Additional information on the microbial community structure can be acquired from amplicon sequencing data trough the framework of co-occurrence analysis. This approach consists in checking which species occur together and which ones esuppress each other in a high number of environmental samples. These patterns are imprinted in the abundances of microorganisms, and are used to build a network with biological species as nodes and edges representing significant associations assigned by a given criteria. After its construction, the network can simply be used as a form of data visualization or for further theoretical analysis. This network analysis uses tools from the field of complex systems aiming to link structure of the microbial community to its function, identify important/keystone species and even make predictions about system's stability to environmental perturbations. Despite being very promising, the successful interpretation of the network analysis relies crucially on our ability to establish a link between co-occurrence patterns and meaningful biological interactions. Therefore it is advisable to approach this step with extra caution and carefully consider all assumptions and limitations of all the different methods used to construct the network\cite{Weiss2016,Faust2012}. Here we highlight some of the most discussed issues, note that most of them are not unique to soil. First, we reiterate that the interactions occur at the level of individual microorganisms while the detectable abundance patterns can only be measured on relatively large and possible highly heterogeneous soil samples. Secondly, the commonly used methods are based on correlations or similarity indexes (e.g. see widely used package CoNet\cite{Faust2012}), which give no direct information on causal relations within the community and are subject to confounding environmental factors. In addition to these difficulties, we have to take into account the compositional nature and high sparsity of data sets, which can significantly overestimate or induce spurious correlations to the co-occurrence patterns. These issues can be corrected with uses of log ratios and protocols for handling low read values (as done by packages such as SparCC\cite{Friedman2012}). All methods based on correlations will also result in networks which include indirect associations between microorganisms, an issue which methods based on inverse covariance are designed to resolve (see SpiecEasi \cite{Kurtz2015}). Finally we would like to mention that alternative approaches exist to also detect non monotonic associations between microorganisms such trough Maximal Information Coefficient (MIC) \cite{Reshef_2011}. In all methods a proper comparison with null models is advised to select only significant associations\cite{Connor2017}. After network is construction additional information, such as spatial distribution of samples, can be used in a complementary analysis to account for possible influence of environmental filtering and dispersal limitation on the community composition\cite{Goberna_2019}. In summary, the field of network inference is a rapidly evolving one and we constantly see new alternatives proposed to solve currently standing issues. Nevertheless we still lack a definite framework which allows to generate co-occurrence networks with a straight forward and easy interpretation. The current approach can still be a useful step to formulate hypothesis on potential microbial interactions and the organization of communities in soil.
Improving ecological insights from sequencing
Consequently, the majority of sequencing studies remain highly descriptive due to their design and the limitations of the nature of the data. Microbial ecology as a field should bridge microbiological isolation approaches and characterization of microbial communities, while reconciling the heterogeneity of soil systems in which microorganisms live. Improving the quantitative nature of sequencing studies, and applying amplicon sequencing of functional genes are two approaches toward expanding the insights researchers are able to gain into soil microbial communities. Recent studies are beginning to combine other forms of data with amplicon sequencing data to improve investigations of ecological patterns.
Combinations of amplicon sequencing and stable-isotope probing have been used as a viable option to link microbial activity to microbial abundance (54).
Other researchers have combined sequencing approaches in order to improve inferences made from amplicon sequencing data (55). Since vertical gains and losses of genetic information are common in prokaryotic world, there is a substantial probability of error associated with functional assignment of taxa based solely on 16S identities. To circumvent this recognized limitation, metagenomic and metatranscriptomics analyses are increasingly being used to describe the functional gene diversity and expression in various environmental samples, although both sequencing and bioinformatic efforts needed for gaining functionally relevant insights into ecosystem processes by these approaches are usually magnitudes higher than those needed for analyzing amplicon sequencing data.