Machine learning applications in microbial ecology, human microbiome studies, and environmental monitoring

Microbial Communities Influence Soil Dissolved Organic Carbon Concentration by Altering Metabolite Composition

Rapid microbial growth in the early phase of plant litter decomposition is viewed as an important component of soil organic matter (SOM) formation. However, the microbial taxa and chemical substrates that correlate with carbon storage are not well resolved. The complexity of microbial communities and diverse substrate chemistries that occur in natural soils make it difficult to identify links between community membership and decomposition processes in the soil environment. To identify potential relationships between microbes, soil organic matter, and their impact on carbon storage, we used sand microcosms to control for external environmental factors such as changes in temperature and moisture as well as the variability in available carbon that exist in soil cores. Using Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) on microcosm samples from early phase litter decomposition, we found that protein- and tannin-like compounds exhibited the strongest correlation to dissolved organic carbon (DOC) concentration. Proteins correlated positively with DOC concentration, while tannins correlated negatively with DOC. Through random forest, neural network, and indicator species analyses, we identified 42 bacterial and 9 fungal taxa associated with DOC concentration. The majority of bacterial taxa (26 out of 42 taxa) belonged to the phylum Proteobacteria while all fungal taxa belonged to the phylum Ascomycota. Additionally, we identified significant connections between microorganisms and protein-like compounds and found that most taxa (12/14) correlated negatively with proteins indicating that microbial consumption of proteins is likely a significant driver of DOC concentration. This research links DOC concentration with microbial production and/or decomposition of specific metabolites to improve our understanding of microbial metabolism and carbon persistence.

MathFeature: feature extraction package for DNA, RNA and protein sequences based on mathematical descriptors

One of the main challenges in applying machine learning algorithms to biological sequence data is how to numerically represent a sequence in a numeric input vector. Feature extraction techniques capable of extracting numerical information from biological sequences have been reported in the literature. However, many of these techniques are not available in existing packages, such as mathematical descriptors. This paper presents a new package, MathFeature, which implements mathematical descriptors able to extract relevant numerical information from biological sequences, i.e. DNA, RNA and proteins (prediction of structural features along the primary sequence of amino acids). MathFeature makes available 20 numerical feature extraction descriptors based on approaches found in the literature, e.g. multiple numeric mappings, genomic signal processing, chaos game theory, entropy and complex networks. MathFeature also allows the extraction of alternative features, complementing the existing packages. To ensure that our descriptors are robust and to assess their relevance, experimental results are presented in nine case studies. According to these results, the features extracted by MathFeature showed high performance (0.6350-0.9897, accuracy), both applying only mathematical descriptors, but also hybridization with well-known descriptors in the literature. Finally, through MathFeature, we overcame several studies in eight benchmark datasets, exemplifying the robustness and viability of the proposed package. MathFeature has advanced in the area by bringing descriptors not available in other packages, as well as allowing non-experts to use feature extraction techniques.

Extending Association Rule Mining to Microbiome Pattern Analysis: Tools and Guidelines to Support Real Applications

Boosted by the exponential growth of microbiome-based studies, analyzing microbiome patterns is now a hot-topic, finding different fields of application. In particular, the use of machine learning techniques is increasing in microbiome studies, providing deep insights into microbial community composition. In this context, in order to investigate microbial patterns from 16S rRNA metabarcoding data, we explored the effectiveness of Association Rule Mining (ARM) technique, a supervised-machine learning procedure, to extract patterns (in this work, intended as groups of species or taxa) from microbiome data. ARM can generate huge amounts of data, making spurious information removal and visualizing results challenging. Our work sheds light on the strengths and weaknesses of pattern mining strategy into the study of microbial patterns, in particular from 16S rRNA microbiome datasets, applying ARM on real case studies and providing guidelines for future usage. Our results highlighted issues related to the type of input and the use of metadata in microbial pattern extraction, identifying the key steps that must be considered to apply ARM consciously on 16S rRNA microbiome data. To promote the use of ARM and the visualization of microbiome patterns, specifically, we developed microFIM (microbial Frequent Itemset Mining), a versatile Python tool that facilitates the use of ARM integrating common microbiome outputs, such as taxa tables. microFIM implements interest measures to remove spurious information and merges the results of ARM analysis with the common microbiome outputs, providing similar microbiome strategies that help scientists to integrate ARM in microbiome applications. With this work, we aimed at creating a bridge between microbial ecology researchers and ARM technique, making researchers aware about the strength and weaknesses of association rule mining approach.

OUP accepted manuscript

Introduction

Antimicrobial stewardship interventions (ASIs) aim to reduce the emergence of antimicrobial resistance. We sought to systematically evaluate how microbiological outcomes have been handled and analysed in randomized controlled trials (RCTs) evaluating ASIs.

Methods

We searched PubMed and Embase from 2011–21. Studies were selected if they were RCTs evaluating ASIs. A narrative synthesis approach was taken, identifying whether the study reported any microbiological data (bacterial genus/species; bacterial colony counts; prevalence of bacterial, microbiologically defined infections; and antibiotic susceptibility, measured pre-randomization or post-randomization in one arm only) or outcomes (post-randomization data compared between arms). Studies with or without microbiological data/outcomes were summarized in terms of study characteristics, methods of reporting and analysis of these outcomes.

Results

We identified 117 studies, with 34 (29.1%) collecting microbiological data and 18 (15.4%) reporting microbiological outcomes. Most studies with microbiological outcomes were conducted in secondary care (12/18, 66.7%) and targeted adult populations (14/18, 77.8%), and the intervention involved biomarker-guided rapid diagnostic testing (7/18, 38.9%). The overall quality of reporting and analysing microbiological outcomes was low and inconsistent. The selected study population in analyses and methods of handling missing data were unclear.

Conclusions

This review demonstrates that the quality of handling and reporting microbiological outcomes in RCTs of ASIs was low. The lack of consistency and clarity made it difficult to compare the findings across studies, limiting policy- and clinical decision-making. Therefore, there is a clear need for the development of guidance for handling microbiological outcomes in RCTs and adopting appropriate methods to evaluate these data carefully.

15 years of microbial biotechnology: the time has come to think big-and act soon

Our epoch is largely characterized by the growing realization and concern about the reality of climate change and environmental deterioration, the surge of global pandemics, the unacceptable inequalities between developed and underdeveloped countries and their unavoidable translation into messy immigration, overpopulation and food crises. While all of these issues have a fundamentally political core, they are not altogether removed from the fact that Earth is primarily a microbial planet and microorganisms are the key agents that make the biosphere (including ourselves) function as it does. It thus makes sense that we bring the microbial world-that is the environmental microbiome-to the necessary multi-tiered conversation (hopefully followed by action) on how to avoid future threats and how to make our globe a habitable common house. Beyond discussion on governance, such a dialogue has technical and scientific aspects that only frontline microbial biotechnology can help to tackle. Fortunately, the field has witnessed the onset of new conceptual and material tools that were missing when the journal started.

Modeling approaches for probing cross-feeding interactions in the human gut microbiome

Microbial communities perform emergent activities that are essentially different from those carried by their individual members. The gut microbiome and its metabolites have a significant impact on the host, contributing to homeostasis or disease. Food molecules shape this community, being fermented through cross-feeding interactions of metabolites such as lactate, acetate, and amino acids, or products derived from macromolecule degradation. Mathematical and experimental approaches have been applied to understand and predict the interactions between microorganisms in complex communities such as the gut microbiota. Rational and mechanistic understanding of microbial interactions is essential to exploit their metabolic activities and identify keystone taxa and metabolites. The latter could be used in turn to modulate or replicate the metabolic behavior of the community in different contexts. This review aims to highlight recent experimental and modeling approaches for studying cross-feeding interactions within the gut microbiome. We focus on short-chain fatty acid production and fiber fermentation, which are fundamental processes in human health and disease. Special attention is paid to modeling approaches, particularly kinetic and genome-scale stoichiometric models of metabolism, to integrate experimental data under different diet and health conditions. Finally, we discuss limitations and challenges for the broad application of these modeling approaches and their experimental verification for improving our understanding of the mechanisms of microbial interactions.

Gastrointestinal Disease Classification in Endoscopic Images Using Attention-Guided Convolutional Neural Networks

Gastrointestinal (GI) diseases constitute a leading problem in the human digestive system. Consequently, several studies have explored automatic classification of GI diseases as a means of minimizing the burden on clinicians and improving patient outcomes, for both diagnostic and treatment purposes. The challenge in using deep learning-based (DL) approaches, specifically a convolutional neural network (CNN), is that spatial information is not fully utilized due to the inherent mechanism of CNNs. This paper proposes the application of spatial factors in improving classification performance. Specifically, we propose a deep CNN-based spatial attention mechanism for the classification of GI diseases, implemented with encoder–decoder layers. To overcome the data imbalance problem, we adapt data-augmentation techniques. A total of 12,147 multi-sited, multi-diseased GI images, drawn from publicly available and private sources, were used to validate the proposed approach. Furthermore, a five-fold cross-validation approach was adopted to minimize inconsistencies in intra- and inter-class variability and to ensure that results were robustly assessed. Our results, compared with other state-of-the-art models in terms of mean accuracy (ResNet50 = 90.28, GoogLeNet = 91.38, DenseNets = 91.60, and baseline = 92.84), demonstrated better outcomes (Precision = 92.8, Recall = 92.7, F1-score = 92.8, and Accuracy = 93.19). We also implemented t-distributed stochastic neighbor embedding (t–SNE) and confusion matrix analysis techniques for better visualization and performance validation. Overall, the results showed that the attention mechanism improved the automatic classification of multi-sited GI disease images. We validated clinical tests based on the proposed method by overcoming previous limitations, with the goal of improving automatic classification accuracy in future work.

Competition, Nodule Occupancy, and Persistence of Inoculant Strains: Key Factors in the Rhizobium-Legume Symbioses

Biological nitrogen fixation by Rhizobium-legume symbioses represents an environmentally friendly and inexpensive alternative to the use of chemical nitrogen fertilizers in legume crops. Rhizobial inoculants, applied frequently as biofertilizers, play an important role in sustainable agriculture. However, inoculants often fail to compete for nodule occupancy against native rhizobia with inferior nitrogen-fixing abilities, resulting in low yields. Strains with excellent performance under controlled conditions are typically selected as inoculants, but the rates of nodule occupancy compared to native strains are rarely investigated. Lack of persistence in the field after agricultural cycles, usually due to the transfer of symbiotic genes from the inoculant strain to naturalized populations, also limits the suitability of commercial inoculants. When rhizobial inoculants are based on native strains with a high nitrogen fixation ability, they often have superior performance in the field due to their genetic adaptations to the local environment. Therefore, knowledge from laboratory studies assessing competition and understanding how diverse strains of rhizobia behave, together with assays done under field conditions, may allow us to exploit the effectiveness of native populations selected as elite strains and to breed specific host cultivar-rhizobial strain combinations. Here, we review current knowledge at the molecular level on competition for nodulation and the advances in molecular tools for assessing competitiveness. We then describe ongoing approaches for inoculant development based on native strains and emphasize future perspectives and applications using a multidisciplinary approach to ensure optimal performance of both symbiotic partners.

Key Technologies for Progressing Discovery of Microbiome-Based Medicines

A growing number of experimental and computational approaches are illuminating the “microbial dark matter” and uncovering the integral role of commensal microbes in human health. Through this work, it is now clear that the human microbiome presents great potential as a therapeutic target for a plethora of diseases, including inflammatory bowel disease, diabetes and obesity. The development of more efficacious and targeted treatments relies on identification of causal links between the microbiome and disease; with future progress dependent on effective links between state-of-the-art sequencing approaches, computational analyses and experimental assays. We argue determining causation is essential, which can be attained by generating hypotheses using multi-omic functional analyses and validating these hypotheses in complex, biologically relevant experimental models. In this review we discuss existing analysis and validation methods, and propose best-practice approaches required to enable the next phase of microbiome research.

Suspect and non-target screening: the last frontier in environmental analysis

Suspect and non-target screening (SNTS) techniques are arising as new analytical strategies useful to disentangle the environmental occurrence of the thousands of exogenous chemicals present in our ecosystems. The unbiased discovery of the wide number of substances present over environmental analysis needs to find a consensus with powerful technical and computational requirements, as well as with the time-consuming unequivocal identification of discovered analytes. Within these boundaries, the potential applications of SNTS include the studies of environmental pollution in aquatic, atmospheric, solid and biological samples, the assessment of new compounds, transformation products and metabolites, contaminant prioritization, bioremediation or soil/water treatment evaluation, and retrospective data analysis, among many others. In this review, we evaluate the state of the art of SNTS techniques going over the normalized workflow from sampling and sample treatment to instrumental analysis, data processing and a brief review of the more recent applications of SNTS in environmental occurrence and exposure to xenobiotics. The main issues related to harmonization and knowledge gaps are critically evaluated and the challenges of their implementation are assessed in order to ensure a proper use of these promising techniques in the near future.