Biology subjects

Liu, Y.-Y.

Publications and source records attributed to Liu, Y.-Y..

At least 37 records · Page 2Linked to original sources

bioRxiv2023

Predicting metabolic response to dietary intervention using deep learning

Due to highly personalized biological and lifestyle characteristics, different individuals may have different metabolite responses to specific foods and nutrients. In particular, the gut microbiota, a collection of trillions of microorganisms living in the gastrointestinal tract, is highly personalized and plays a key role in the metabolite responses to foods and nutrients. Accurately predicting metabolite responses to dietary interventions based on individuals gut microbial compositions holds great promise for precision nutrition. Existing prediction methods are typically limited to traditional machine learning models. Deep learning methods dedicated to such tasks are still lacking. Here we develop a method McMLP (Metabolite response predictor using coupled Multilayer Perceptrons) to fill in this gap. We provide clear evidence that McMLP outperforms existing methods on both synthetic data generated by the microbial consumer-resource model and real data obtained from six dietary intervention studies. Furthermore, we perform sensitivity analysis of McMLP to infer the tripartite food-microbe-metabolite interactions, which are then validated using the ground-truth (or literature evidence) for synthetic (or real) data, respectively. The presented tool has the potential to inform the design of microbiota-based personalized dietary strategies to achieve precision nutrition.

microbiology↗

bioRxiv2022

Association Rule Mining of the Human Gut Microbiome

The human gut carries a vast and diverse microbial community that is essential for human health. Understanding the structure of this complex community requires quantitative approaches. Traditional co-occurrence and correlation analyses typically focus on pair-wise relationships and ignore higher-order relationships. Association rule mining (ARM) is a well-developed technique in data mining and has been applied to human microbiome data to identify higher-order relationships. Yet, existing attempts suffer from small sample sizes and low taxonomic resolution. Here we leverage the curatedMetagenomic Database (CMD) to resolve those issues. We first infer association rules from gut microbiome samples of a large cohort of healthy individuals (n=2,815) in CMD. Then we compare those rules with that inferred from samples of individuals with different diseases: Inflammatory Bowel Disease (IBD, n=768), Colorectal cancer (CRC, n=368), Impaired Glucose Tolerance (IGT, n=199), and Type 2 Diabetes (T2D, n=164). Finally, we demonstrate that using ARM as a feature selection tool can improve the performance of microbiome-based disease classification. Together, this study provides a comprehensive study of higher-order microbial relationships in the human gut microbiome and highlights the importance of incorporating association rules in microbiome-based disease classification.

microbiology↗

bioRxiv2022

Pairing Metagenomics and Metaproteomics to Pinpoint Ecological Niches and Metabolic Essentiality of Microbial Communities

The genome of a microorganism encodes its potential functions that can be implemented through expressed proteins. It remains elusive how a proteins selective expression depends on its metabolic essentiality to microbial growth or its ability to claim resources as ecological niches. To reveal a proteins metabolic or ecological role, we developed a computational pipeline, which pairs metagenomics and metaproteomics data to quantify each proteins gene-level and protein-level functional redundancy simultaneously. We first illustrated the idea behind the pipeline using simulated data of a consumer-resource model. We then validated it using real data from human and mouse gut microbiome samples. In particular, we analyzed ABC-type transporters and ribosomal proteins, confirming that the metabolic and ecological roles predicted by our pipeline agree well with prior knowledge. Finally, we performed in vitro cultures of a human gut microbiome sample and investigated how oversupplying various sugars involved in ecological niches influences the community structure and protein abundance. The presented results demonstrate the performance of our pipeline in identifying proteins metabolic and ecological roles, as well as its potential to help us design nutrient interventions to modulate the human microbiome.

ecology↗

bioRxiv2022

Eliminate false positives in metagenomic profiling based on type IIB restriction sites

Accurate species identification and abundance estimation are critical for the interpretation of whole metagenome shotgun sequencing (WMS) data. Numerous computational methods, broadly referred to as metagenomic profilers, have been developed to identify species in microbiome samples by classification of sequencing reads and quantification of their relative abundances. Yet, existing metagenomic profilers typically suffer from false positive identifications and consequently biased relative abundance estimation (as false positives can be accounted for more than 90% of total identified species). Here, we present a new metagenomic profiler MAP2B (MetAgenomic Profiler based on type IIB restriction site) to resolve those issues. We first illustrate the pitfalls of using relative abundance as the only feature in determining false positives. We then propose a feature set to distinguish false positives from true positives. By benchmarking the performance in metagenomic profiling using data from CAMI2 (Critical Assessment of Metagenome Interpretation: second round of challenge), we illustrate the superior performance of MAP2B (F1 score ~ 0.93) over existing metagenomic profilers (F1 score ranges from 0.18 to 0.58). We further tested the performance of MAP2B using real WMS data from an ATCC mock community, confirming its superior performance and robustness against sequencing depth. In addition, by leveraging WMS data from an IBD cohort, we demonstrate the taxonomic features obtained by MAP2B can better discriminate disease status and predict metabolomic profiles.

bioinformatics↗

bioRxiv2022

Teasing out Missing Reactions in Genome-scale Metabolic Networks through Deep Learning

GEnome-scale Metabolic models (GEMs) are powerful tools to predict cellular metabolism and physiological states in living organisms. However, due to our imperfect knowledge of metabolic processes, even highly curated GEMs have knowledge gaps (e.g., missing reactions). Existing gap-filling methods typically require phenotypic data as input to tease out missing reactions. We still lack a computational method for rapid and accurate gap-filling of metabolic networks before experimental data is available. Here we present a deep learning-based method -- CHEbyshev Spectral HyperlInk pREdictor (CHESHIRE) -- to predict missing reactions in GEMs purely from metabolic network topology. We demonstrate that CHESHIRE outperforms other topology-based methods in predicting artificially removed reactions over 926 high- and intermediate-quality GEMs. Furthermore, CHESHIRE is able to improve the phenotypic predictions of 49 draft GEMs for fermentation products and amino acids secretions. Both types of validation suggest that CHESHIRE is a powerful tool for GEM curation to reveal unknown links between reactions and observed metabolic phenotypes.

bioinformatics↗

bioRxiv2022

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 new 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.

microbiology↗

bioRxiv2022

Gut feelings: Associations of emotions and emotion regulation with the gut microbiome in women

Accumulating evidence suggests that positive and negative emotions, as well as emotion regulation processes, influence the onset and the progression of multiple human diseases. However, the underlying mechanisms connecting these emotion-related factors with physical health are not fully understood. Recent work linking the gut microbiome with both mental and physical health suggests it may be a potential pathway. Yet, its association with emotions and emotion regulation are understudied. In this study, we examined whether positive and negative emotions, as well as two emotion regulation strategies (i.e., cognitive reappraisal and suppression), were associated with the diversity, overall structure, and specific species/pathways of the gut microbiome in 206 healthy women. We found that the alpha diversity was negatively associated with suppression. Moreover, using multivariate analysis, we found that positive emotions were inversely associated with the relative abundance of Firmicutes bacterium CAG 94 and Ruminococcaceae bacterium D16, while negative emotions were directly related to relative abundance of these same species. Moreover, we found associations of emotions and emotion regulation with microbial metabolic pathways. For example, negative emotions were inversely related to biosynthesis of pantothenate, coenzyme A, and adenosine. Taken together, our findings offer human evidence supporting linkages of emotions and related regulatory processes with the gut microbiome. These findings highlight the critical importance of incorporating the human gut microbiome in our understanding of emotion-related factors and their associations with physical health.

microbiology↗

bioRxiv2022

Ecological Dynamics Imposes Fundamental Challenges in Microbial Source Tracking

Quantifying the contributions of possible environmental sources ("sources") to a specific microbial community ("sink") is a classical problem in microbiology known as microbial source tracking (MST). Solving the MST problem will not only help us understand how microbial communities were formed, but also have far-reaching applications in pollution control, public health, and forensics. Numerous computational methods, referred to as MST solvers hereafter, have been developed in the past and applied to various real datasets to demonstrate their utility across different contexts. Yet, those MST solvers do not consider microbial interactions and priority effects in microbial communities. Here, we revisit the performance of several representative MST solvers. We show compelling evidence that solving the MST problem using existing MST solvers is impractical when ecological dynamics plays a role in community assembly. In particular, we clearly demonstrate that the presence of either microbial interactions or priority effects will render the MST problem mathematically unsolvable for any MST solver. We further analyze data from fecal microbiota transplantation studies, finding that the state-of-the-art MST solvers fail to identify donors for most of the recipients. Finally, we perform community coalescence experiments to demonstrate that the state-of-the-art MST solvers fail to identify the sources for most of the sinks. Our findings suggest that ecological dynamics imposes fundamental challenges in solving the MST problem using computational approaches.

microbiology↗

bioRxiv2022

Dissecting the Role of the Human Microbiome in COVID-19 via Metagenome-assembled Genomes

Coronavirus disease 2019 (COVID-19), primarily a respiratory disease caused by infection with Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), is often accompanied by gastrointestinal symptoms. However, little is known about the relation between the human microbiome and COVID-19, largely due to the fact that previous studies fail to provide high taxonomic resolution to identify microbes that likely interact with SARS-CoV-2 infection. Here we used whole-metagenome shotgun sequencing data together with assembly and binning strategies to reconstruct metagenome-assembled genomes (MAGs) from a total of 514 nasopharyngeal and fecal samples of patients with COVID-19 and controls. We reconstructed a total of 11,584 medium-and high-quality microbial MAGs and obtained 5,403 non-redundant MAGs (nrMAGs) with strain-level resolution. We found that, thanks to the high taxonomic resolution of nrMAGs, the gut microbiome signatures can accurately distinguish COVID-19 cases from healthy controls and predict the progression of COVID-19. Moreover, we identified a set of nrMAGs with a putative causal role in the clinical manifestations of COVID-19 and revealed their functional pathways that potentially interact with SARS-CoV-2 infection. The presented results highlight the importance of incorporating the human gut microbiome in our understanding of SARS-CoV-2 infection and disease progression. The genomic content of nrMAGs presented here has the potential to inform microbiome-based therapeutic developments for COVID-19 progression and post-COVID conditions.

microbiology↗

bioRxiv2022

StrainPanDA: linked reconstruction of strain composition and gene content profiles via pangenome-based decomposition of metagenomic data

BackgroundMicrobial strains of variable functional capacities co-exist in microbiomes. Current bioinformatics methods of strain analysis cannot provide the direct linkage between strain composition and their gene contents from metagenomic data. MethodsHere we present StrainPanDA (Strain-level Pangenome Decomposition Analysis), a novel method that uses the pangenome coverage profile of multiple metagenomic samples to simultaneously reconstruct the composition and gene content variation of co-existing strains in microbial communities. ResultsWe systematically validate the accuracy and robustness of StrainPanDA using synthetic datasets. To demonstrate the power of gene-centric strain profiling, we then apply StrainPanDA to analyze the gut microbiome samples of infants, as well as patients treated with fecal microbiota transplantation. We show that the linked reconstruction of strain composition and gene content profiles is critical for understanding the relationship between microbial adaptation and strain-specific functions (e.g., nutrient utilization, pathogenicity). ConclusionsStrainPanDA can be applied to metagenomic datasets to detect association between molecular functions and microbial/host phenotypes to formulate testable hypotheses and gain novel biological insights at the strain or subspecies level.

bioinformatics↗

bioRxiv2022

A Statistical Physics Approach for Disease Module Detection

Extensive evidence indicates that the pathobiological processes of a complex disease are associated with perturbation within specific disease neighborhoods of the human protein-protein interaction (PPI) network (a.k.a. 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 with genome-wide association studies (GWAS) of six 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.

systems biology↗

bioRxiv2022

A Computational Method to Dissect Colonization Resistance of the Gut Microbiota against Pathogens

The indigenous gut microbes have co-evolved with their hosts for millions of years. Those gut microbes play a crucial role in host health and disease. In particular, they protect the host against incursion by exogenous and often harmful microorganisms, a mechanism known as colonization resistance (CR). Yet, identifying the exact microbes responsible for the gut microbiota-mediated CR against a particular pathogen remains a fundamental challenge in microbiome research. Here, we develop a computational method --- Generalized Microbe-Phenotype Triangulation (GMPT) to systematically identify causal microbes that directly influence the microbiota-mediated CR against a pathogen. We systematically validate GMPT using a classical population dynamics model in community ecology, and then apply it to microbiome data from two mouse studies on C. difficile infection. The developed method will not only significantly advance our understanding of CR mechanisms but also pave the way for the rational design of microbiome-based therapies for preventing and treating enteric infections.

bioinformatics↗

bioRxiv2021

Higher-order interaction of brain microstructural and functional connectome

Despite a relatively fixed anatomical structure, the human brain can support rich cognitive functions, triggering particular interest in investigating structure-function relationships. Myelin is a vital brain microstructure marker, yet the individual microstructure-function relationship is poorly understood. Here, we explore the brain microstructure-function relationships using a higher-order framework. Global (network-level) higher-order microstructure-function relationships negatively correlate with male participants personality scores and decline with aging. Nodal (node-level) higher-order microstructure-function relationships are not aligned uniformly throughout the brain, being stronger in association cortices and lower in sensory cortices, showing gender differences. Notably, higher-order microstructure-function relationships are maintained from the whole-brain to local circuits, which uncovers a compelling and straightforward principle of brain structure-function interactions. Additionally, targeted artificial attacks can disrupt these higher-order relationships, and the main results are robust against several factors. Together, our results increase the collective knowledge of higher-order structure-function interactions that may underlie cognition, individual differences, and aging.

neuroscience↗

bioRxiv2021

Assessment of community efforts to advance computational prediction of protein-protein interactions

Comprehensive insights from the human protein-protein interaction (PPI) network, known as 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 new PPIs. Many such approaches have been proposed. However, 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 24 representative network-based methods to predict PPIs across five different interactomes, including a synthetic interactome generated by the duplication-mutation-complementation model, and the interactomes of four different organisms: A. thaliana, C. elegans, S. cerevisiae, and H. sapiens. We selected the top-seven methods through a computational validation on the human interactome. We next experimentally validated their top-500 predicted PPIs (in total 3,276 predicted PPIs) using the yeast two-hybrid assay, finding 1,177 new human PPIs (involving 633 proteins). Our results indicate that task-tailored similarity-based methods, which leverage the underlying network characteristics of PPIs, show superior performance over other general link prediction methods. Through experimental validation, we confirmed that the top-ranking methods show promising performance externally. For example, from the top 500 PPIs predicted by an advanced similarity-base method [MPS(B&T)], 430 were successfully tested by Y2H with 376 testing positive, yielding a precision of 87.4%. These results establish advanced similarity-based methods as powerful tools for the prediction of human PPIs.

systems biology↗

bioRxiv2021

Ecological dynamics of the gut microbiome in response to dietary fiber

Dietary fibers are generally thought to benefit intestinal health. Their impacts on the composition and metabolic function of the gut microbiome, however, vary greatly across individuals. Previous research showed that each individuals response to fibers depends on their baseline gut microbiome, but the ecology driving microbiota remodeling during fiber intake remained unclear. Here, we studied the long-term dynamics of gut microbiome and short-chain fatty acids (SCFAs) in isogenic mice with distinct microbiota baselines fed with the fermentable fiber inulin compared to the non-fermentable fiber cellulose. We found that inulin produced generally rapid response followed by gradual stabilization to new equilibria, and those dynamics were baseline-dependent. We parameterized an ecology model from the timeseries data, which revealed a group of bacteria whose growth significantly increases in response to inulin. and whose baseline abundance and interspecies competition explains the baseline-dependence of microbiome density and community composition dynamics. Fecal levels of of SCFAs, such as propionate, is associated with the abundance of inulin responders, yet inter-individual variation of gut microbiome impedes the prediction of SCFAs by machine learning models. Finally, we showed that our methods and major findings are generalizable to dietary resistant starch. This study emphasizes the importance of ecological modeling to understand microbiome responses to dietary changes and the need for personalized interventions.

ecology↗

bioRxiv2021

Revealing Protein-Level Functional Redundancy in the Human Gut Microbiome using Ultra-deep Metaproteomics

Functional redundancy is a key property of ecosystems and represents the fact that phylogenetically unrelated taxa can play similar functional roles within an ecosystem. The redundancy of potential functions of human microbiome has been recently quantified using metagenomics data. Yet, the redundancy of functions which are actually expressed within the human microbiome remains largely unexplored. Here, we quantify the protein-level functional redundancy in the human gut microbiome using metaproteomics and network approaches. In particular, our ultra-deep metaproteomics approach revealed high protein-level functional redundancy and high nestedness in proteomic content networks - bipartite graphs that connect taxa with their expressed functions. We further examined multiple metaproteomics datasets and showed that various environmental factors, including individuality, biogeography, xenobiotics, and disease, significantly altered the protein-level functional redundancy. Finally, by projecting the bipartite proteomic content networks into unipartite weighted genus networks, functional hub genera across individual microbiomes were discovered, suggesting that there may be a universal principle of functional organization in microbiome assembly. HighlightsO_LIUltra-deep metaproteomics reveals high protein-level functional redundancy in the human gut microbiome C_LIO_LIWithin-sample proteomic content networks display universal topology C_LIO_LIVarious environmental factors influence the redundancy of expressed functions C_LIO_LIFunctional hub genera are present across different datasets C_LI

systems biology↗

bioRxiv2021

Gut Microbiota predicts Healthy Late-life Aging in Male Mice

Calorie restriction (CR) extends lifespan and retards age-related chronic diseases in most species. There is growing evidence that the gut microbiota has a pivotal role in host health and age-related pathological conditions. Yet, it is still unclear how CR and the gut microbiota are related to healthy aging. Here we report findings from a small longitudinal study of male C57BL/6 mice maintained on either ad libitum or mild (15%) CR diets from 21 months of age and tracked until natural death. We demonstrate that CR results in a significant reduction in frailty index (FI), a well-established indicator of aging. We observed significant alterations in bacterial load, diversity, and compositional patterns of the mouse gut microbiota during the aging process. Interrogating the FI-related microbial features using machine learning techniques, we show that gut microbial signatures from 21-month-old mice can predict the healthy aging of 30-month-old mice with reasonable accuracy. This study deepens our understanding of the links between CR, gut microbiota, and frailty in the aging process of mice.

microbiology↗

bioRxiv2021

Predicting microbiome compositions 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, here we present a deep learning framework that automatically learns the map between species assemblages and community compositions from training data only, without knowledge of any of the above processes. First, we systematically validate our framework using synthetic data generated by classical population dynamics models. Then, we apply it to experimental data of both in vitro and in vivo communities, including ocean and soil microbial communities, Drosophila melanogaster gut microbiota, and human gut and oral microbiota. In particular, we show how 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.

microbiology↗