Current problems in sustainable agriculture include the threat of biodiversity loss, a significant issue in recent years. The threats most often mentioned include climate change, erosion, depletion of soil organic matter, agricultural intensification, and land-use changes. Therefore, the search for sustainable plant cultivation strategies is essential for maintaining the quality of the farming environment. These strategies are used to develop biotechnological solutions for sustainable and organic agriculture. The new vision of agriculture emphasizes the close connection between crop management and the soil and plant microbiome. Therefore, plant and soil microbiome management is increasingly used to: enhance the resistance of specific crops to pests, pathogens, drought or nutrients, select/develop pest control practices that are best in the context of sustainable food production, fully integrate biological agents/changes or microbiome control with crop management depending on the location, conditions and environmental factors (e.g., climate zone, soil type, biotic and abiotic stresses), develop biopreparations and methods for detecting phytopathogens for sustainable crop production management.

Therefore, the interactions between plants and microbiomes and their hosts are essential functional contexts. Plant–microbiome interactions have co-evolved to maintain the plant’s overall stability, functionality, and fitness as a holobiont (Trivedi et al., 2020). A plant holobiont is defined as an ecological and functional unit composed of the plant host and its associated microbiota, including bacteria, fungi, archaea, protists, viruses, and other microorganisms inhabiting both above- and belowground plant compartments (Mesny et al., 2023; Vandenkoornhuyse et al., 2015). These microbial partners interact with the host through metabolic, signaling, and immune-mediated processes, collectively shaping plant development, health, stress resilience, and adaptive capacity within a given environment (Berg et al., 2016; Berg et al., 2020). Plants are recognized as metaorganisms because the plant-associated microbiome comprises all microorganisms colonizing plant surfaces and internal tissues, including the rhizosphere, phyllosphere, endosphere, and spermosphere (Berg et al., 2017). These microorganisms can exert beneficial, neutral, or detrimental effects on the host and play key roles in nutrient cycling, pathogen suppression, modulation of plant immunity, and responses to biotic and abiotic stresses (Frąc et al., 2018; Jansson and Hofmockel, 2020; Siegieda et al., 2023; Trivedi et al., 2020).

A plant microbiome comprises beneficial, neutral, and pathogenic microorganisms. The benefits of microorganisms to their host plants can be direct – including transformation and translocation of relevant nutrients in the soil to make them available to plants, protection against plant pathogens through antibiosis competition, production of hydrolytic enzymes, and mitigation of environmental stresses (Trivedi et al., 2016a). Benefits can also be indirect, as they enhance plant resistance responses (Pieterse et al., 2014). As illustrated in Figure 1, the complex interactions among plant compartments demonstrate how microorganisms actively contribute to soil health and promote host plant growth. Through mechanisms such as the production of extracellular polymeric substances (EPS), systemically induced root exudation of metabolites (SIREM), and signaling via homoserine lactones (HSL), microbial communities facilitate nutrient cycling, stress resilience, and plant–soil communication. These interconnected processes emphasize the functional integration of the soil and plant microbiome in maintaining ecosystem stability and productivity.

However, diseases are characterized by microbial dysbiosis and subsequent responses from specific microbes that can act as antagonists or synergists toward pathogens (Trivedi et al., 2020). A healthy microbiome, also known as the eubiosis state, can be defined by high microbial richness, maintenance of its symbiotic functions for the host, and resistance to changes under physiological stress (Kuntz et al., 2015). The plant-associated microbiome is often described in terms of trophic modes and ecological guilds, which are defined by specific attributes of individual microorganisms. Examples include pathotrophs, which obtain nutrients by harming host cells, and phagotrophs; symbiotrophs, which obtain nutrients by exchanging resources with host cells; and saprotrophs, which obtain nutrients by breaking down dead host cells (Oszust and Frąc, 2021).

Despite significant progress in microbiome research, major knowledge gaps persist in understanding the functional mechanisms linking microbial community composition with plant performance and soil health. Most studies focus on taxonomic profiling using high-throughput sequencing but lack insight into metabolic functions, signaling pathways, and interspecies interactions that drive ecosystem stability (Bender et al., 2016; Hartmann and Widmer, 2006). Furthermore, the temporal dynamics of microbiomes – how communities change across plant growth stages, seasons, and environmental stress events – remain largely uncharacterized (Shade et al., 2012). Similarly, the context-dependence of microbial functions under different soil types, climates, and management systems has not been systematically addressed, limiting the translation of laboratory findings into field-scale applications (Raaijmakers and Mazzola, 2016; Fierer, 2017).

Several unresolved questions hinder our ability to harness microbiomes for sustainable agriculture. These include:

Understanding these questions is vital to predicting how soil microbial communities will respond to global change drivers such as warming, drought, and nutrient limitation (Cavicchioli et al., 2019).

Current approaches to studying the plant–soil microbiome are constrained by methodological and technological limitations. Although next-generation sequencing provides massive amounts of genetic data, it often fails to capture functional expression and metabolic activity in situ (Prosser, 2015; Jansson and Hofmockel, 2020). Culture-dependent methods still recover only a small fraction of microbial diversity, while metagenomic and metatranscriptomic techniques require complex bioinformatics pipelines and standardized protocols for data comparability (Franzosa et al., 2015; Knight et al., 2018). Additionally, short-term experimental designs and the lack of integrated multi-omics frameworks restrict our capacity to connect microbial genes to ecological outcomes (Bahram et al., 2018; van der Heijden and Hartmann, 2016). Addressing these limitations requires interdisciplinary strategies that combine long-term monitoring, experimental ecology, and predictive modeling to bridge the gap between descriptive and functional microbiome science (Berg et al., 2020; Trivedi et al., 2020).

Such topics as an exploration of interactions within the plant–soil-microbiome structure, their resilience to pathogens, studying their composition and employing advanced bioinformatic tools to find connections and impact of biodiversity shifts toward crop health and productivity, and ways to restore and induce healthy microbiome although were briefly discussed in the past, are not yet presented in condensed, linked together manner, what is the primary focus of this review.

This review examines the interactions between plants, soil, and their microbial communities, focusing on how influencing microbial composition can enhance agricultural sustainability and preserve or restore biodiversity.

The challenges posed by broad urbanization, a changing climate, and a rapidly growing global population, which are driving increasing demand for food, have led to the development of new approaches to innovative farming, such as vertical and urban agriculture. Vertical farming involves cultivating crops in containers stacked vertically within controlled environments, employing technologies such as media-based methods, hydroponics, aeroponics, and LED lighting. Such an approach allows optimization not only of space but also of water and nutrient use through the application of closed-loop systems, minimizing reliance on herbicides and pesticides and further improving sustainability and resource efficiency (Benke and Tomkins, 2017). On the other hand, urban farming integrates farming into metropolitan areas, such as transforming rooftops, vacant spaces, or indoor areas into farmland. Such urban farming can enhance food security and resilience against disruptions to the food supply chain, while promoting community engagement and environmental conservation, or improving urban landscapes (Specht et al., 2014; Thomaier et al., 2015).

Another topic closely related to urban and vertical farming is the production of so-called “superfoods.” Often, foods such as sprouts and microgreens are considered superfoods. This food is rich in macro- and micronutrients and possesses favorable properties and effects on human health, thanks to its high content of vitamins, minerals, and antioxidants (Franco Lucas et al., 2021). The integration of a controlled vertical farming environment with superfood production ensures consistent quality and yield, maintaining year-round availability regardless of external conditions. Moreover, urban agriculture can significantly reduce the carbon footprint associated with food storage and transportation by lowering logistics requirements (Kulak et al., 2013). As such, monitoring and maintaining the microbiome composition of vertically and urban-cultivated plants is even more critical, as the stability of the plant holobiont ensures the predictability and stability of production and resilience against plant diseases (Du et al., 2025).

Climate change impacts significant aspects of our lives. But not only is human life altered. Rising temperatures and CO₂ levels alter the soil microbiome’s abundance, behavior, and diversity, affecting its fertility and soil–plant microbiome interactions, which, in turn, change plant resistance to stresses and vulnerability to crop diseases. Currently, the most efficient methods for studying changes in soil and plant microbiome diversity involve sequencing and characterizing microbial DNA. Advances in DNA sequencing methods, computer power, and bioinformatic tools have enabled the study of the genetic diversity of microbial communities, including previously uncultivable species. Two strategies for determining microbiome composition have emerged: amplicon and metagenomic sequencing. Amplicon sequencing focuses on defining and amplifying the sequence of a single gene fragment. Most frequently targeted are marker genes that are taxonomically and phylogenetically informative (Hugenholtz and Pace, 1996; Schoch et al., 2012). Such sequencing is often referred to as metataxonomics (Marchesi and Ravel, 2015). The relatively low cost of metataxonomic methods allows for studying how microbial profiles and their diversity change in response to environmental changes. Although this approach is promising and accessible, it has some critical constraints. First, amplicon sequencing is prone to PCR-related bias, including artifacts that skew the distribution of PCR products due to unequal amplification (Acinas et al., 2005; Hong et al., 2009). Moreover, choosing primers is crucial for obtaining high-quality data and is constantly being re-evaluated.

In contrast to amplicon sequencing, metagenomic sequencing sequences all the DNA present in a studied sample. This is done using a method known as “shotgun” metagenomics. Genetic material from a community is cut into shorter fragments using various techniques (such as sonication, mechanical tearing, or enzymatic restriction) and, after adding sequencing adapters, sequenced (Sharpton, 2014). Metagenomic sequencing can provide more comprehensive information about the genes involved in metabolic pathways in a sample, as well as generate metagenome-assembled genomes (MAGs). However, shotgun sequencing, especially of soil samples, requires a sequencing depth that is not comparable to that of amplicon sequencing. Sequencing of microbiomes is heavily hindered by host DNA “contamination,” requiring even greater sequencing depth (Sharpton, 2014).

To study shifts in soil microbiomes driven by climate change, appropriate sequencing techniques must be used. Illumina sequencing-by-synthesis is currently the gold standard in metataxonomics due to its ability to sequence either the V4 region of the 16S rRNA gene or the V3–V4 region of the 16S rRNA gene, as well as ITS1 or ITS2, in large quantities. Due to the observation that the longer the read across marker region, the better the accuracy of microbial identification, developments in recent years have brought different approaches to metataxonomics, focusing on obtaining much longer reads (Payne et al., 2019; Wang Y. et al., 2021), spanning the entire 16S rRNA gene (Wagner et al., 2016) or the entire ITS1-5.8S-ITS2 region (Purahong et al., 2019). Although there have been early-stage problems with read quality (Goodwin et al., 2016; Tedersoo et al., 2018), there already is the possibility to utilize customer-ready third-generation long-read sequencing technologies, with PacBio offering Single-Molecule-Real-Time (SMRT) sequencing within its latest Revio and Vega platforms, while Oxford Nanopore Sequencing Technologies offers nanopore sequencing-based technology with both mobile and accessible platform Minion and higher throughput ones like Gridion and Promethion sequencers (Goodwin et al., 2016; Tedersoo et al., 2018).

Thanks to the integration of metataxonomic and metagenomics approaches enabled by The Data Integration Analysis for Biomarker discovery using a Latent cOmponents (DIABLO) framework (Singh et al., 2019) in mixOmics (Rohart et al., 2017), with zero radius Operational taxonomic units (zOTU) constructed by USEARCH (Edgar, 2010) and Ribosomal Database Project (Cole et al., 2014), while also using Metagenomic Rapid Annotations using Subsystems Technology (MG-RAST), Ferrarezi et al. (2023) confirmed positive effect of plant growth promoting bacteria Azospirillum brasilense inoculations on maize growth. Moreover, a metagenomics metastudy by Masuda et al. (2024) found that Anaeromyxobacteraceae and Geobacteraceae within Deltaproteobacteria are groups of nitrogen-fixing microorganisms present within microbiomes in ecosystems with different land usage types and geographic origins. Functional profiling enabled by the metagenomics approach to soil studies allowed Pang et al. (2021) to determine that continuous sugarcane cultivation significantly reduction the abundance of the functional pathway, such as genes related to nitrogen and sulfur cycling in soil, reduced diversity of soil bacterial and fungal communities, significantly reduced the number of bacteria associated with soil nitrogen and sulfur cycling functions, and enriched pathogenic bacteria.

Recent years have brought significant advances in computational power, making it possible to apply supervised Machine Learning (ML) techniques to studies of soil. ML is widely used with microbiome data, as naive Bayesian classifiers have found utility in assigning taxonomy to Amplicon Sequence Variants or Operational taxonomic Units (Pedregosa et al., 2011; Wang et al., 2007), while model for learning error profile of reads within sequencer run based on LOESS (locally estimated scatterplot smoothing) function fitting was developed to overcome limitations of OTU clustering with emergence of ASV approach to metataxonomic data. Moreover, methods such as the random forest (RF) classifier and the L2 Support Vector Machine (L2-SVM) with a linear kernel have been successfully deployed to predict many soil properties and even crop productivity (Chang et al., 2017; Wilhelm et al., 2022).

What is essential is that the RF method was found to perform best for typical microbiome data (Thompson et al., 2019; Zhou and Gallins, 2019), while L2-SVM is often chosen for its speed (Topçuoğlu et al., 2020). Depending on the choice of ML algorithm, classifier, or regressor, categorical and numerical variables can be predicted.

Recent advances in metagenomics, metatranscriptomics, and metabolomics have revolutionized our understanding of plant-microbe interactions, enabling the identification of functional genes, signaling pathways, and metabolic exchanges that govern plant health and soil ecosystem dynamics (Levy et al., 2018; Hacquard et al., 2015). These multi-omics approaches reveal not only the taxonomic composition but also the functional potential and real-time activity of microbial communities in response to environmental and plant-derived cues (Delgado-Baquerizo et al., 2018). Integrating these data with artificial intelligence (AI) and machine learning (ML) has enabled researchers to build predictive models that link microbiome composition to plant traits, stress tolerance, and disease outcomes (Xu et al., 2018). For instance, ML algorithms are now applied in disease prediction, microbiome-based crop yield forecasting, and trait selection for breeding programs that favor beneficial microbial associations (Xu et al., 2018; Zhao et al., 2023; Chang et al., 2017). Together, these approaches are transforming plant–microbiome research from descriptive ecology into predictive and actionable systems biology, paving the way for precision agriculture and microbiome engineering.

Machine learning algorithms can now be easily employed using QIIME2 plugin “sample-classifier,” which supports supervised ML methods for classification and regression of sample properties (Bolyen et al., 2019; Bokulich et al., 2018). There are also attempts to utilize soil and plant metagenomic data, along with machine learning, to predict their susceptibility to plant diseases. It is a powerful tool that may become useful in predicting how microbiomes will adapt to climate change and how these changes will affect soil properties and health (Figure 2). Recently, Demilie (2024) suggested that the best results for predicting plant diseases are obtained when deploying ML models on each compartment of the studied plant and its environment. It is also stated that new approaches, such as the real-time application of ML methods to predict or detect diseases, are being developed, and that studies focusing on these approaches should be encouraged. Bioinformatics can be deployed to assess microbial community structure and functions, predict plant diseases, and assess soil health and quality in a changing climate, based on soil microbiome analysis and computation.

Increasing soil performance, persistence, and inoculation efficiency with microbial-based products is a priority to harness their potential and reduce risks of adverse outcomes. Although numerous studies have focused on aspects of soil microbiomes, including structure, biodiversity, and functions, significant knowledge gaps remain, and substantial research is needed to understand the interactions within soil–plant–microbiomes in different ecosystems, thereby gaining a deeper understanding of soil functionality and microbiome-based services for agroecosystems. The challenges in this context can serve as principal future perspectives in soil microbiome research, especially under changing climate conditions. In conclusion, we suggest studies to include (a) a deeper explanation of how soil microbiomes’ structure and functionality can contribute to the future development of climate-resilient and resistant plants to biotic and abiotic stresses; (b) integration of different omics approaches to increase higher-resolution characterization of soil microbiomes to define holistic healthy (eubiotic) and unhealthy (dysbiotic, pathobiomic) plant and soil microbiomes both to develop predictive models of plant disease occurrence and to improve climate models; (c) implementation of more cultivation-based assays into microbiome research to more explicitly describe ecotypes and adaptation modes of specific microbial groups to a changing environment; to date, there are not well enough described and designed studies in these areas. These prospects are supported by the latest soil priorities in the European Union, which highlight soil health restoration as an essential challenge by 2050. These priorities are contained in the Horizon Europe mission “Caring for soil is caring for life”, supported by “A Soil Deal for Europe”, and contribute to multiple European Green Deal targets on climate resilience, sustainable farming, zero pollution, and biodiversity. Moreover, the strategies and approaches presented in this review align with the European Commission’s Microbiome World Pathway, which identifies systemic challenges and necessary actions by 2030 (European Commission, 2020).

Finally, given the challenges facing modern and future agriculture, revolutionary approaches and solutions to climate and sustainability problems, food production, security, and plant health defense must be pursued. The methods, strategies, and challenges should not only refer to large-scale agricultural and horticultural crops, but also range from agroecological management techniques based on microbially mediated ecosystem functions to advanced sustainable farming techniques, such as vertical and urban farming, and the production of superfoods, including microgreens. The application of microbial inoculants, biofertilizers, and genetically enhanced biocontrol agents offers significant promise for sustainable agriculture, yet it also raises ecological and regulatory concerns that warrant careful consideration. Introducing non-native or engineered microorganisms into soil ecosystems can disrupt native microbial community structure and function, potentially leading to competitive exclusion or altered nutrient cycling (Li et al., 2022; Müller et al., 2016). Moreover, the risk of horizontal gene transfer (HGT) between introduced strains and indigenous microbes poses challenges for biosafety, as antibiotic resistance or virulence genes could be unintentionally disseminated (van Elsas et al., 2003; Heuer and Smalla, 2012). Environmental persistence of inoculants and their metabolites may also cause unforeseen ecological feedbacks, particularly under varying soil and climate conditions (Mumtaz et al., 2025). Regulatory frameworks for microbial-based products remain fragmented globally, often lagging behind technological advances in synthetic biology and microbial engineering (Elazzazy et al., 2025; Pellegrini et al., 2025). Therefore, risk assessment protocols integrating genomic, ecological, and functional data are essential to ensure that microbiome-based innovations promote sustainability without compromising ecosystem integrity or biosafety.