Dataset Information

Integrated-omics analysis with explainable deep networks on pathobiology of infant bronchiolitis.

ABSTRACT: Bronchiolitis is the leading cause of infant hospitalization. However, the molecular networks driving bronchiolitis pathobiology remain unknown. Integrative molecular networks, including the transcriptome and metabolome, can identify functional and regulatory pathways contributing to disease severity. Here, we integrated nasopharyngeal transcriptome and metabolome data of 397 infants hospitalized with bronchiolitis in a 17-center prospective cohort study. Using an explainable deep network model, we identified an omics-cluster comprising 401 transcripts and 38 metabolites that distinguishes bronchiolitis severity (test-set AUC, 0.828). This omics-cluster derived a molecular network, where innate immunity-related metabolites (e.g., ceramides) centralized and were characterized by toll-like receptor (TLR) and NF-κB signaling pathways (both FDR < 0.001). The network analyses identified eight modules and 50 existing drug candidates for repurposing, including prostaglandin I₂ analogs (e.g., iloprost), which promote anti-inflammatory effects through TLR signaling. Our approach facilitates not only the identification of molecular networks underlying infant bronchiolitis but the development of pioneering treatment strategies.

SUBMITTER: Ooka T

PROVIDER: S-EPMC11341550 | biostudies-literature | 2024 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Integrated-omics analysis with explainable deep networks on pathobiology of infant bronchiolitis.

Ooka Tadao T Usuyama Naoto N Shibata Ryohei R Kyo Michihito M Mansbach Jonathan M JM Zhu Zhaozhong Z Camargo Carlos A CA Hasegawa Kohei K

NPJ systems biology and applications 20240822 1

Bronchiolitis is the leading cause of infant hospitalization. However, the molecular networks driving bronchiolitis pathobiology remain unknown. Integrative molecular networks, including the transcriptome and metabolome, can identify functional and regulatory pathways contributing to disease severity. Here, we integrated nasopharyngeal transcriptome and metabolome data of 397 infants hospitalized with bronchiolitis in a 17-center prospective cohort study. Using an explainable deep network model, ...[more]

PMID: 39174575

Similar Datasets

Project description:BackgroundYoung children with rhinovirus (RV) infection-particularly bronchiolitis-are at high risk for developing childhood asthma. Emerging evidence suggests clinical heterogeneity within RV bronchiolitis. However, little is known about these biologically distinct subgroups (endotypes) and their relations with asthma risk.ObjectiveWe aimed to identify RV bronchiolitis endotypes and examine their longitudinal relations with asthma risk.MethodsAs part of a multicenter prospective cohort study of infants (age <12 months) hospitalized for bronchiolitis, we integrated clinical, RV species (RV-A, RV-B, and RV-C), nasopharyngeal microbiome (16S rRNA gene sequencing), cytokine, and metabolome (liquid chromatography tandem mass spectrometry) data collected at hospitalization. We then applied network and clustering approaches to identify bronchiolitis endotypes. We also examined their longitudinal association with risks of developing recurrent wheeze by age 3 years and asthma by age 5 years.ResultsOf 122 infants hospitalized for RV bronchiolitis (median age, 4 months), we identified 4 distinct endotypes-mainly characterized by RV species, microbiome, and type 2 cytokine (T2) response: endotype A, virusRV-CmicrobiomemixedT2low; endotype B, virusRV-AmicrobiomeHaemophilusT2low; endotype C, virusRSV/RVmicrobiomeStreptococcusT2low; and endotype D, virusRV-CmicrobiomeMoraxellaT2high. Compared with endotype A infants, endotype D infants had a significantly higher rate of recurrent wheeze (33% vs 64%; hazard ratio, 2.23; 95% CI, 1.00-4.96; P = .049) and a higher risk for developing asthma (28% vs 59%; odds ratio, 3.74: 95% CI, 1.21-12.6; P = .03).ConclusionsIntegrated-omics analysis identified biologically meaningful RV bronchiolitis endotypes in infants, such as one characterized by RV-C infection, Moraxella-dominant microbiota, and high T2 cytokine response, at higher risk for developing recurrent wheeze and asthma. This study should facilitate further research toward validating our inferences.

Project description:Predicting compound-protein affinity is beneficial for accelerating drug discovery. Doing so without the often-unavailable structure data is gaining interest. However, recent progress in structure-free affinity prediction, made by machine learning, focuses on accuracy but leaves much to be desired for interpretability. Defining intermolecular contacts underlying affinities as a vehicle for interpretability; our large-scale interpretability assessment finds previously used attention mechanisms inadequate. We thus formulate a hierarchical multiobjective learning problem, where predicted contacts form the basis for predicted affinities. We solve the problem by embedding protein sequences (by hierarchical recurrent neural networks) and compound graphs (by graph neural networks) with joint attentions between protein residues and compound atoms. We further introduce three methodological advances to enhance interpretability: (1) structure-aware regularization of attentions using protein sequence-predicted solvent exposure and residue-residue contact maps; (2) supervision of attentions using known intermolecular contacts in training data; and (3) an intrinsically explainable architecture where atomic-level contacts or "relations" lead to molecular-level affinity prediction. The first two and all three advances result in DeepAffinity+ and DeepRelations, respectively. Our methods show generalizability in affinity prediction for molecules that are new and dissimilar to training examples. Moreover, they show superior interpretability compared to state-of-the-art interpretable methods: with similar or better affinity prediction, they boost the AUPRC of contact prediction by around 33-, 35-, 10-, and 9-fold for the default test, new-compound, new-protein, and both-new sets, respectively. We further demonstrate their potential utilities in contact-assisted docking, structure-free binding site prediction, and structure-activity relationship studies without docking. Our study represents the first model development and systematic model assessment dedicated to interpretable machine learning for structure-free compound-protein affinity prediction.

Dataset Information

Integrated-omics analysis with explainable deep networks on pathobiology of infant bronchiolitis.

Publications

Integrated-omics analysis with explainable deep networks on pathobiology of infant bronchiolitis.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets