Biology subjects

Hosoda, S.

Publications and source records attributed to Hosoda, S..

5 recordsLinked to original sources

bioRxiv2024

Prediction of antibiotic resistance mechanisms using a protein language model

MotivationAntibiotic resistance has emerged as a major global health threat, with an increasing number of bacterial infections becoming difficult to treat. Predicting the underlying resistance mechanisms of antibiotic resistance genes (ARGs) is crucial for understanding and combating this problem. However, existing methods struggle to accurately predict resistance mechanisms for ARGs with low similarity to known sequences and lack sufficient interpretability of the prediction models. ResultsIn this study, we present a novel approach for predicting ARG resistance mechanisms using Protein-BERT, a protein language model based on deep learning. Our method outperforms state-of-the-art techniques on diverse ARG datasets, including those with low homology to the training data, highlighting its potential for predicting the resistance mechanisms of unknown ARGs. Attention analysis of the model reveals that it considers biologically relevant features, such as conserved amino acid residues and antibiotic target binding sites, when making predictions. These findings provide valuable insights into the molecular basis of antibiotic resistance and demonstrate the interpretability of protein language models, offering a new perspective on their application in bioinformatics. AvailabilityThe source code is available for free at https://github.com/hmdlab/ARG-BERT. The output results of the model are published at https://waseda.box.com/v/ARG-BERT-suppl. Contactmhamada@waseda.jp

bioinformatics↗

bioRxiv2024

BayesianSSA: a Bayesian statistical model based on structural sensitivity analysis for predicting responses to enzyme perturbations in metabolic networks

BackgroundChemical bioproduction has attracted attention as a key technology in a decarbonized society. In computational design for chemical bioproduction, it is necessary to predict changes in metabolic fluxes when up-/down-regulating enzymatic reactions, that is, responses of the system to enzyme perturbations. Structural sensitivity analysis (SSA) was previously developed as a method to predict qualitative responses to enzyme perturbations on the basis of the structural information of the reaction network. However, the network structural information can sometimes be insufficient to predict qualitative responses unambiguously, which is a practical issue in bioproduction applications. To address this, in this study, we propose BayesianSSA, a Bayesian statistical model based on SSA. BayesianSSA extracts environmental information from perturbation datasets collected in environments of interest and integrates it into SSA predictions. ResultsWe applied BayesianSSA to synthetic and real datasets of the central metabolic pathway of Escherichia coli. Our result demonstrates that BayesianSSA can successfully integrate environmental information extracted from perturbation data into SSA predictions. In addition, the posterior distribution estimated by BayesianSSA can be associated with the known pathway reported to enhance succinate export flux in previous studies. ConclusionsWe believe that BayesianSSA will accelerate the chemical bioproduction process and contribute to advancements in the field.

bioinformatics↗

bioRxiv2021

Representation learning applications in biological sequence analysis

Remarkable advances in high-throughput sequencing have resulted in rapid data accumulation, and analyzing biological (DNA/RNA/protein) sequences to discover new insights in biology has become more critical and challenging. To tackle this issue, the application of natural language processing (NLP) to biological sequence analysis has received increased attention, because biological sequences are regarded as sentences and k-mers in these sequences as words. Embedding is an essential step in NLP, which converts words into vectors. This transformation is called representation learning and can be applied to biological sequences. Vectorized biological sequences can be used for function and structure estimation, or as inputs for other probabilistic models. Given the importance and growing trend in the application of representation learning in biology, here, we review the existing knowledge in representation learning for biological sequence analysis.

bioinformatics↗

bioRxiv2021

Umibato: estimation of time-varying microbial interaction using continuous-time regression hidden Markov model

MotivationAccumulating evidence has highlighted the importance of microbial interaction networks. Methods have been developed for estimating microbial interaction networks, of which the generalized Lotka-Volterra equation (gLVE)-based method can estimate a directed interaction network. The previous gLVE-based method for estimating microbial interaction networks did not consider time-varying interactions. ResultsIn this study, we developed unsupervised learning based microbial interaction inference method using Bayesian estimation (Umibato), a method for estimating time-varying microbial interactions. The Umibato algorithm comprises Gaussian process regression (GPR) and a new Bayesian probabilistic model, the continuous-time regression hidden Markov model (CTRHMM). Growth rates are estimated by GPR, and interaction networks are estimated by CTRHMM. CTRHMM can estimate time-varying interaction networks using interaction states, which are defined as hidden variables. Umibato outperformed the existing methods on synthetic datasets. In addition, it yielded reasonable estimations in experiments on a mouse gut microbiota dataset, thus providing novel insights into the relationship between consumed diets and the gut microbiota. AvailabilityThe C++ and python source codes of the Umibato software are available at https://github.com/shion-h/Umibato Contactshion_hosoda@asagi.waseda.jp, mhamada@waseda.jp

microbiology↗

bioRxiv2019

Revealing microbial assemblage structure in the human gut microbiome using latent Dirichlet allocation

Recent research has revealed that there are various microbial species in the human gut microbiome. To clarify the structure of the human gut microbiome, many data mining methods have been applied to microbial composition data. Cluster analysis, one of the key data mining methods that have been used in human gut microbiome research, can classify the human gut microbiome into three clusters, called enterotypes. The human gut microbiome has been suggested to be composed of the microbial assemblages or groups of co-occurring microbes, and one human gut microbiome can contain several microbial assemblages. However, cluster analysis can cluster samples into groups without capturing minor assemblages. In addition, a reliable method of assemblage detection has not been established, and little is known about the distributions of microbial assemblages at a population-level scale. Accordingly, the purpose of this study was to clarify the microbial assemblages in the human gut microbiome. In this study, we detected gut microbiome assemblages using a latent Dirichlet allocation (LDA) method, which was first proposed for the classification of documents in natural language processing. We applied LDA to a large-scale human gut metagenome dataset and found that a four-assemblage LDA model can represent relationships between enterotypes and assemblages with high interpretability. This model indicates that each individual tends to have several assemblages, and each of three assemblages corresponded to each enterotype. However, the C-assemblage can exist in all enterotypes. Interestingly, the dominant genera of the C-assemblage (Clostridium, Eubacterium, Faecalibacterium, Roseburia, Coprococcus, and Butyrivibrio) included butyrate-producing species such as Faecalibacterium prausnitzii. Finally, we revealed that genera mainly appearing in the same assemblage were correlated to each other. We conducted an assemblage analysis on a large-scale human gut metagenome dataset using LDA, a powerful method for detection of microbial assemblages. This approach has the potential to reveal the structure of the human gut microbiome.

microbiology↗