To address the limitation of overlooking crucial ecological interactions due to relying on single time point samples, we developed a computational approach that analyzes individual samples based on the interspecific microbial relationships. We verify, using both numerical simulations as well as real and shuffled microbial profiles from the human oral cavity, that the method can classify single samples based on their interspecific interactions. By analyzing the gut microbiome of people with autistic spectrum disorder, we found that our interaction-based method can improve the classification of individual subjects based on a single microbial sample. These results demonstrate that the underlying ecological interactions can be practically utilized to facilitate microbiome-based diagnosis and precision medicine.

暂无摘要。

Microbiome biomarker discovery for patient diagnosis, prognosis, and risk evaluation is attracting broad interest. Selected groups of microbial features provide signatures that characterize host disease states such as cancer or cardio-metabolic diseases. Yet, the current predictive models stemming from machine learning still behave as black boxes and seldom generalize well. Their interpretation is challenging for physicians and biologists, which makes them difficult to trust and use routinely in the physician-patient decision-making process. Novel methods that provide interpretability and biological insight are needed. Here, we introduce \"predomics\", an original machine learning approach inspired by microbial ecosystem interactions that is tailored for metagenomics data. It discovers accurate predictive signatures and provides unprecedented interpretability. The decision provided by the predictive model is based on a simple, yet powerful score computed by adding, subtracting, or dividing cumulative abundance of microbiome measurements.
Tested on >100 datasets, we demonstrate that predomics models are simple and highly interpretable. Even with such simplicity, they are at least as accurate as state-of-the-art methods. The family of best models, discovered during the learning process, offers the ability to distil biological information and to decipher the predictability signatures of the studied condition. In a proof-of-concept experiment, we successfully predicted body corpulence and metabolic improvement after bariatric surgery using pre-surgery microbiome data.
Predomics is a new algorithm that helps in providing reliable and trustworthy diagnostic decisions in the microbiome field. Predomics is in accord with societal and legal requirements that plead for an explainable artificial intelligence approach in the medical field.

Dispersal and environmental selection are two of the most important factors that govern the distributions of microbial communities in nature. While dispersal rates are often inferred by measuring the degree to which community similarity diminishes with increasing geographic distance, determining the extent to which environmental selection impacts the distribution of microbes is more complex. To address this knowledge gap, we performed a large reciprocal transplant experiment to simulate the dispersal of US East Coast salt marsh Spartina alterniflora rhizome-associated microbial sediment communities across a latitudinal gradient and determined if any shifts in microbial community composition occurred as a result of the transplantation. Using bacterial 16S rRNA gene sequencing, we did not observe large-scale changes in community composition over a five-month S. alterniflora summer growing season and found that transplanted communities more closely resembled their origin sites than their destination sites. Furthermore, transplanted communities grouped predominantly by region, with two sites from the north and three sites to the south hosting distinct bacterial taxa, suggesting that sediment communities transplanted from north to south tended to retain their northern microbial distributions, and south to north maintained a southern distribution. A small number of potential indicator 16S rRNA gene sequences had distributions that were strongly correlated to both temperature and nitrogen, indicating that some organisms are more sensitive to environmental factors than others. These results provide new insight into the microbial biogeography of salt marsh sediments and suggest that established bacterial communities in frequently-inundated environments may be both highly resistant to invasion and resilient to some environmental shifts. However, the extent to which environmental selection impacts these communities is taxon specific and variable, highlighting the complex interplay between dispersal and environmental selection for microbial communities in nature.

Many microbial ecosystems can be seen as microbial \'food chains\' where the different reaction steps can be seen as such: the waste products of the organisms at a given reaction step are consumed by organisms at the next reaction step. In the present paper we study a model of a two-step biological reaction with feedback inhibition, which was recently presented as a reduced and simplified version of the anaerobic digestion model ADM1 of the International Water Association (IWA). It is known that in the absence of maintenance (or decay) the microbial \'food chain\' is stable. In a previous study, using a purely numerical approach and ADM1 consensus parameter values, it was shown that the model remains stable when decay terms are added. However, the authors could not prove in full generality that it remains true for other parameter values. In this paper we prove that introducing decay in the model preserves stability whatever its parameters values are and for a wide range of kinetics.

The rapid expansion of new sequencing technologies has enabled large-scale functional exploration of numerous microbial ecosystems, by establishing catalogs of functional genes and by comparing their prevalence in various microbiota. However, sequence similarity does not necessarily reflect functional conservation, since just a few modifications in a gene sequence can have a strong impact on the activity and the specificity of the corresponding enzyme or the recognition for a sensor. Similarly, some microorganisms harbor certain identified functions yet do not have the expected related genes in their genome. Finally, there are simply too many protein families whose function is not yet known, even though they are highly abundant in certain ecosystems. In this context, the discovery of new protein functions, using either sequence-based or activity-based approaches, is of crucial importance for the discovery of new enzymes and for improving the quality of annotation in public databases. This paper lists and explores the latest advances in this field, along with the challenges to be addressed, particularly where microfluidic technologies are concerned.