Microbiome profiling is essential for understanding the complex living network within soil that plays an important role in agricultural systems (Mas-Carrió et al., 2018). It provides vital information on soil ecology (Chen et al., 2021) and helps in identifying microbiomes indicative of soil fertility and crop production, which is crucial for developing sustainable agriculture (Zhang et al., 2022).

Despite the importance of soil microbiome profiling, conventional techniques such as culturing and PCR-based methodologies face significant limitations. One major challenge is the high diversity and heterogeneity of soil microbial community (MC), which complicates the detection and monitoring of microbial responses to different management practices (Erlandson et al., 2024). Additionally, there is a lack of soil-specific reference databases available to classifiers, which hinders the accurate representation of the soil community complexity (Edwin et al., 2024).

Recent advancements in soil microbiome characterization have been driven by the development and application of high-throughput sequencing (HTS) techniques (Garg et al., 2024) which simplify description of MC composition in various habitats compared to traditional methods (Zhang G. et al., 2024; Zhang T. et al., 2024) and allow the development of modern indicators of soil biological quality (Djemiel et al., 2022). Traditional methods like culturing and Sanger sequencing are limited by biases and low throughput, often underrepresenting microbial diversity. HTS analyzes thousands to millions of DNA sequences in parallel, offering identification of both abundant and rare taxa, enabling assessment of functional potential and exploration of MC dynamics across spatial and temporal scales. Additionally, HTS reduces time and cost by automating processes, providing rapid, large-scale analysis of soil ecosystems, crucial for studying biogeochemical cycles, plant-microbe interactions, and environmental responses (Caporaso et al., 2012).

Amplicon sequencing of the 16S ribosomal RNA (16S rRNA) gene is considered a reliable and efficient method for taxonomic classification of MC and analyzing their associated characteristics (Yarza et al., 2014). Third-generation sequencing solutions, represented by PacBio and ONT, appear to be superior platforms compared to traditional methods such as Illumina and Sanger sequencing due to their ability to produce long reads, allowing for full-length 16S rRNA gene sequencing, which offers finer taxonomic resolution (Garg et al., 2024). Traditional short-read methods (100–400 bp), such as Illumina, usually target hypervariable regions, e.g., V3–V4 (Klindworth et al., 2013), which can lead to ambiguous taxonomic assignments (Callahan et al., 2019). Long-read sequencing overcomes this limitation and improves species-level identification.

PacBio platform, utilizing the circular consensus sequencing (CCS) model, is capable of fully covering the complete 16S rRNA sequences, providing high-resolution species-level identification with an exceptional accuracy exceeding 99.9% (Johnson et al., 2019). The high accuracy of PacBio reads is achieved through multiple passes of the same DNA molecule, which offers detailed insights into specific genomic regions (Garg et al., 2024). The existing data on ONT use in 16S rRNA sequencing is more controversial due to higher error rates compared to PacBio (Zhang et al., 2023). Meanwhile, recent modifications of ONT, with the latest reagent kits, flow cells with double reader-head (particularly, R10.4.1 flow cell used in current study), and basecalling algorithms significantly improved the base accuracy to over 99% (Ermini and Driguez, 2024). Additionally, application of algorithms developed precisely for MC profiling from full-length 16S rRNA sequences, such as Emu, ensures generation of fewer false positives and false negatives than alternative methods, thus reducing the error rates (Curry et al., 2022). Furthermore, a recent study reported Q-scores close to Q28 (~99.84% base accuracy) for ONT reads, emphasizing the notable improvement in sequencing quality achieved by Nanopore platforms in recent years (Zhang G. et al., 2024; Zhang T. et al., 2024). This trend of increasing basecalling accuracy, coupled with improved library preparation protocols and real-time data processing, highlights the rapidly growing potential of ONT for accurate and large-scale metataxonomic studies.

In recent years, both PacBio and ONT sequencing platforms have significantly improved in accuracy, enabling more precise analysis of MC. However, challenges remain, such as the higher error rates inherent to ONT reads and the reliance on error-correction algorithms in PacBio sequencing (Zhang et al., 2023).

Although comparative studies of microbiome sequencing data, including soil microbiomes, have been conducted, they typically focus on pairwise platform comparisons and often lack biological replication (Stevens et al., 2023; Yeo et al., 2024; Biada et al., 2025). In contrast, our study includes three independent biological replicates per soil sample, enabling a more robust comparison of three sequencing technologies (Illumina, PacBio, ONT) and minimizing random variation. This approach enhances the reliability of diversity estimates and strengthens conclusions regarding the comparative performance of the platforms. However, few studies have evaluated the combined impact of sequencing technology, read depth, and analytical tools on soil microbiota profiling using full-length 16S rRNA gene sequencing. Our work addresses this gap by directly comparing all three platforms under controlled conditions, incorporating multiple sequencing depths and a standardized bioinformatic pipeline.

Building on this experimental design, we systematically assessed the performance of PacBio, ONT, and Illumina platforms in detecting bacterial diversity in soil samples. By analyzing full-length 16S rRNA gene sequences across different read depths (10,000, 20,000, 25,000, and 35,000 reads per sample), we compared alpha and beta diversity metrics and examined the taxonomic resolution of each platform. This study provides new insights into the relative strengths and limitations of each technology for soil microbiome research.

Soils are complex ecosystems characterized by immense variability in ecological properties and MC structure on both global and local scales (Fierer, 2017). Numerous studies have documented the intricate relationships between microbial traits and soil characteristics, with most focusing on individual parameters such as pH, temperature, vegetation, texture, soil moisture content, nitrogen levels, organic matter, or contaminants (He et al., 2006; Ge et al., 2008; Fierer et al., 2009; Lauber et al., 2009; Griffiths et al., 2011; Lanzen et al., 2016; Liu et al., 2016; Zhou et al., 2016). These studies have provided valuable insights into specific interactions, yet they often fail to capture the multifaceted nature of soil ecosystems.

The relevance of the study lies in the growing need to accurately profile soil microbiomes, which play a crucial role in ecosystem functioning and agricultural productivity. Our study’s novelty stems from the detailed comparison of PacBio, ONT and Illumina sequencing technologies at varying read depths, highlighting their strengths and limitations in detecting bacterial species in soil samples. To enhance the robustness of our analysis, we included three biological replicates per soil sample and retained only taxa consistently detected across replicates. This strategy reduced random variation and improved the reliability of diversity estimates, allowing for more confident interpretation of platform performance.

Recent advancements in sequencing technologies, including PacBio, ONT, and Illumina platforms, have provided unprecedented opportunities to explore soil microbiomes in detail. These tools enable researchers to link microbial diversity with soil properties more comprehensively, offering insights into the complex mechanisms that govern MC composition and function. Such integrative approaches are crucial for advancing our understanding of the ecological processes underpinning soil ecosystems and their responses to environmental changes.

Alpha diversity analysis applying the Shannon index, a widely used metric accounting both species richness and evenness within a sample (Roswell et al., 2021), showed statistically significant differences between ONT and PacBio data at 10,000 reads, but not at 20,000, 25,000 and 35,000 reads. The results indicate that while low read depths may show significant differences in diversity metrics, due to their sensitivity to sequencing effort (Schloss, 2024), increasing the read depth mitigates this effect, leading to more comparable results between sequencing technologies. Such results at low sequencing depth can also be caused by randomness during rarefaction of samples’ data. This aligns with the rarefaction curve analysis, which demonstrated that bacterial diversity in the samples reaches a plateau after 35,000 reads, indicating sufficient taxonomic coverage. These findings emphasize the importance of selecting an optimal read count for capturing the full scope of microbial diversity and are consistent with previously published data (Hussein et al., 2017; Mysara et al., 2017). Higher efficiency of PacBio at lower read depth can be attributed to its longer read lengths and higher accuracy due to employment of circular consensus sequencing technology which is based on the generation of multiple reads of the same DNA molecule, thereby reducing random errors (Wang et al., 2019; Kim et al., 2024). PacBio’s high sensitivity in microbiome analysis has been previously demonstrated, showing 100% specificity and sensitivity in taxonomic classification of a mock community with 20 bacterial species (Earl et al., 2018). ONT, on the other hand, offers advantages such as portability and cost-efficiency, but its sensitivity and accuracy are currently considered inferior to PacBio (Straub et al., 2020).

Despite technological differences, overall trends in alpha diversity remained consistent across sequencing platforms. Long-read technologies such as PacBio and ONT provide better coverage of hypervariable regions within the 16S rRNA gene, thereby improving taxonomic classification (Nygaard et al., 2020; Santos et al., 2020; Szoboszlay et al., 2023; Buetas et al., 2024). However, there are trade-offs to consider when selecting a platform for microbiome analysis: while Illumina offers a more established and standardized bioinformatics pipeline, PacBio and ONT generate longer reads that can enhance resolution and accuracy in profiling bacterial communities. Moreover, current long-read classification tools, such as Emu, lack explicit chimera detection modules, which may affect data quality in complex communities. This highlights the need for further development of chimera-aware algorithms tailored to long-read 16S rRNA data. Other tools, such as VSEARSH (Rognes et al., 2016), can perform de novo chimera detection. However, in this study the chimera removal step is absent. Instead of this in order to make our taxonomic profiling more accurate we applied filtering of observed OTUs by their presence in biological repeats. In the broader context, ongoing improvements in basecalling algorithms (e.g., Dorado super mode), error-correction tools (such as Medaka and NanoCLUST), and new taxonomic classifiers specifically tailored for long, noisy reads (e.g., Emu) are helping to improve the reliability of ONT-based microbial profiling (Zhang G. et al., 2024; Zhang T. et al., 2024).

Beta diversity analysis revealed that samples formed two distinct clusters corresponding to the type of sequencing technology, indicating that each technology contributes uniquely to the detection of bacterial diversity and influences the outcomes of MC analysis. This distinct clustering suggests that PacBio and ONT may have different biases in detecting certain bacterial taxa, underscoring the importance of considering the choice of sequencing technology when interpreting MC data (Kim et al., 2024). Analysis of the overlap and uniqueness of species identified by each technology revealed that more than 57% of the total species were observed applying both technologies, indicating their complementarity in detecting bacterial diversity. This complementarity suggests that using both technologies in tandem could provide a more comprehensive view of the soil microbiome, a finding supported by previous research (Kim et al., 2024).

Both ONT and PacBio demonstrated a similar efficiency in identifying the top 6 bacterial genera. This suggests that key bacterial genera in the samples can be reliably detected even at lower sequencing depths, while additional sequencing primarily contributes to identifying low-abundance taxa.

Differences in the number of bacterial species detected in soil samples between PacBio and ONT may be attributed to several factors related to the technological characteristics of these platforms as well as bioinformatics data processing. A total of 0.52% of reads from PacBio and 16.7% of reads from ONT failed filtering, indicating the lower accuracy of ONT. This lower accuracy can lead to errors in species identification, particularly in homopolymeric regions or when analyzing closely related sequences. In addition to differences in read accuracy and filtering rates, our results suggest that PacBio may be more effective than ONT in detecting low-abundance taxa in soil samples. This discrepancy could arise from differences in bioinformatic processing pipelines. For ONT, error correction algorithms often remove low-frequency variants, mistaking them for sequencing artifacts. In contrast, PacBio’s CCS provides high per-read accuracy (>99%) without aggressive filtering, thereby preserving rare sequence variants (Johnson et al., 2019). Furthermore, soil microbial communities include many high-GC-content organisms, which may be systematically underrepresented in ONT datasets due to known biases in basecalling and pore-level performance. PacBio appears to be less affected by GC bias, which may further enhance its ability to recover rare taxa from complex environments like soil. Recent developments in ONT bioinformatics, particularly error correction tools based on machine learning and deep learning frameworks, are further enhancing read accuracy and taxonomic resolution. For example, DeChat, a novel tool that incorporates repeat- and haplotype-aware models, significantly improves the correction of Nanopore reads in complex genomic regions (Liu et al., 2024). Although not directly applied in our current 16S rRNA amplicon study, such methods hold promise for future integration into long-read metataxonomic pipelines, especially when dealing with highly diverse or repetitive microbial genomes. However, one of the inherent limitations of 16S rRNA gene sequencing lies in its inability to reliably distinguish between closely related bacterial species that share high sequence similarity in this gene. For instance, Veillonella rogosae and V. parvula exhibit 98% homology in their 16S rRNA gene sequences, which often leads to misidentification (Arif et al., 2008). This limitation underscores the need for whole-metagenome sequencing approaches to achieve more accurate species-level identification.

The Illumina MiSeq platform, with a read length limitation of up to 600 bp, does not allow for full-length 16S rRNA gene sequencing. Instead, its hypervariable regions can be used individually or in combination to evaluate bacterial community structures. Previous studies have demonstrated that the choice of primers and targeted regions significantly influences microbiome profiling results. These factors should therefore be carefully considered during the sequencing (Klindworth et al., 2013; Albertsen et al., 2015; Tremblay et al., 2015; Fouhy et al., 2016; Rintala et al., 2017; Fuks et al., 2018; Zhang et al., 2018).

A substantial body of research has explored the impact of primer selection on microbiome community profiling. For example, studies have shown that the analysis of human microbiomes and microbial alpha diversity in fecal samples varies depending on whether the V3–V4 or V4–V5 region is targeted (Rintala et al., 2017). Conversely, other studies have reported minimal differences in community profiling when regions such as V1–V3, V3–V4, or V4 are used within the same sample type. Comparable research conducted on environmental samples, such as water, suggests that the V4 region may be more suitable for precise sequence assignment in the bacterial domain while also offering increased coverage (Zhang et al., 2018).

Albertson et al. reported differences in the distribution of bacterial taxa when targeting V1–V3, V3–V4, and V4 regions, yet observed similar alpha diversity values across these regions (Albertsen et al., 2015). However, gaps remain in our understanding, particularly for highly diverse samples such as those derived from soil. Current studies lack detailed insights into the influence of 16S rRNA regions on bacterial community profiling for both environmental and biological samples. Furthermore, there is an absence of comprehensive analyses that include the use of combined regions such as V1–V3, V3–V4, V4–V5, and V6–V8, which are frequently employed to enhance taxonomic accuracy in amplicon sequencing (Soriano-Lerma et al., 2020).

Future research should address existing gaps by investigating how different 16S rRNA regions and primer combinations influence the profiling of complex and diverse MC (Strokach et al., 2025). This discrepancy underscores the critical role of primer selection in microbiome studies. While region-specific primers, such as those targeting the V4 region, may enhance taxonomic resolution for specific groups (Kozich et al., 2013; Zhou et al., 2024), they risk overlooking minor bacterial populations essential for understanding MC complexity. In contrast, universal primers used for full-length 16S rRNA gene sequencing provide a more comprehensive view of community evenness, making them better suited for studies aiming to capture a broad range of taxa (Schloss et al., 2009; Caporaso et al., 2012). These findings highlight the need to carefully consider primer design and target region selection when planning sequencing-based microbiome studies. Future research should further explore the trade-offs between primer specificity and universality to optimize microbial diversity assessments across different environments (Abellan-Schneyder et al., 2021), ultimately improving the reliability of microbiome analyses and guiding the selection of sequencing strategies for both targeted and high-resolution profiling.

The observed differences in taxonomic profiles between regions sequenced on PacBio and Illumina platforms may be attributed to the primers used during the initial PCR step in library preparation. Primers targeting the full-length 16S rRNA gene, as used in PacBio sequencing, are designed to be more universal and inclusive, allowing for a broader representation of bacterial taxa, including low-abundance groups. In contrast, primers targeting the V4 region, often employed in Illumina sequencing, are more specific to certain bacterial species, which can limit their effectiveness in identifying minor taxa and reduce overall taxonomic resolution (Klindworth et al., 2013; Apprill et al., 2015).

In our analysis, we also observed differences in taxonomic composition depending on the sequencing technology used. Sequencing with the MiSeq platform showed an increased presence of the genera Peribacillus, Gaiella, and Neobacillus, whereas the ONT platform identified a higher number of taxa belonging to Palsa-739, Blastococcus, Sphingomicrobium, Sphingobium, and Bacillus. These results are consistent with literature data indicating that the sequencing platform has a significant impact on the relative abundance of taxa. Specifically, MiSeq sequencing tends to detect a higher proportion of Actinobacteria, Chloroflexi, and Gemmatimonadetes, while the abundance of Acidobacteria, Bacteroides, Firmicutes, Proteobacteria, and Verrucomicrobia is lower compared to the MinION platform (Stevens et al., 2023). Thus, taxonomic identification based on Illumina reads appears to be biased toward Actinobacteria and related groups, whereas ONT enables the detection of a greater number of Bacilli representatives.

Our analysis demonstrated a high level of concordance in bacterial composition down to the genus level when using both sequencing platforms (Illumina and PacBio). A comparison of the presence and absence of taxonomically assigned bacteria showed that most genera were detected on both platforms, indicating that the obtained data are comparable and suitable for future genus-level comparisons (Buetas et al., 2024). Similar results, but using ONT, were reported in the study by Matsuo et al., which assessed the efficiency of full-length 16S rRNA gene sequencing with the MinION^™ technology for analyzing human fecal samples. Their findings also confirmed that different sequencing platforms, particularly Illumina and ONT, provide reproducible results at the genus level (Matsuo et al., 2021).

In summary, all three sequencing platforms (ONT, PacBio, and Illumina) are suitable for determining the composition of MC based on the 16S rRNA gene. In our study, we analyzed three different soil types, and the results of metataxonomics sequencing obtained from all platforms allowed for a clear differentiation between them. However, it is important to note that when using the Illumina technology to analyze only the V4 variable region, no distinct clustering of samples based on soil type was observed. Additionally, each platform exhibits taxonomic biases, which may complicate the comparison of results across studies utilizing different sequencing methods.

A comparative analysis of the PacBio and ONT platforms for soil microbiome profiling revealed a high degree of similarity in the obtained data, with PacBio demonstrating slightly higher efficiency in identifying low-abundance taxa at lower sequencing depths. Our results highlight the importance of selecting an optimal sequencing depth and considering the specific characteristics of each technology when analyzing MC. We demonstrated that all three platforms (ONT, PacBio, and Illumina) are capable of assessing bacterial diversity in soil samples and distinguishing them into groups based on their origin. Furthermore, the data obtained from ONT are comparable to those from PacBio, indicating that sequencing errors inherent to ONT do not significantly impact the interpretation of well-represented taxa. Due to their long-read capability and the ability to analyze the full-length 16S rRNA gene, both ONT and PacBio offer advantages over Illumina for bacterial diversity assessment. Our study confirms that the choice of sequencing technology can substantially influence microbiome data interpretation. Future research should focus on improving bioinformatics approaches, particularly for the detection of rare taxa, while also accounting for taxonomic biases associated with different sequencing platforms.