Data-driven prediction of colonization outcomes for complex microbial communities.

Microbial interactions can lead to different colonization outcomes of exogenous species, be they pathogenic or beneficial in nature. Predicting the colonization of exogenous species in complex communities remains a fundamental challenge in microbial ecology, mainly due to our limited knowledge of the diverse mechanisms governing microbial dynamics. Here, we propose a data-driven approach independent of any dynamics model to predict colonization outcomes of exogenous species from the baseline compositions of microbial communities. We systematically validate this approach using synthetic data, finding that machine learning models can predict not only the binary colonization outcome but also the post-invasion steady-state abundance of the invading species. Then we conduct colonization experiments for commensal gut bacteria species Enterococcus faecium and Akkermansia muciniphila in hundreds of human stool-derived in vitro microbial communities, confirming that the data-driven approaches can predict the colonization outcomes in experiments. Furthermore, we find that while most resident species are predicted to have a weak negative impact on the colonization of exogenous species, strongly interacting species could significantly alter the colonization outcomes, e.g., Enterococcus faecalis inhibits the invasion of E. faecium invasion. The presented results suggest that the data-driven approaches are powerful tools to inform the ecology and management of microbial communities.

Identifying keystone species in microbial communities using deep learning.

Previous studies suggested that microbial communities can harbour keystone species whose removal can cause a dramatic shift in microbiome structure and functioning. Yet, an efficient method to systematically identify keystone species in microbial communities is still lacking. Here we propose a data-driven keystone species identification (DKI) framework based on deep learning to resolve this challenge. Our key idea is to implicitly learn the assembly rules of microbial communities from a particular habitat by training a deep-learning model using microbiome samples collected from this habitat. The well-trained deep-learning model enables us to quantify the community-specific keystoneness of each species in any microbiome sample from this habitat by conducting a thought experiment on species removal. We systematically validated this DKI framework using synthetic data and applied DKI to analyse real data. We found that those taxa with high median keystoneness across different communities display strong community specificity. The presented DKI framework demonstrates the power of machine learning in tackling a fundamental problem in community ecology, paving the way for the data-driven management of complex microbial communities.

Nutritional redundancy in the human diet and its application in phenotype association studies.

Studying human dietary intake may help us identify effective measures to treat or prevent many chronic diseases whose natural histories are influenced by nutritional factors. Here, by examining five cohorts with dietary intake data collected on different time scales, we show that the food intake profile varies substantially across individuals and over time, while the nutritional intake profile appears fairly stable. We refer to this phenomenon as 'nutritional redundancy' and attribute it to the nested structure of the food-nutrient network. This network enables us to quantify the level of nutritional redundancy for each diet assessment of any individual. Interestingly, this nutritional redundancy measure does not strongly correlate with any classical healthy diet scores, but its performance in predicting healthy aging shows comparable strength. Moreover, after adjusting for age, we find that a high nutritional redundancy is associated with lower risks of cardiovascular disease and type 2 diabetes.

Data-driven prediction of colonization outcomes for complex microbial communities.

Complex microbial interactions can lead to different colonization outcomes of exogenous species, be they pathogenic or beneficial in nature. Predicting the colonization of exogenous species in complex communities remains a fundamental challenge in microbial ecology, mainly due to our limited knowledge of the diverse physical, biochemical, and ecological processes governing microbial dynamics. Here, we proposed a data-driven approach independent of any dynamics model to predict colonization outcomes of exogenous species from the baseline compositions of microbial communities. We systematically validated this approach using synthetic data, finding that machine learning models (including Random Forest and neural ODE) can predict not only the binary colonization outcome but also the post-invasion steady-state abundance of the invading species. Then we conducted colonization experiments for two commensal gut bacteria species Enterococcus faecium and Akkermansia muciniphila in hundreds of human stool-derived in vitro microbial communities, confirming that the data-driven approach can successfully predict the colonization outcomes. Furthermore, we found that while most resident species were predicted to have a weak negative impact on the colonization of exogenous species, strongly interacting species could significantly alter the colonization outcomes, e.g., the presence of Enterococcus faecalis inhibits the invasion of E. faecium . The presented results suggest that the data-driven approach is a powerful tool to inform the ecology and management of complex microbial communities.

Identifying chronic obstructive pulmonary disease from integrative omics and clustering in lung tissue.

BACKGROUND: Chronic obstructive pulmonary disease (COPD) is a highly morbid and heterogenous disease. While COPD is defined by spirometry, many COPD characteristics are seen in cigarette smokers with normal spirometry. The extent to which COPD and COPD heterogeneity is captured in omics of lung tissue is not known. METHODS: We clustered gene expression and methylation data in 78 lung tissue samples from former smokers with normal lung function or severe COPD. We applied two integrative omics clustering methods: (1) Similarity Network Fusion (SNF) and (2) Entropy-Based Consensus Clustering (ECC). RESULTS: SNF clusters were not significantly different by the percentage of COPD cases (48.8% vs. 68.6%, p = 0.13), though were different according to median forced expiratory volume in one second (FEV1) % predicted (82 vs. 31, p = 0.017). In contrast, the ECC clusters showed stronger evidence of separation by COPD case status (48.2% vs. 81.8%, p = 0.013) and similar stratification by median FEV1% predicted (82 vs. 30.5, p = 0.0059). ECC clusters using both gene expression and methylation were identical to the ECC clustering solution generated using methylation data alone. Both methods selected clusters with differentially expressed transcripts enriched for interleukin signaling and immunoregulatory interactions between lymphoid and non-lymphoid cells. CONCLUSIONS: Unsupervised clustering analysis from integrated gene expression and methylation data in lung tissue resulted in clusters with modest concordance with COPD, though were enriched in pathways potentially contributing to COPD-related pathology and heterogeneity.

Identifying keystone species in microbial communities using deep learning.

Previous studies suggested that microbial communities harbor keystone species whose removal can cause a dramatic shift in microbiome structure and functioning. Yet, an efficient method to systematically identify keystone species in microbial communities is still lacking. This is mainly due to our limited knowledge of microbial dynamics and the experimental and ethical difficulties of manipulating microbial communities. Here, we propose a Data-driven Keystone species Identification (DKI) framework based on deep learning to resolve this challenge. Our key idea is to implicitly learn the assembly rules of microbial communities from a particular habitat by training a deep learning model using microbiome samples collected from this habitat. The well-trained deep learning model enables us to quantify the community-specific keystoneness of each species in any microbiome sample from this habitat by conducting a thought experiment on species removal. We systematically validated this DKI framework using synthetic data generated from a classical population dynamics model in community ecology. We then applied DKI to analyze human gut, oral microbiome, soil, and coral microbiome data. We found that those taxa with high median keystoneness across different communities display strong community specificity, and many of them have been reported as keystone taxa in literature. The presented DKI framework demonstrates the power of machine learning in tackling a fundamental problem in community ecology, paving the way for the data-driven management of complex microbial communities.

Assessment of community efforts to advance network-based prediction of protein-protein interactions.

Comprehensive understanding of the human protein-protein interaction (PPI) network, aka the human interactome, can provide important insights into the molecular mechanisms of complex biological processes and diseases. Despite the remarkable experimental efforts undertaken to date to determine the structure of the human interactome, many PPIs remain unmapped. Computational approaches, especially network-based methods, can facilitate the identification of previously uncharacterized PPIs. Many such methods have been proposed. Yet, a systematic evaluation of existing network-based methods in predicting PPIs is still lacking. Here, we report community efforts initiated by the International Network Medicine Consortium to benchmark the ability of 26 representative network-based methods to predict PPIs across six different interactomes of four different organisms: A. thaliana, C. elegans, S. cerevisiae, and H. sapiens. Through extensive computational and experimental validations, we found that advanced similarity-based methods, which leverage the underlying network characteristics of PPIs, show superior performance over other general link prediction methods in the interactomes we considered.

Benchmarking omics-based prediction of asthma development in children.

BACKGROUND: Asthma is a heterogeneous disease with high morbidity. Advancement in high-throughput multi-omics approaches has enabled the collection of molecular assessments at different layers, providing a complementary perspective of complex diseases. Numerous computational methods have been developed for the omics-based patient classification or disease outcome prediction. Yet, a systematic benchmarking of those methods using various combinations of omics data for the prediction of asthma development is still lacking. OBJECTIVE: We aimed to investigate the computational methods in disease status prediction using multi-omics data. METHOD: We systematically benchmarked 18 computational methods using all the 63 combinations of six omics data (GWAS, miRNA, mRNA, microbiome, metabolome, DNA methylation) collected in The Vitamin D Antenatal Asthma Reduction Trial (VDAART) cohort. We evaluated each method using standard performance metrics for each of the 63 omics combinations. RESULTS: Our results indicate that overall Logistic Regression, Multi-Layer Perceptron, and MOGONET display superior performance, and the combination of transcriptional, genomic and microbiome data achieves the best prediction. Moreover, we find that including the clinical data can further improve the prediction performance for some but not all the omics combinations. CONCLUSIONS: Specific omics combinations can reach the optimal prediction of asthma development in children. And certain computational methods showed superior performance than other methods.

Predicting metabolomic profiles from microbial composition through neural ordinary differential equations.

Characterizing the metabolic profile of a microbial community is crucial for understanding its biological function and its impact on the host or environment. Metabolomics experiments directly measuring these profiles are difficult and expensive, while sequencing methods quantifying the species composition of microbial communities are well-developed and relatively cost-effective. Computational methods that are capable of predicting metabolomic profiles from microbial compositions can save considerable efforts needed for metabolomic profiling experimentally. Yet, despite existing efforts, we still lack a computational method with high prediction power, general applicability, and great interpretability. Here we develop a method - mNODE (Metabolomic profile predictor using Neural Ordinary Differential Equations), based on a state-of-the-art family of deep neural network models. We show compelling evidence that mNODE outperforms existing methods in predicting the metabolomic profiles of human microbiomes and several environmental microbiomes. Moreover, in the case of human gut microbiomes, mNODE can naturally incorporate dietary information to further enhance the prediction of metabolomic profiles. Besides, susceptibility analysis of mNODE enables us to reveal microbe-metabolite interactions, which can be validated using both synthetic and real data. The presented results demonstrate that mNODE is a powerful tool to investigate the microbiome-diet-metabolome relationship, facilitating future research on precision nutrition.

A statistical physics approach for disease module detection.

Extensive evidence indicates that the pathobiological processes of a complex disease are associated with perturbation in specific neighborhoods of the human protein-protein interaction (PPI) network (also known as the interactome), often referred to as the disease module. Many computational methods have been developed to integrate the interactome and omics profiles to extract context-dependent disease modules. Yet, existing methods all have fundamental limitations in terms of rigor and/or efficiency. Here, we developed a statistical physics approach based on the random-field Ising model (RFIM) for disease module detection, which is both mathematically rigorous and computationally efficient. We applied our RFIM approach to genome-wide association studies (GWAS) of ten complex diseases to examine its performance for disease module detection. We found that our RFIM approach outperforms existing methods in terms of computational efficiency, connectivity of disease modules, and robustness to the interactome incompleteness.

Predicting microbiome compositions from species assemblages through deep learning.

Microbes can form complex communities that perform critical functions in maintaining the integrity of their environment or their hosts' well-being. Rationally managing these microbial communities requires improving our ability to predict how different species assemblages affect the final species composition of the community. However, making such a prediction remains challenging because of our limited knowledge of the diverse physical, biochemical, and ecological processes governing microbial dynamics. To overcome this challenge, we present a deep learning framework that automatically learns the map between species assemblages and community compositions from training data only, without knowing any of the above processes. First, we systematically validate our framework using synthetic data generated by classical population dynamics models. Then, we apply our framework to data from in vitro and in vivo microbial communities, including ocean and soil microbiota, Drosophila melanogaster gut microbiota, and human gut and oral microbiota. We find that our framework learns to perform accurate out-of-sample predictions of complex community compositions from a small number of training samples. Our results demonstrate how deep learning can enable us to understand better and potentially manage complex microbial communities.

Discovery of a novel nonsteroidal selective glucocorticoid receptor modulator by virtual screening and bioassays.

Synthetic glucocorticoids (GCs) have been widely used in the treatment of a broad range of inflammatory diseases, but their clinic use is limited by undesired side effects such as metabolic disorders, osteoporosis, skin and muscle atrophies, mood disorders and hypothalamic-pituitary-adrenal (HPA) axis suppression. Selective glucocorticoid receptor modulators (SGRMs) are expected to have promising anti-inflammatory efficacy but with fewer side effects caused by GCs. Here, we reported HT-15, a prospective SGRM discovered by structure-based virtual screening (VS) and bioassays. HT-15 can selectively act on the NF-κB/AP1-mediated transrepression function of glucocorticoid receptor (GR) and repress the expression of pro-inflammation cytokines (i.e., IL-1ß, IL-6, COX-2, and CCL-2) as effectively as dexamethasone (Dex). Compared with Dex, HT-15 shows less transactivation potency that is associated with the main adverse effects of synthetic GCs, and no cross activities with other nuclear receptors. Furthermore, HT-15 exhibits very weak inhibition on the ratio of OPG/RANKL. Therefore, it may reduce the side effects induced by normal GCs. The bioactive compound HT-15 can serve as a starting point for the development of novel therapeutics for high dose or long-term anti-inflammatory treatment.

Integrating gut microbiome and host immune markers to understand the pathogenesis of Clostridioides difficile infection.

Clostridioides difficile (C.difficile) infection is the most common cause of healthcare-associated infection and an important cause of morbidity and mortality among hospitalized patients. A comprehensive understanding of C.difficile infection (CDI) pathogenesis is crucial for disease diagnosis, treatment, and prevention. Here, we characterized gut microbial compositions and a broad panel of innate and adaptive immunological markers in 243 well-characterized human subjects (including 187 subjects with both microbiota and immune marker data), who were divided into four phenotype groups: CDI, Asymptomatic Carriage, Non-CDI Diarrhea, and Control. We found that the interactions between gut microbiota and host immune markers are very sensitive to the status of C.difficile colonization and infection. We demonstrated that incorporating both gut microbiome and host immune marker data into classification models can better distinguish CDI from other groups than can either type of data alone. Our classification models display robust diagnostic performance to differentiate CDI from Asymptomatic carriage (AUC~0.916), Non-CDI Diarrhea (AUC~0.917), or Non-CDI that combines all other three groups (AUC~0.929). Finally, we performed symbolic classification using selected features to derive simple mathematic formulas that explicitly quantify the interactions between the gut microbiome and host immune markers. These findings support the potential roles of gut microbiota and host immune markers in the pathogenesis of CDI. Our study provides new insights for a microbiome-immune marker-derived signature to diagnose CDI and design therapeutic strategies for CDI.

Deciphering functional redundancy in the human microbiome.

Although the taxonomic composition of the human microbiome varies tremendously across individuals, its gene composition or functional capacity is highly conserved - implying an ecological property known as functional redundancy. Such functional redundancy has been hypothesized to underlie the stability and resilience of the human microbiome, but this hypothesis has never been quantitatively tested. The origin of functional redundancy is still elusive. Here, we investigate the basis for functional redundancy in the human microbiome by analyzing its genomic content network - a bipartite graph that links microbes to the genes in their genomes. We find that this network exhibits several topological features that favor high functional redundancy. Furthermore, we develop a simple genome evolution model to generate genomic content network, finding that moderate selection pressure and high horizontal gene transfer rate are necessary to generate genomic content networks with key topological features that favor high functional redundancy. Finally, we analyze data from two published studies of fecal microbiota transplantation (FMT), finding that high functional redundancy of the recipient's pre-FMT microbiota raises barriers to donor microbiota engraftment. This work elucidates the potential ecological and evolutionary processes that create and maintain functional redundancy in the human microbiome and contribute to its resilience.

Link Prediction through Deep Generative Model.

Inferring missing links based on the currently observed network is known as link prediction, which has tremendous real-world applications in biomedicine, e-commerce, social media, and criminal intelligence. Numerous methods have been proposed to solve the link prediction problem. Yet, many of these methods are designed for undirected networks only and based on domain-specific heuristics. Here we developed a new link prediction method based on deep generative models, which does not rely on any domain-specific heuristic and works for general undirected or directed complex networks. Our key idea is to represent the adjacency matrix of a network as an image and then learn hierarchical feature representations of the image by training a deep generative model. Those features correspond to structural patterns in the network at different scales, from small subgraphs to mesoscopic communities. When applied to various real-world networks from different domains, our method shows overall superior performance against existing methods.

Comparative study of classifiers for human microbiome data.

Accumulated evidence has shown that commensal microorganisms play key roles in human physiology and diseases. Dysbiosis of the human-associated microbial communities, often referred to as the human microbiome, has been associated with many diseases. Applying supervised classification analysis to the human microbiome data can help us identify subsets of microorganisms that are highly discriminative and hence build prediction models that can accurately classify unlabeled samples. Here, we systematically compare two state-of-the-art ensemble classifiers: Random Forests (RF), eXtreme Gradient Boosting decision trees (XGBoost) and two traditional methods: The elastic net (ENET) and Support Vector Machine (SVM) in the classification analysis of 29 benchmark human microbiome datasets. We find that XGBoost outperforms all other methods only in a few benchmark datasets. Overall, the XGBoost, RF and ENET display comparable performance in the remaining benchmark datasets. The training time of XGBoost is much longer than others, partially due to the much larger number of hyperparameters in XGBoost. We also find that the most important features selected by the four classifiers partially overlap. Yet, the difference between their classification performance is almost independent of this overlap.

Identifying concepts from medical images via transfer learning and image retrieval.

Automatically identifying semantic concepts from medical images provides multimodal insights for clinical research. To study the effectiveness of concept detection on large scale medical images, we reconstructed over 230,000 medical image-concepts pairs collected from the ImageCLEFcaption 2018 evaluation task. A transfer learning-based multi-label classification model was used to predict multiple high-frequency concepts for medical images. Semantically relevant concepts of visually similar medical images were identified by the image retrieval-based topic model. The results showed that the transfer learning method achieved F1 score of 0.1298, which was comparable with the state of art methods in the ImageCLEFcaption tasks. The image retrieval-based method contributed to the recall performance but reduced the overall F1 score, since the retrieval results of the search engine introduced irrelevant concepts. Although our proposed method achieved second-best performance in the concept detection subtask of ImageCLEFcaption 2018, there will be plenty of further work to improve the concept detection with better understanding the medical images.

Control energy of complex networks towards distinct mixture states.

Controlling complex networked systems is a real-world puzzle that remains largely unsolved. Despite recent progress in understanding the structural characteristics of network control energy, target state and system dynamics have not been explored. We examine how varying the final state mixture affects the control energy of canonical and conformity-incorporated dynamical systems. We find that the control energy required to drive a network to an identical final state is lower than that required to arrive a non-identical final state. We also demonstrate that it is easier to achieve full control in a conformity-based dynamical network. Finally we determine the optimal control strategy in terms of the network hierarchical structure. Our work offers a realistic understanding of the control energy within the final state mixture and sheds light on controlling complex systems.

Repeat HIV Testing and Incident Rates among Individuals Attending Voluntary Counseling and Testing Clinics in Wuxi, China: A Retrospective Study.

OBJECTIVE: We aimed to elucidate the rates of repeat HIV testing and incident HIV diagnosis, and baseline CD4+ T cell count among individuals attending HIV voluntary counseling and testing (VCT) clinics in Wuxi, China. METHODS: A repeat HIV testing within 12 months was recorded if individuals had their first test with negative results, during 2013-2014 and retested within 12 months. An incident HIV diagnosis was recorded if individuals had their first test with negative results, during 2013-2015 and had a subsequent positive result at any point by the end of 2015. Data on HIV testing and diagnosis among individuals attending 32 VCT clinics from 2013 to 2015 and HIV diagnosis from other clinical services in Wuxi, China, were retrieved. A multivariate logistic regression model was used to analyze factors associated with repeat HIV testing. Cox regression was used to evaluate factors associated with incident HIV diagnosis. RESULTS: From 2013 to 2014, 11,504 individuals tested HIV negative at their first recorded test, with 655 (5.7%) retesting within 12 months. Higher repeat HIV testing within 12 months was associated with male gender [adjusted odds ratio (aOR) = 1.7, 95% confidence interval (CI): 1.4-2.2], risk behaviors [commercial heterosexual behaviors (aOR = 1.4, CI: 1.1-1.6), male-male sexual behaviors (aOR = 3.7, CI: 2.7-4.9)], injection drug use (aOR = 9.9, CI: 6.5-15.1), and having taken HIV tests previously (aOR = 2.0, CI: 1.6-2.4). From 2013 to 2015, 1,088 individuals tested negative on HIV test at their visit and at ⋝ 2 subsequent tests; of them 30 had incident HIV diagnosis. The overall rate of incident HIV diagnosis among all VCT individuals was 1.6 (95% CI: 1.1-2.1) per 100 person-years. Incident HIV diagnosis was associated with male gender [adjusted hazard ratio (aHR) = 8.5, 95% CI: 1.9-38.1], attending hospital-based VCT clinics (aHR = 7.8, 95% CI: 1.1-58.3), and male-male sexual behavior (aHR = 8.4, 95% CI: 1.5-46.7). Individuals diagnosed at VCT clinics had higher CD4+ T cell count compared with those diagnosed at other clinical services (median 407 vs. 326 copies/mm3, P = 0.003). CONCLUSION: VCT individuals in Wuxi, China, had a low repeat HIV testing rate and high HIV incidence. VCT-clinic-based interventions aimed at increasing repeat HIV testing are needed to detect more cases at an earlier stage, especially among individuals at high risk for HIV infection such as men who have sex with men.

Effect of correlations on controllability transition in network control.

The network control problem has recently attracted an increasing amount of attention, owing to concerns including the avoidance of cascading failures of power-grids and the management of ecological networks. It has been proven that numerical control can be achieved if the number of control inputs exceeds a certain transition point. In the present study, we investigate the effect of degree correlation on the numerical controllability in networks whose topological structures are reconstructed from both real and modeling systems, and we find that the transition point of the number of control inputs depends strongly on the degree correlation in both undirected and directed networks with moderately sparse links. More interestingly, the effect of the degree correlation on the transition point cannot be observed in dense networks for numerical controllability, which contrasts with the corresponding result for structural controllability. In particular, for directed random networks and scale-free networks, the influence of the degree correlation is determined by the types of correlations. Our approach provides an understanding of control problems in complex sparse networks.