TMBstable: a variant caller controls performance variation across heterogeneous sequencing samples.

In cancer genomics, variant calling has advanced, but traditional mean accuracy evaluations are inadequate for biomarkers like tumor mutation burden, which vary significantly across samples, affecting immunotherapy patient selection and threshold settings. In this study, we introduce TMBstable, an innovative method that dynamically selects optimal variant calling strategies for specific genomic regions using a meta-learning framework, distinguishing it from traditional callers with uniform sample-wide strategies. The process begins with segmenting the sample into windows and extracting meta-features for clustering, followed by using a pre-trained meta-model to select suitable algorithms for each cluster, thereby addressing strategy-sample mismatches, reducing performance fluctuations and ensuring consistent performance across various samples. We evaluated TMBstable using both simulated and real non-small cell lung cancer and nasopharyngeal carcinoma samples, comparing it with advanced callers. The assessment, focusing on stability measures, such as the variance and coefficient of variation in false positive rate, false negative rate, precision and recall, involved 300 simulated and 106 real tumor samples. Benchmark results showed TMBstable's superior stability with the lowest variance and coefficient of variation across performance metrics, highlighting its effectiveness in analyzing the counting-based biomarker. The TMBstable algorithm can be accessed at https://github.com/hello-json/TMBstable for academic usage only.

The role of apparent diffusion coefficient in the grading of adult isocitrate dehydrogenase-mutant astrocytomas: relationship with the Ki-67 proliferation index.

BACKGROUND: The grading of adult isocitrate dehydrogenase (IDH)-mutant astrocytomas is a crucial prognostic factor. PURPOSE: To investigate the value of conventional magnetic resonance imaging (MRI) features and apparent diffusion coefficient (ADC) in the grading of adult IDH-mutant astrocytomas, and to analyze the correlation between ADC and the Ki-67 proliferation index. MATERIAL AND METHODS: The clinical and MRI data of 82 patients with adult IDH-mutant astrocytoma who underwent surgical resection and molecular genetic testing with IDH and 1p/19q were retrospectively analyzed. The conventional MRI features, ADCmin, ADCmean, and nADC of the tumors were compared using the Kruskal-Wallis single factor ANOVA and chi-square tests. Receiver operating characteristic (ROC) curves were drawn to evaluate conventional MRI and ADC accuracy in differentiating tumor grades. Pearson correlation analysis was performed to determine the correlation between ADC and the Ki-67 proliferation index. RESULTS: The difference in enhancement, ADCmin, ADCmean, and nADC among WHO grade 2, 3, and 4 tumors was statistically significant (all P <0.05). ADCmin showed the preferable diagnostic accuracy for grading WHO grade 2 and 3 tumors (AUC=0.724, sensitivity=63.4%, specificity=80%, positive predictive value (PPV)=62.0%; negative predictive value (NPV)=82.5%), and distinguishing grade 3 from grade 4 tumors (AUC=0.764, sensitivity=70%, specificity=76.2%, PPV=75.0%, NPV=71.4%). Enhancement + ADC model showed an optimal predictive accuracy (grade 2 vs. 3: AUC = 0.759; grade 3 vs. 4: AUC = 0.799). The Ki-67 proliferation index was negatively correlated with ADCmin, ADCmean, and nADC (all P <0.05), and positively correlated with tumor grade. CONCLUSION: Conventional MRI features and ADC are valuable to predict pathological grading of adult IDH-mutant astrocytomas.

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.

Perceptions about respiratory syncytial virus (RSV) and attitudes toward the RSV vaccine among the general public in China: A cross-sectional survey.

Our study aims to assess the public's perceptions of respiratory syncytial virus (RSV) and attitudes toward the RSV vaccine and to identify associated factors in China. A nationwide cross-sectional survey conducted using an online platform between August 16 and September 14, 2023. Questions related to socio-demographics, awareness, knowledge, perceptions of susceptibility and severity of RSV, and attitudes toward the RSV vaccine were included in the questionnaire. We used the chi-square test and logistic regression model to explore the associated factors. Overall, 2133 individuals were included in this study. Nearly a quarter of participants (24.3%) indicated that they had never heard of RSV. The proportion of individuals aged over 50 years reporting never having heard of RSV (36.5%) and having a low knowledge level of RSV (55.3%) was significantly higher that of other younger age groups. More than half of individuals (55.7%) exhibited low level of perceptions of susceptibility concerning RSV infection. A total of 68.4% of the participants expressed willingness to receive the RSV vaccine. Younger age was positively associated with a higher willingness to be vaccinated. The most frequent reason for declining the vaccine was "Concern about vaccine's safety or side effects." About 60% of individuals considered a price of RSV vaccine below 200 CNY (28 USD) as acceptable. The awareness and perceived susceptibility to RSV infection were limited to the Chinese public. It is necessary to take measures to address the low awareness and knowledge of RSV and acceptability of the RSV vaccine among older adults.

Influence of Femtosecond Laser Surface Modification on Tensile Properties of Titanium Alloy.

Titanium alloy components often experience damage from impact loads during usage, which makes improving the mechanical properties of TC4 titanium alloys crucial. This paper investigates the influence of laser scanning irradiation on the tensile properties of thin titanium alloy sheets. Results indicate that the tensile strength of thin titanium alloy sheets exhibits a trend of initial increase followed by a decrease. Different levels of enhancement are observed in the elongation at break of a cross-section. Optimal improvement in the elongation at break is achieved when the laser fluence is around 8 J/cm2, while the maximum increase in tensile strength occurs at approximately 10 J/cm2. Using femtosecond laser surface irradiation, this study compares the maximum enhancement in the tensile strength of titanium alloy base materials, which is approximately 8.54%, and the maximum increase in elongation at break, which reaches 25.61%. In addition, the results verify that cracks in tensile fractures of TC4 start from the middle, while laser-induced fracture cracks occur from both ends.

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.

Genetic and Morpho-Physiological Differences among Transgenic and No-Transgenic Cotton Cultivars.

Three carbon-chain extension genes associated with fatty acid synthesis in upland cotton (Gossypium hirsutum), namely GhKAR, GhHAD, and GhENR, play important roles in oil accumulation in cotton seeds. In the present study, these three genes were cloned and characterized. The expression patterns of GhKAR, GhHAD, and GhENR in the high seed oil content cultivars 10H1014 and 10H1041 differed somewhat compared with those of 10H1007 and 2074B with low seed oil content at different stages of seed development. GhKAR showed all three cultivars showed higher transcript levels than that of 2074B at 10-, 40-, and 45-days post anthesis (DPA). The expression pattern of GhHAD showed a lower transcript level than that of 2074B at both 10 and 30 DPA but a higher transcript level than that of 2074B at 40 DPA. GhENR showed a lower transcript level than that of 2074B at both 15 and 30 DPA. The highest transcript levels of GhKAR and GhENR were detected at 15 DPA in 10H1007, 10H1014, and 10H1041 compared with 2074B. From 5 to 45 DPA cotton seed, the oil content accumulated continuously in the developing seed. Oil accumulation reached a peak between 40 DPA and 45 DPA and slightly decreased in mature seed. In addition, GhKAR and GhENR showed different expression patterns in fiber and ovule development processes, in which they showed high expression levels at 20 DPA during the fiber elongation stage, but their expression level peaked at 15 DPA during ovule development processes. These two genes showed the lowest expression levels at the late seed maturation stage, while GhHAD showed a peak of 10 DPA in fiber development. Compared to 2074B, the oil contents of GhKAR and GhENR overexpression lines increased 1.05~1.08 folds. These results indicated that GhHAD, GhENR, and GhKAR were involved in both seed oil synthesis and fiber elongation with dual biological functions in cotton.

Assessing vaccine effectiveness for varicella in Wuxi, China: a time-series analysis.

OBJECTIVE: The varicella vaccine is not included in the national childhood immunization schedules in China. Varicella epidemics and outbreaks are frequently reported, and the evidence for the effectiveness of the varicella vaccine remains unclear. The aim of this study was to investigate varicella vaccine effectiveness in Wuxi, China. METHODS: Varicella surveillance data were extracted from the China Information System for Disease Control and Prevention, and vaccination data were obtained from the Vaccination Integrated Service Management Information System of Jiangsu Province, China. Time-series analysis approaches were used to estimate varicella vaccine effectiveness. RESULTS: A total of 16,093 varicella cases among children aged 1-6 years between January 2016 and December 2020 were analysed. A total of 217,297 children completed a two-dose varicella vaccination series. Compared with districts with lower vaccination rates, districts in Wuxi with higher varicella vaccination rates had a lower proportion of cases (p < 0.001). In the time-series approach, 0.8% fewer varicella cases were associated with a 1% increase in the two-dose varicella vaccination rate (p < 0.001), and similar effects were found in both the male and female populations. CONCLUSIONS: Two-dose varicella vaccination was recommended as an effective health intervention to prevent varicella in Wuxi, China. Varicella vaccination is urgently needed in routine childhood immunisation programs.

The introduction of two-dose varicella vaccination was an effective intervention to prevent varicella in Wuxi, China.Varicella vaccination is urgently needed in routine childhood immunization programmes.

Impact of implementing a free varicella vaccination policy on incidence in Wuxi City, China: an interrupted time series analysis.

Varicella vaccination is optional and requires self-payment. On 1 December 2018, Wuxi City launched a free varicella vaccination program for children. This study aimed to evaluate the changes in varicella incidence before and after the implementation of the policy. The data were obtained from official information systems and statistical yearbooks. We divided the period into chargeable (January 2017 to November 2018) and free (December 2018 to December 2021) periods. Interrupt time series analysis was used to conduct a generalised least-squares regression analysis for the two periods. A total of 51,071 varicella cases were reported between January 2017 and December 2021. After the implementation of the policy, there was a statistically significant decrease in the incidence of varicella (ß2 = -0.140, P = 0.017), and the slope of the incidence also decreased by 0.012 (P = 0.015). Following policy implementation, the incidence decreased in all age groups, with the largest decline observed among children aged 8-14 years (ß2 = -1.109, P = 0.009), followed by children aged ≤7 years (ß2 = -0.894, P = 0.013). Our study found a significant reduction in the incidence of varicella in the total population after the introduction of free varicella vaccination in Wuxi City.

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.

The relationship between geometry and hemodynamics of the stenotic carotid artery based on computational fluid dynamics.

OBJECTIVE: The purpose of this work was to investigate the relationship between the geometric factors and the hemodynamics of the stenotic carotid artery. METHODS: We retrospectively reviewed data of patients with carotid stenosis (40%-95%). The Navier-Stokes equations were solved using ANSYS CFX 18.0. Correlation analysis was based on Spearman's test. Geometric variables (p < 0.1 in the univariate analysis) were entered into the logistical regression. A receiver-operating characteristics analysis was used to detect hemodynamically significant lesions. RESULTS: 81 patients (96 arteries) were evaluated. The logistic regression analysis revealed that the translesional pressure ratio was significantly correlated with the stenosis degree (OR = 1.147, p < 0.001) and the angle between internal carotid artery and external carotid artery (angle γ) (OR = 0.933, p = 0.01). The translesional wall shear stress ratio was significantly correlated with stenosis degree (OR = 1.094, p < 0.001), lesion length (OR = 0.873, p = 0.01), lumen area of internal carotid artery (OR = 0.867, p = 0.002), and lumen area of common carotid artery (OR = 1.058, p = 0.01). For predicting low translesional pressure ratio, the AUC was 0.71 (p < 0.001) for angle γ, and was 0.87 (p < 0.001) for stenosis degree. For predicting high translesional wall shear stress ratio, the AUC was 0.62 (p = 0.04) for lumen area of internal carotid artery, and was 0.77 (p < 0.001) for stenosis degree. CONCLUSIONS: Apart from stenosis degree, other geometric characteristics of lesions may also have an influence on hemodynamics of the stenotic carotid artery.

OCRFinder: a noise-tolerance machine learning method for accurately estimating open chromatin regions.

Open chromatin regions are the genomic regions associated with basic cellular physiological activities, while chromatin accessibility is reported to affect gene expressions and functions. A basic computational problem is to efficiently estimate open chromatin regions, which could facilitate both genomic and epigenetic studies. Currently, ATAC-seq and cfDNA-seq (plasma cell-free DNA sequencing) are two popular strategies to detect OCRs. As cfDNA-seq can obtain more biomarkers in one round of sequencing, it is considered more effective and convenient. However, in processing cfDNA-seq data, due to the dynamically variable chromatin accessibility, it is quite difficult to obtain the training data with pure OCRs or non-OCRs, and leads to a noise problem for either feature-based approaches or learning-based approaches. In this paper, we propose a learning-based OCR estimation approach with a noise-tolerance design. The proposed approach, named OCRFinder, incorporates the ideas of ensemble learning framework and semi-supervised strategy to avoid potential overfitting of noisy labels, which are the false positives on OCRs and non-OCRs. Compared to different noise control strategies and state-of-the-art approaches, OCRFinder achieved higher accuracies and sensitivities in the experiments. In addition, OCRFinder also has an excellent performance in ATAC-seq or DNase-seq comparison experiments.

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.

Functional characterization of TBL genes revealed the role of GhTBL7 and GhTBL58 in cotton fiber elongation.

TBL (Trichome Birefringence Like) gene family members are involved in trichome initiation and xylan acetylation in several plant species. In our research, we identified 102 TBLs from G. hirsutum. The phylogenetic tree classified TBL genes into five groups. Collinearity analysis of TBL genes indicated 136 paralogous gene pairs in G. hirsutum. Gene duplication indicated that WGD or segmental duplication contributed to the GhTBL gene family expansion. Promoter cis-elements of GhTBLs were related to growth and development, seed-specific regulation, light, and stress responses. GhTBL genes (GhTBL7, GhTBL15, GhTBL21, GhTBL25, GhTBL45, GhTBL54, GhTBL67, GhTBL72, and GhTBL77) exhibited upregulated response under exposure to cold, heat, NaCl, and PEG. GhTBL genes exhibited high expression during fiber development stages. Two GhTBL genes (GhTBL7 and GhTBL58) showed differential expression at 10 DPA fiber, as 10 DPA is a fast fiber elongation stage and fiber elongation is a very important stage of cotton fiber development. Subcellular localization of GhTBL7 and GhTBL58 revealed that these genes reside inside the cell membrane. Promoter GUS activity of GhTBL7 and GhTBL58 exhibited deep staining in roots. To further validate the role of these genes in cotton fiber elongation, we silenced these genes and observed a significant reduction in the fiber length at 10 DPA. In conclusion, the functional study of cell membrane-associated genes (GhTBL7 and GhTBL58) showed deep staining in root tissues and potential function during cotton fiber elongation at 10 DPA fiber.

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.

Cortical thickness differences are associated with cellular component morphogenesis of astrocytes and excitatory neurons in nonsuicidal self-injuring youth.

Nonsuicidal self-injury (NSSI) generally occurs in youth and probably progresses to suicide. An examination of cortical thickness differences (ΔCT) between NSSI individuals and controls is crucial to investigate potential neurobiological correlates. Notably, ΔCT are influenced by specific genetic factors, and a large proportion of cortical thinning is associated with the expression of genes that overlap in astrocytes and pyramidal cells. However, in NSSI youth, the mechanisms underlying the relations between the genetic and cell type-specific transcriptional signatures to ΔCT are unclear. Here, we studied the genetic association of ΔCT in NSSI youth by performing a partial least-squares regression (PLSR) analysis of gene expression data and 3D-T1 brain images of 45 NSSI youth and 75 controls. We extracted the top-10 Gene Ontology terms for the enrichment results of upregulated PLS component 1 genes related to ΔCT to conduct the cell-type classification and enrichment analysis. Enrichment of cell type-specific genes shows that cellular component morphogenesis of astrocytes and excitatory neurons accounts for the observed NSSI-specific ΔCT. We validated the main results in independent datasets to verify the robustness and specificity. We concluded that the brain ΔCT is associated with cellular component morphogenesis of astrocytes and excitatory neurons in NSSI youth.

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.