AIC_lm_null, \nStd_AIC_lm_null, Mean_BIC_lm_null, Median_BIC_lm_null, and \nStd_BIC_lm_null; and for non -linear models: Mean_R2_nls_null, \nMedian_R2_nls_null, Std_R2_nls_null, Mean_AIC_nls_null, \nMedian_AIC_nls_null, Std_AIC_nls_null, Mean_BIC_nls_null, \nMedian_BIC_nls_null, and Std_BIC_nls_null, Mean_a_nls_null, \nMedian_a_nls_null, Std_a_nls_null, Mean_b_nls_null, Median_b_nls_null, and \nStd_b_nls_null. P -values were calculated to determine whether the observed \nrelationships differed significantly from random expectations, based on the coefficient of \ndetermination, and were saved in the columns P_R2_lm_null (linear models) and \nP_R2_nls_null (non-linear models). \nTo further examine whether a correlation exists between genetic and geographic distances \nand to assess its consistency across phyla, Mantel tests were performed using the R script \n8_calculate_mantel_test_by_otu.R. For each OTU, genetic and geographic dista nce \nmatrices were generated, and the Mantel test was conducted using the mantel function \nfrom the vegan package (version 2.6) (55) with 999 permutations, the results were \nrecovered in the columns Mantel_R and P_Value_Mantel. However, the Mantel test can \nbe problematic due to inherent dependencies in the data (56). In this context, the \npermutations used to calculate the p -value may not adequately reflect the expected null \ndistribution under true independence; therefore, this results should be interpreted with \ncaution. After completing these analyses, only OTUs with  a Mantel test p -value < 0.05 \nwere retained, and comparisons between phyla were conducted as described above. \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nBiogeography Based on Haplotype Networks \nHaplotype network construction and diversity metrics \nUnique haplotypes were first identified from the alignment for each informative OTU \nusing the haplotype function (51). The number of haplotypes per informative OTU was \nthen recorded in a column named num_haplotypes. Next, using the haploNet function \n(51), haplotype networks based on Minimum Spanning Networks (MSNs) were \nconstructed for each informative OTU and used in subsequent analyses. We then calculate \nthe Haplotype diversity (Hd), branch diversity (Bd) and their combined metric (Hbd) (23) \nusing R script 11_calculate_haplotype_diversity_by_otu.R. We saved the results in \ncolumns named Hd, Bd and HBd, respectively, in the informative_OTUs_results.csv \nfile. Additionally, the maximum value of the squared haplotype frequencies per OTU was \nrecorded in a column Efh2_Hd. \nThe above metrics treat each ASV equally, regardless of its abundance (i.e. they are based \nonly on ASV presence/absence within each haplotype). To account for the relative \nabundance of each ASV (as a proxy for its prevalence in the sample), we incorporated the \nassociated read counts. Specifically, we used two metrics (1) the total reads count for each \nASV and (2) a logarithmic transformation, (log(reads+1)), using the R script \n12_calculate_weighted_haplotype_diversity_by_otu.R. By weighting sequences \naccording to their total read counts and a logarithmic transformation, haplotypes with \nhigher abundance have a greater influence on the calculated diversity metrics. Therefore, \nif a sequence-based diversity metric exceeds its read-weighted counterpart, this indicates \nthat rarer haplotypes disproportionately influence the haplotypic diversity (i.e. diversity \nis overestimated in the unweighted case). This approach is particularly useful for \ncommunities with high variability in abundances, as it allows patterns of d ominance or \ninequality to emerge. However, it may underestimate diversity in OTUs that contain many \nlow-abundance haplotypes. The results of these abundance -weighted calculations were \nrecorded in the following columns: Hd_reads and Hd_log_reads (haplotype diversity); \nBd_reads and Bd_log_reads (branch diversity); Hbd_reads and Hbd_log_reads \n(combined haplotype and branch diversity); and Efh2_reads and Efh2_log_reads \n(maximum squared haplotype frequencies).  \n \nOceanic genetic structure  \nAfter calculating haplotype diversity metrics, we evaluated genetic structuring between \nand within populations (each corresponding to a different ocean). Analysis of Molecular \nVariance (AMOVA) was performed for each OTU using R script \n13_calculate_amova_by_otu.R. OTUs were selected for AMOVA based on the following \ncriteria: (i) Presence in at least two different oceans; (ii) At least two populations; and \n(iii) At least two distinct haplotypes within those populations. We used the popr.amova \nfunction to perform the AMOVA. Variance components and sums of squares were \ncalculated for between-population and within-population comparisons. The results were \nsaved in the informative_OTUs_results.csv file, in columns named \nBetween_samples_SumSq, Within_samples_SumSq, Total_SumSq, \nBetween_samples_Sigma, Within_samples_Sigma and Total_Sigma. These metrics \nwere then compared across phyla as described above. Because the sequence length is \nrelatively short (approximately 313 bp), the AMOVA results should be interpreted with \ncaution.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nTo mitigate potential biases introduced by short sequences, we evaluated genetic \nstructuring among oceans using the haplotype frequencies and distributions for each \ninformative OTU. The haplotypic diversity metrics total diversity (Ht), within-population \ndiversity (Hs), genetic differentiation coefficient (Fst) and gene flow (Nm) were calculated \nfor each OTU using script 14_calculate_Hs_fst_nm_by_otu.R. These four metrics were \nstored in columns Ht, Hs, Fst, and Nm; and comparisons between phyla were conducted \nas described above. \nA modified haplotype diversity metric (PH d) was created to infer the distribution of \nhaplotypes among populations (here defined by different oceans) using script \n15_calculate_PHd_and_global_ocean_dominance_by_otu.R. This new metric \nincorporates the relative distribution of haplotypes across populations. By using weighted \nrelative haplotype frequencies, PH d quantifies the contribution of population -level \ndifferences to the genetic diversity of each OTU. The classical haplotype diversity \nformula was adapted to account for the distribution of haplotypes across populations. In \nthis adaptation, the relative frequency of each haplotype is redefined as its frequency in a \ngiven population divided by its total frequency in the OTU. The modified PHd is therefore \ncalculated as: \n𝑃𝐻𝑑 = 1 − ∑\n𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛(𝑜𝑐𝑒𝑎𝑛𝑠)\n(𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 ℎ𝑎𝑝𝑙𝑜𝑡𝑦𝑝𝑒𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛\n𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 ℎ𝑎𝑝𝑙𝑜𝑡𝑦𝑝𝑒𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑂𝑇𝑈 ) ^2 \n \nThe PHd values were saved in a column named PHd. The relative haplotype frequencies \nfor each ocean were stored in columns named <ocean>_sqfreq. These PHd metrics were \nthen compared across phyla using the same procedure as for the other diversity metrics. \nFurthemore we compare PHd with Hs metric to helps to identify whether the populations \nof an informative OTU are genetically structured or relatively homogeneous across \noceans. If PHd exceeds Hs (a positive difference), this indicates that the populations have \ndistinct genetic compositions and are partially isolated from one another, reflecting \nsignificant genetic structuring and no single population dominating the haplotype \ndiversity. Conversely , if PH d is less than H s (a negative difference), one or a few \npopulations account for most of the global haplotypic diversity in the OTU, suggesting \nthat these populations dominate the haplotypic diversity. \nFor each informative OTU, we identified the ocean with the highest haplotype relative \ncontribution (the maximum sqfreq value) as a proxy for the putative population of source \nand vice versa for recent introduction populations. However, this analysis is considered a \npreliminary proxy to guide further investigation. To distinguish likely source populations \nfrom recent introductions, we calculated ratios of relative haplotype frequencies between \noceans using script 20_introductions_vs_dominant_oceans_byOTU.R. Specifically, \nwithin each OTU we used the ocean with the highest relative frequency as the reference \n(ratio = 1), and we computed for each other ocean the ratio of the reference frequency to \nthat ocean’s frequency. We then examined the distribution of thes e ratios across all \ninformative OTUs to detect shifts toward lower values. A notable change in the \ndistribution occurred at a ratio of 0.25. Consequently, for each OTU we classified any \nocean with a ratio below 0.25 as a “recently introduced ocean/populati on,” since such \noceans have very low relative haplotype frequency compared to the highest haplotypic \nfrequency ocean. Furthermore, to identify a clear source ocean for each informative OTU, \nwe applied additional filtering criteria. We selected OTUs in whic h exactly one ocean \nhad a ratio of 1.0 (i.e. a single ocean had the maximum relative frequency) and excluded \ninformative OTUs where multiple oceans shared this maximum frequency, since that \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nwould preclude a clear source. We also excluded informative OTUs in which any other \nocean had a ratio exceeding 0.5, to avoid ambiguity. Although these thresholds are \nsomewhat arbitrary, they provide a more restrictive criterion that allows a more reliable  \ninference of the putative source ocean for the remaining informative OTUs. We create a \npie chart graph of the proportion of dominant oceans per phylum ancestral_oceans.pdf. \nFurthermore, we combine the number of OTUs dominant and introduced oceans in the \nfile dominant_introduced_oceans.csv. \nTo validate this proxy for source (dominant) and introduced ocean populations, we used \nphylum Chordata taxonomically annotated informative OTUs as a positive control. We \nincluded only those informative OTUs that had at least one ocean classified as “recent ly \nintroduced”. For each selected informative OTU, the first amplicon sequence variant \n(ASV) was compared to entries in the NCBI database using BLAST (57). We retained \nthose OTUs for which the BLAST hit showed ≥ 97% sequence identity to a species in the \ndatabase, ensuring that species -level identification was feasible. Taxonomic annotation \nresults for these OTUs are recorded in file introduced_chordata_otus.csv. \nFor each informative OTU, we inferred connectivity between different oceans from its \nhaplotype network to evaluate genetic dispersal and exchange. This analysis was \nperformed using the script 16_generate_ocean_connectivity_by_otu.R. In each haplotype \nnetwork, nodes represent haplotypes and edges represent connections between them. We \nannotated each edge with the ocean of origin for each haplotype in the pair. Only \nconnections linking haplotypes  from different oceans were retained; intra -oceanic \nconnections were discarded because they do not inform inter-ocean connectivity. We then \naggregated the data to count the number of connections between each pair of oceans. For \neach informative OTU, we recorded the total number of oceanic haplotypic connections \nin the column Total_Connections and the number of connections per ocean in columns \nnamed <ocean>_connections. Then, we classified oceanic pairs as “nearby” if they are \ngeographically adjacent and “distant” if they are geographically separated using the script \n17_calculate_ocean_conectivity.R, the number of connections were recovered in the \ncolumns nearby_oceans_connections and distant_oceans_connections. OTUs in \nwhich all haplotypes occur in the same ocean therefore have only “nearby” connectivity \nand no “distant” connectivity. We also characterised intra -hemispheric and inter -\nhemispheric connectivity, focusing on connections between oceans in the Northe rn and \nSouthern Hemispheres, recovered in the columns Inter-Hemisferica and Intra-\nHemisferica. \nFor each phylum, we generated network graphs illustrating connections among oceans \nusing the script 18_oceanic_network_by_phyla.py and 19_global_oceanic_network.py. \nIn these graphs, each node represents an ocean and each edge represents the total number \nof connections between two oceans, weighted by relative strength. We summed all \nconnections across OTUs for each phylum and applied a log1p transformation (the natural \nlogarithm of one plus this sum) to normalise the data. We then applied the Louvain \ncommunity detection algorithm to identify clusters (communities) within each network, \ncolouring nodes by their community membership. The networks were constructed using \nthe networkx library, with edge weights reflecting connection strength the network per \nphylum were recovered as network_<phylum>.pdf and .jpg. \nFinally, we analysed correlations among the different calculated variables for each \ninformative OTU. We computed correlation matrices and associated p -values using \nPearson’s correlation coefficient, generating a table listing only those correlations that \nwere statistically significant (p < 0.05) significant_correlations.csv file. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \n \nReferences and Notes \n1.  J. Whitfield, Is Everything Everywhere? Science 310, 960–961 (2005). \n2.  L. G. M. B. Becking, Geobiologie of inleiding tot de milieukunde (W.P. Van Stockum & Zoon, 1934). \n3.  E. Pinseel, K. Sabbe, E. Verleyen, W. Vyverman, A New Dawn for Protist Biogeography. Global \nEcology and Biogeography 33, e13925 (2024). \n4.  C. de Vargas, S. Audic, N. Henry, J. Decelle, F. Mahé, R. Logares, E. Lara, C. Berney, N. Le Bescot, \nI. Probert, M. Carmichael, J. Poulain, S. Romac, S. Colin, J. -M. Aury, L. Bittner, S. Chaffron, M. \nDunthorn, S. Engelen, O. Flegontova, L. Guidi, A. Ho rák, O. Jaillon, G. Lima -Mendez, J. Lukeš, S. \nMalviya, R. Morard, M. Mulot, E. Scalco, R. Siano, F. Vincent, A. Zingone, C. Dimier, M. Picheral, \nS. Searson, S. Kandels-Lewis, Tara Oceans Coordinators, S. G. Acinas, P. Bork, C. Bowler, G. Gorsky, \nN. Grimsley, P. Hingamp, D. Iudicone, F. Not, H. Ogata, S. Pesant, J. Raes, M. E. Sieracki, S. Speich, \nL. Stemmann, S. Sunagawa, J. Weissenbach, P. Wincker, E. Karsenti, Eukaryotic plankton diversity \nin the sunlit ocean. Science 348, 1261605 (2015). \n5.  G. Sommeria-Klein, R. Watteaux, F. M. Ibarbalz, J. J. Pierella Karlusich, D. Iudicone, C. Bowler, H. \nMorlon, Global drivers of eukaryotic plankton biogeography in the sunlit ocean. Science 374, 594–\n599 (2021). \n6.  P. C. Junger, H. Sarmento, C. R. Giner, M. Mestre, M. Sebastián, X. A. G. Morán, J. Arístegui, S. \nAgustí, C. M. Duarte, S. G. Acinas, R. Massana, J. M. Gasol, R. Logares, Global biogeography of the \nsmallest plankton across ocean depths. Sci. Adv. 9, eadg9763 (2023). \n7.  G. S. Bradburd, P. L. Ralph, Spatial Population Genetics: It’s About Time. Annual Review of Ecology, \nEvolution, and Systematics 50, 427–449 (2019). \n8.  L. J. Rissler, Union of phylogeography and landscape genetics. Proceedings of the National Academy \nof Sciences 113, 8079–8086 (2016). \n9.  S. Manel, M. K. Schwartz, G. Luikart, P. Taberlet, Landscape genetics: combining landscape ecology \nand population genetics. Trends in Ecology & Evolution 18, 189–197 (2003). \n10.  J. C. Avise, J. Arnold, R. M. Ball, E. Bermingham, T. Lamb, J. E. Neigel, C. A. Reeb, N. C. Saunders, \nINTRASPECIFIC PHYLOGEOGRAPHY: The Mitochondrial DNA Bridge Between Population \nGenetics and Systematics. Annual Review of Ecology, Evolution, and Systematics 18, 489–522 (1987). \n11.  P. Taberlet, E. Coissac, M. Hajibabaei, L. H. Rieseberg, Environmental DNA: ENVIRONMENTAL \nDNA. Molecular Ecology 21, 1789–1793 (2012). \n12.  P. D. N. Hebert, S. Ratnasingham, J. R. de Waard, Barcoding animal life: cytochrome c oxidase subunit \n1 divergences among closely related species. Proc. R. Soc. Lond. B 270 (2003). \n13.  R. González-Miguéns, A. Gàlvez-Morante, M. Skamnelou, M. Antó, E. Casacuberta, D. J. Richter, D. \nVaulot, J. del Campo, I. Ruiz -Trillo, A Novel Taxonomic Database for eukaryotic Mitochondrial \nCytochrome Oxidase subunit I Gene (eKOI): Enhancing taxonomi c resolution at community-level in \nmetabarcoding analyses. bioRxiv [Preprint] (2024). https://doi.org/10.1101/2024.12.05.626972.  \n14.  M. Couton, F. Viard, F. Altermatt, Opportunities and inherent limits of using environmental DNA for \npopulation genetics. Environmental DNA 5, 1048–1064 (2023). \n15.  Materials and methods and additional supplementary text sections are available as supplementary \nmaterials. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n16.  K. J. Gaston, K. J. Gaston, The Structure and Dynamics of Geographic Ranges  (Oxford University \nPress, Oxford, New York, 2003)Oxford Series in Ecology and Evolution. \n17.  N. D. Tsutsui, A. V. Suarez, D. A. Holway, T. J. Case, Reduced genetic variation and the success of \nan invasive species. Proceedings of the National Academy of Sciences 97, 5948–5953 (2000). \n18.  C. E. Lee, Evolutionary genetics of invasive species. Trends in Ecology & Evolution  17, 386–391 \n(2002). \n19.  J. Roman, J. A. Darling, Paradox lost: genetic diversity and the success of aquatic invasions. Trends \nin Ecology & Evolution 22, 454–464 (2007). \n20.  D. Posada, K. A. Crandall, Intraspecific gene genealogies: trees grafting into networks. Trends in \nEcology & Evolution 16, 37–45 (2001). \n21.  D. H. Huson, D. Bryant, Application of Phylogenetic Networks in Evolutionary Studies. Molecular \nBiology and Evolution 23, 254–267 (2006). \n22.  M. Nei, Molecular Evolutionary Genetics (Columbia University Press, New York, 1987). \n23.  E. Garcia, D. Wright, R. Gatins, M. B. Roberts, H. T. Pinheiro, E. Salas, J. -Y. Chen, J. R. Winnikoff, \nG. Bernardi, Haplotype network branch diversity, a new metric combining genetic and topological \ndiversity to compare the complexity of haplotype networks. PLOS ONE 16, e0251878 (2021). \n24.  B. S. Weir, C. C. Cockerham, Estimating F -Statistics for the Analysis of Population Structure. \nEvolution 38, 1358–1370 (1984). \n25.  M. Slatkin, Gene Flow in Natural Populations. Annual Review of Ecology, Evolution, and Systematics \n16, 393–430 (1985). \n26.  J. T. Carlton, Biological Invasions and Cryptogenic Species. Ecology 77, 1653–1655 (1996). \n27.  J. L. Molnar, R. L. Gamboa, C. Revenga, M. D. Spalding, Assessing the global threat of invasive \nspecies to marine biodiversity. Frontiers in Ecology and the Environment 6, 485–492 (2008). \n28.  B. W. Bowen, M. R. Gaither, J. D. DiBattista, M. Iacchei, K. R. Andrews, W. S. Grant, R. J. Toonen, \nJ. C. Briggs, Comparative phylogeography of the ocean planet. Proceedings of the National Academy \nof Sciences 113, 7962–7969 (2016). \n29.  S. Katsanevakis, I. Wallentinus, A. Zenetos, E. Leppäkoski, M. E. Çinar, B. Oztürk, M. Grabowski, \nD. Golani, A. C. Cardoso, Impacts of invasive alien marine species on ecosystem services and \nbiodiversity: a pan-European review. AI 9, 391–423 (2014). \n30.  N. Streftaris, A. Z. & E. Papathanassiou, “Globalisation in Marine Ecosystems: The Story of Non -\nIndigenous Marine Species Across European Seas” in Oceanography and Marine Biology (CRC Press, \n2005). \n31.  J. D. DiBattista, J. Howard Choat, M. R. Gaither, J. -P. A. Hobbs, D. F. Lozano -Cortés, R. F. Myers, \nG. Paulay, L. A. Rocha, R. J. Toonen, M. W. Westneat, M. L. Berumen, On the origin of endemic \nspecies in the Red Sea. Journal of Biogeography 43, 13–30 (2016). \n32.  M. Coll, C. Piroddi, J. Steenbeek, K. Kaschner, F. B. R. Lasram, J. Aguzzi, E. Ballesteros, C. N. \nBianchi, J. Corbera, T. Dailianis, R. Danovaro, M. Estrada, C. Froglia, B. S. Galil, J. M. Gasol, R. \nGertwagen, J. Gil, F. Guilhaumon, K. Kesner -Reyes, M.-S. Kitsos, A. Koukouras, N. Lampadariou, \nE. Laxamana, C. M. L.-F. de la Cuadra, H. K. Lotze, D. Martin, D. Mouillot, D. Oro, S. Raicevich, J. \nRius-Barile, J. I. Saiz -Salinas, C. S. Vicente, S. Somot, J. Templado, X. Turon, D. Vafidis, R. \nVillanueva, E. V oultsiadou, The Biodiversity of the Mediterranean Sea: Estimates, Patterns, and \nThreats. PLOS ONE 5, e11842 (2010). \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n33.  C. Capinha, F. Essl, H. Seebens, D. Moser, H. M. Pereira, The dispersal of alien species redefines \nbiogeography in the Anthropocene. Science 348, 1248–1251 (2015). \n34.  H. Seebens, N. Schwartz, P. J. Schupp, B. Blasius, Predicting the spread of marine species introduced \nby global shipping. Proceedings of the National Academy of Sciences 113, 5646–5651 (2016). \n35.  S. P. Lawler, P. J. Morin, Food Web Architecture and Population Dynamics in Laboratory Microcosms \nof Protists. The American Naturalist 141, 675–686 (1993). \n36.  A. McMinn, A. Martin, Dark survival in a warming world. Proceedings of the Royal Society B: \nBiological Sciences 280, 20122909 (2013). \n37.  H. Wu, C. Shen, Q. Wang, R. B. Aronson, C. Chen, J. Xue, Survivorship characteristics and adaptive \nmechanisms of phytoplankton assemblages in ballast water. J. Ocean. Limnol. 37, 580–588 (2019). \n38.  A. Ardura, J. Rick, J. L. Martinez, A. Zaiko, E. Garcia -Vazquez, Stress resistance for unraveling \npotential biopollutants. Insights from ballast water community analysis through DNA. Marine \nPollution Bulletin 163, 111935 (2021). \n39.  O. Folmer, M. Black, W. Hoeh, R. Lutz, R. Vrijenhoek, DNA primers for amplification of \nmitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Mol Mar Biol \nBiotechnol 3, 294–299 (1994). \n40.  J. Geller, C. Meyer, M. Parker, H. Hawk, Redesign of PCR primers for mitochondrial cytochrome c \noxidase subunit I for marine invertebrates and application in all-taxa biotic surveys. Molecular Ecology \nResources 13, 851–861 (2013). \n41.  M. Leray, J. Y. Yang, C. P. Meyer, S. C. Mills, N. Agudelo, V. Ranwez, J. T. Boehm, R. J. Machida, \nA new versatile primer set targeting a short fragment of the mitochondrial COI region for \nmetabarcoding metazoan diversity: application for characterizing coral reef fish gut contents. Frontiers \nin Zoology 10, 34 (2013). \n42.  O. S. Wangensteen, C. Palacín, M. Guardiola, X. Turon, DNA metabarcoding of littoral hard -bottom \ncommunities: high diversity and database gaps revealed by two molecular markers. PeerJ 6, e4705 \n(2018). \n43.  R. González-Miguéns, E. Cano, A. Guillén-Oterino, A. Quesada, D. J. G. Lahr, D. Tenorio-Rodríguez, \nD. de Salvador-Velasco, D. Velázquez, M. I. Carrasco -Braganza, R. T. Patterson, E. Lara, D. Singer, \nA needle in a haystack: A new metabarcoding approach  to survey diversity at the species level of \nArcellinida (Amoebozoa: Tubulinea). Molecular Ecology Resources n/a (2023). \n44.  M. Martin, Cutadapt removes adapter sequences from high -throughput sequencing reads. \nEMBnet.journal 17, 10–12 (2011). \n45.  B. J. Callahan, P. J. McMurdie, M. J. Rosen, A. W. Han, A. J. A. Johnson, S. P. Holmes, DADA2: \nHigh-resolution sample inference from Illumina amplicon data. Nat Methods 13, 581–583 (2016). \n46.  P. J. A. Cock, T. Antao, J. T. Chang, B. A. Chapman, C. J. Cox, A. Dalke, I. Friedberg, T. Hamelryck, \nF. Kauff, B. Wilczynski, M. J. L. de Hoon, Biopython: freely available Python tools for computational \nmolecular biology and bioinformatics. Bioinformatics 25, 1422–1423 (2009). \n47.  T. Rognes, T. Flouri, B. Nichols, C. Quince, F. Mahé, VSEARCH: a versatile open source tool for \nmetagenomics. PeerJ 4, e2584 (2016). \n48.  K. Katoh, K. Misawa, K. Kuma, T. Miyata, MAFFT: a novel method for rapid multiple sequence \nalignment based on fast Fourier transform. Nucleic Acids Research 30, 3059–3066 (2002). \n49.  R. González-Miguéns, C. Soler-Zamora, F. Useros, S. Nogal- Prata, C. Berney, A. Blanco-Rotea, M. \nI. Carrasco -Braganza, D. Salvador -Velasco, A. Guillén - Oterino, D. Tenorio -Rodríguez, D. \nVelázquez, T. J. Heger, I. Sanmartín, E. Lara, Cyphoderia ampulla (Cyp hoderiidae: Rhizaria), a tale \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nof freshwater sailors: The causes and consequences of ecological transitions through the salinity barrier \nin a family of benthic protists. Molecular Ecology (2022). \n50.  M. Kimura, A simple method for estimating evolutionary rates of base substitutions through \ncomparative studies of nucleotide sequences. J Mol Evol 16, 111–120 (1980). \n51.  E. Paradis, pegas: an R package for population genetics with an integrated –modular approach. \nBioinformatics 26, 419–420 (2010). \n52.  F. Tajima, Statistical Method for Testing the Neutral Mutation Hypothesis by DNA Polymorphism. \nGenetics 123, 585–595 (1989). \n53.  O. J. Dunn, Multiple Comparisons Among Means. Journal of the American Statistical Association 56, \n52–64 (1961). \n54.  N. A. Bokulich, S. Subramanian, J. J. Faith, D. Gevers, J. I. Gordon, R. Knight, D. A. Mills, J. G. \nCaporaso, Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. \nNat Methods 10, 57–59 (2013). \n55.  P. Dixon, VEGAN, a package of R functions for community ecology. Journal of Vegetation Science \n14, 927–930 (2003). \n56.  P. Legendre, M.-J. Fortin, D. Borcard, Should the Mantel test be used in spatial analysis? Methods in \nEcology and Evolution 6, 1239–1247 (2015). \n57.  S. F. Altschul, W. Gish, W. Miller, E. W. Myers, D. J. Lipman, Basic local alignment search tool. \nJournal of Molecular Biology 215, 403–410 (1990). \n \nAcknowledgments: we express our gratitude to A. Berlinches, E. Lara, G. Bercedo and \nM. Villar-DePablo and the MCG lab group for helpful discussions. Work funded \nby the European Union (ERC, MISSINGRELATIVES, 101097659 ). Views and \nopinions expressed are however those of the author(s) only and do not necessarily \nreflect those of the European Union or the European Research Council. Neither \nthe European Union nor the granting authority can be held responsible for them . \nWe also acknowledge support to Departament de Recerca i Universitats de la \nGeneralitat de Catalunya (exp. 2021 SGR 00751) and support by PIE -\n202120E047- Conexiones-Life, and MICIU/AEI /10.13039/501100011033  - \nFSE+ (JDC2023-050439-I). \n \nAuthor contributions:  \nConceptualization: RGM, IRT \nMethodology: RGM \nInvestigation: RGM \nVisualization: RGM \nFunding acquisition: RGM, IRT \nProject administration: RGM \nSupervision: EC, IRT \nDatabase maintenance: AGM, GT, CB \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\nWriting – original draft: RGM, IRT \nWriting – review & editing: RGM, AGM, GT, CB, EC, IRT \nCompeting interests: Authors declare that they have no competing interests. \n \nData and materials availability: All data, code, results, including eKOI metabarcoding \ndatabase, are available via Figshare (10.6084/m9.figshare.29144645). All other \ndata are available in the manuscript or the supplementary materials. \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. 1. eKOI metabarcoding database . ( A) Map showing the locations of samples \nincluded in the database. The size and color of the points represent the log(total reads) for \neach sample. To the left of the map, a scatter plot displays the relationship between \nlog(ASVs) and log(reads) for the sampled localities, colored according to log(reads). (B) \nrose plots representing the proportion of taxonomic annotated ASV as log(ASV) on the \ninner layer and log(reads) on the outer layer in the different eukaryotic phyla used in the \nanalysis. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. 2. Population structure between eukaryotic phyla. (A) Boxplots representing the \nPHd values for each OTU between phyla, where values close to 0 indicate low structuring \nin terms of intra-OTU haplotype diversity. Red circles with crosses represent the median \nHs values for each phyla. The number near each phyla name represents the number of \nOTU recovered after filtration. The numbers at the top indicate the difference between \nPHd and H s. ( B) PHd values grouped between different phyla. ( C) Number of recent \nintroduced oceans per OTU grouped between different phyla. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. 3. Number of recently introduced and dominant informative OTUs per ocean. \n(A) Boxplots showing the number of informative OTUs considered invasive in each \nocean and (B) the number of informative OTUs per ocean identified as their \ndominant/sink. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S1. Rare haplotypes (< 10 reads). Boxplot represents the proportion of haplotypes \nwith less than 10 total reads. The median is indicated by black diamonds, and the box \ncolors correspond to different taxonomic groups. Blue circles represent the median \nproportion of haplotypes with more than 10 reads but composed of at least one ASV \nwith less than 10 reads. Red squares represent the median proportion of oceans \ncomposed of less than 10 reads. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S2. Nucleotide diversity metrics. (A) Boxplot representing the values of Tajima’s \nD for OTUs grouped by phyla, with p -values less than 0.05. ( B)  Boxplot of nucleotide \ndiversity values for OTUs grouped by phyla. \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S3.  Geographic distances . The boxplots represents the maximum distribution \nbetween two localities per informative OTU. The red circles represents the median of the \nmean geographic distances between all the localities for each informative OTU. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S4. Parameter “a” of the nonlinear (exponential) model . Boxplots representing \nthe values of parameter “a” from the non -linear models for each informative OTU, \nrelating genetic and geographic distances. Orange points represent null models generated \nfor each phyla, with squares marking the median values. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S5. Significant Mantel test . Boxplots representing the R values from the Mantel \ntest, including only informative OTUs with a p-value less than 0.05. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S6. Haplotypic diversity metrics . (A) Boxplot representing haplotype diversity \nvalues, (B) branch diversity, and (C) haplotype-branch diversity, all based on the number \nof sequences. In all three cases, the median considering the number of sequences is \nrepresented by diamonds, the median based on total reads is represented by squares, and \nthe median using log(reads) is represented by circles.  \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S7. Population structure. (A) Boxplot representing the Fst values, and (B) boxplot \nof Hs values per informative OTU grouped by phyla. Circles represent the median of Nm \nvalues, and squares represent the medians for Ht values. \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S8. Characterization dominant and introduced ocean per OTU. (A) Distribution \nof haplotype relative frequency ratios between oceans of all informative OTUs. ( B) \nnumber of introduced oceans per informative OTU grouped by phyla. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint \n\n \nFig. S9. Ratios of haplotype oceanic connections. Bar plot representing the composition \nof the ratios for the different types of ocean connections, based on the haplotypic \nnetworks, for each phylum. The left side of the bar plot illustrates the proportion of \nconnections between inter- versus intra-hemispheric oceans, while the right side shows \nthe proportion of connections between separated versus adjacent oceans. \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 2, 2025. ; https://doi.org/10.1101/2025.06.02.657477doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}