Development of method using language processing techniques for extracting information on drug-health food product interactions.

AIMS: Health food products (HFPs) are foods and products related to maintaining and promoting health. HFPs may sometimes cause unforeseen adverse health effects by interacting with drugs. Considering the importance of information on the interactions between HFPs and drugs, this study aimed to establish a workflow to extract information on Drug-HFP Interactions (DHIs) from open resources. METHODS: First, Information on drugs, enzymes, their interactions, and known DHIs was collected from multiple public databases and literature sources. Next, a network consisted of enzymes, HFP, and drugs was constructed, assuming enzymes as candidates for hubs in Drug-HFP interactions (Method 1). Furthermore, we developed methods to analyze the biomedical context of each drug and HFP to predict potential DHIs out of the DHIs obtained in Method 1 by applying BioWordVec, a widely used biomedical terminology quantifier (Method 2-1 and 2-2). RESULTS: 44,965 DHIs (30% known) were identified in Method 1, including 38 metabolic enzymes, 157 HFPs, and 1256 drugs. Method 2-1 selected 7401 DHIs (17% known) from the DHIs of Method 1, while Method 2-2 chose 2819 DHIs (30% known). Based on the different assumptions in these methods where Method 2-1 specifically selects HFPs interacting with specific enzymes and Method 2-2 specifically selects HFPs with similar function with drugs, the propsed methods resulted in extracting a wide variety of DHIs. CONCLUSIONS: By integrating the results of language processing techniques with those of the network analysis, a workflow to efficiently extract unknown and known DHIs was constructed.

Integrated pathway mining and selection of an artificial CYP79-mediated bypass to improve benzylisoquinoline alkaloid biosynthesis.

BACKGROUND: Computational mining of useful enzymes and biosynthesis pathways is a powerful strategy for metabolic engineering. Through systematic exploration of all conceivable combinations of enzyme reactions, including both known compounds and those inferred from the chemical structures of established reactions, we can uncover previously undiscovered enzymatic processes. The application of the novel alternative pathways enables us to improve microbial bioproduction by bypassing or reinforcing metabolic bottlenecks. Benzylisoquinoline alkaloids (BIAs) are a diverse group of plant-derived compounds with important pharmaceutical properties. BIA biosynthesis has developed into a prime example of metabolic engineering and microbial bioproduction. The early bottleneck of BIA production in Escherichia coli consists of 3,4-dihydroxyphenylacetaldehyde (DHPAA) production and conversion to tetrahydropapaveroline (THP). Previous studies have selected monoamine oxidase (MAO) and DHPAA synthase (DHPAAS) to produce DHPAA from dopamine and oxygen; however, both of these enzymes produce toxic hydrogen peroxide as a byproduct. RESULTS: In the current study, in silico pathway design is applied to relieve the bottleneck of DHPAA production in the synthetic BIA pathway. Specifically, the cytochrome P450 enzyme, tyrosine N-monooxygenase (CYP79), is identified to bypass the established MAO- and DHPAAS-mediated pathways in an alternative arylacetaldoxime route to DHPAA with a peroxide-independent mechanism. The application of this pathway is proposed to result in less formation of toxic byproducts, leading to improved production of reticuline (up to 60 mg/L at the flask scale) when compared with that from the conventional MAO pathway. CONCLUSIONS: This study showed improved reticuline production using the bypass pathway predicted by the M-path computational platform. Reticuline production in E. coli exceeded that of the conventional MAO-mediated pathway. The study provides a clear example of the integration of pathway mining and enzyme design in creating artificial metabolic pathways and suggests further potential applications of this strategy in metabolic engineering.

Extracting the latent needs of dementia patients and caregivers from transcribed interviews in japanese: an initial assessment of the availability of morpheme selection as input data with Z-scores in machine learning.

BACKGROUND: Given the increasing number of dementia patients worldwide, a new method was developed for machine learning models to identify the 'latent needs' of patients and caregivers to facilitate patient/public involvement in societal decision making. METHODS: Japanese transcribed interviews with 53 dementia patients and caregivers were used. A new morpheme selection method using Z-scores was developed to identify trends in describing the latent needs. F-measures with and without the new method were compared using three machine learning models. RESULTS: The F-measures with the new method were higher for the support vector machine (SVM) (F-measure of 0.81 with the new method and F-measure of 0.79 without the new method for patients) and Naive Bayes (F-measure of 0.69 with the new method and F-measure of 0.67 without the new method for caregivers and F-measure of 0.75 with the new method and F-measure of 0.73 without the new method for patients). CONCLUSION: A new scheme based on Z-score adaptation for machine learning models was developed to predict the latent needs of dementia patients and their caregivers by extracting data from interviews in Japanese. However, this study alone cannot be used to assign significance to the adaptation of the new method because of no enough size of sample dataset. Such pre-selection with Z-score adaptation from text data in machine learning models should be considered with more modified suitable methods in the near future.

Exploration and Evaluation of Machine Learning-Based Models for Predicting Enzymatic Reactions.

Unannotated gene sequences in databases are increasing due to sequencing advances. Therefore, computational methods to predict functions of unannotated genes are needed. Moreover, novel enzyme discovery for metabolic engineering applications further encourages annotation of sequences. Here, enzyme functions are predicted using two general approaches, each including several machine learning algorithms. First, Enzyme-models (E-models) predict Enzyme Commission (EC) numbers from amino acid sequence information. Second, Substrate-Enzyme models (SE-models) are built to predict substrates of enzymatic reactions together with EC numbers, and Substrate-Enzyme-Product models (SEP-models) are built to predict substrates, products, and EC numbers. While accuracy of E-models is not optimal, SE-models and SEP-models predict EC numbers and reactions with high accuracy using all tested machine learning-based methods. For example, a single Random Forests-based SEP-model predicts EC first digits with an Average AUC score of over 0.94. Various metrics indicate that the current strategy of combining sequence and chemical structure information is effective at improving enzyme reaction prediction.

Reconstruction of metabolic pathway for isobutanol production in Escherichia coli.

BACKGROUND: The microbial production of useful fuels and chemicals has been widely studied. In several cases, glucose is used as the raw material, and almost all microbes adopt the Embden-Meyerhof (EM) pathway to degrade glucose into compounds of interest. Recently, the Entner-Doudoroff (ED) pathway has been gaining attention as an alternative strategy for microbial production. RESULTS: In the present study, we attempted to apply the ED pathway for isobutanol production in Escherichia coli because of the complete redox balance involved. First, we generated ED pathway-dependent isobutanol-producing E. coli. Thereafter, the inactivation of the genes concerning organic acids as the byproducts was performed to improve the carbon flux to isobutanol from glucose. Finally, the expression of the genes concerning the ED pathway was modified. CONCLUSIONS: The optimized isobutanol-producing E. coli produced 15.0 g/L of isobutanol as the final titer, and the yield from glucose was 0.37 g/g (g-glucose/g-isobutanol).

Transporter engineering in biomass utilization by yeast.

Biomass resources are attractive carbon sources for bioproduction because of their sustainability. Many studies have been performed using biomass resources to produce sugars as carbon sources for cell factories. Expression of biomass hydrolyzing enzymes in cell factories is an important approach for constructing biomass-utilizing bioprocesses because external addition of these enzymes is expensive. In particular, yeasts have been extensively engineered to be cell factories that directly utilize biomass because of their manageable responses to many genetic engineering tools, such as gene expression, deletion and editing. Biomass utilizing bioprocesses have also been developed using these genetic engineering tools to construct metabolic pathways. However, sugar input and product output from these cells are critical factors for improving bioproduction along with biomass utilization and metabolic pathways. Transporters are key components for efficient input and output activities. In this review, we focus on transporter engineering in yeast to enhance bioproduction from biomass resources.

M-path: a compass for navigating potential metabolic pathways.

MOTIVATION: Construction of synthetic metabolic pathways promises sustainable production of diverse chemicals and materials. To design synthetic metabolic pathways of high value, computational methods are needed to expand present knowledge by mining comprehensive chemical and enzymatic information databases. Several computational methods have been already reported for the metabolic pathway design, but until now computation complexity has limited the diversity of chemical and enzymatic data used. RESULTS: We introduce a computational platform, M-path, to explore synthetic metabolic pathways including putative enzymatic reactions and compounds. M-path is an iterative random algorithm, which makes efficient use of chemical and enzymatic databases to find potential synthetic metabolic pathways. M-path can readily control the search space and perform well compared with exhaustively enumerating possible pathways. A web-based pathway viewer is also developed to check extensive metabolic pathways with evaluation scores on the basis of chemical similarities. We further produce extensive synthetic metabolic pathways for a comprehensive set of alpha amino acids. The scalable nature of M-path enables us to calculate potential metabolic pathways for any given chemicals.

Evaluation of genes involved in oxidative phosphorylation in yeast by developing a simple and rapid method to measure mitochondrial ATP synthetic activity.

BACKGROUND: Measurement of mitochondrial ATP synthesis is a critical way to compare cellular energetic performance. However, fractionation of mitochondria requires large amounts of cells, lengthy purification procedures, and an extreme caution to avoid damaging intact mitochondria, making it the highest barrier to high-throughput studies of mitochondrial function. To evaluate 45 genes involved in oxidative phosphorylation in Saccharomyces cerevisiae, we aimed to develop a simple and rapid method to measure mitochondrial ATP synthesis. RESULTS: To obtain functional mitochondria, S. cerevisiae cells were lysed with zymolyase followed by two-step, low- then high-speed centrifugation. Using a firefly luciferin-luciferase assay, the ATP synthetic activity of the mitochondria was determined. Decreasing the ATP synthesis in the presence of mitochondrial inhibitors confirmed functionality of the isolated crude mitochondria. Deletion of genes encoding mitochondrial ATP synthesis-related protein showed their dependency on the oxidative phosphorylation in S. cerevisiae. CONCLUSIONS: Compared with conventional procedures, this measurement method for S. cerevisiae Mitochondrial ATP Synthetic activity in High-throughput (MASH method) is simple and requires a small amount of cells, making it suitable for high-throughput analyses. To our knowledge, this is the first report on a rapid purification process for yeast mitochondria suitable for high-throughput screening.

Improvement of oxidized glutathione fermentation by thiol redox metabolism engineering in Saccharomyces cerevisiae.

Glutathione is a valuable tripeptide widely used in the pharmaceutical, food, and cosmetic industries. In industrial fermentation, glutathione is currently produced primarily using the yeast Saccharomyces cerevisiae. Intracellular glutathione exists in two forms; the majority is present as reduced glutathione (GSH) and a small amount is present as oxidized glutathione (GSSG). However, GSSG is more stable than GSH and is a more attractive form for the storage of glutathione extracted from yeast cells after fermentation. In this study, intracellular GSSG content was improved by engineering thiol oxidization metabolism in yeast. An engineered strain producing high amounts of glutathione from over-expression of glutathione synthases and lacking glutathione reductase was used as a platform strain. Additional over-expression of thiol oxidase (1.8.3.2) genes ERV1 or ERO1 increased the GSSG content by 2.9-fold and 2.0-fold, respectively, compared with the platform strain, without decreasing cell growth. However, over-expression of thiol oxidase gene ERV2 showed almost no effect on the GSSG content. Interestingly, ERO1 over-expression did not decrease the GSH content, raising the total glutathione content of the cell, but ERV1 over-expression decreased the GSH content, balancing the increase in the GSSG content. Furthermore, the increase in the GSSG content due to ERO1 over-expression was enhanced by additional over-expression of the gene encoding Pdi1, whose reduced form activates Ero1 in the endoplasmic reticulum. These results indicate that engineering the thiol redox metabolism of S. cerevisiae improves GSSG and is critical to increasing the total productivity and stability of glutathione.

Development of a multi-gene expression system in Xanthophyllomyces dendrorhous.

BACKGROUND: Red yeast, Xanthophyllomyces dendrorhous (Phaffia rhodozyma) is the only yeast known to produce astaxanthin, an anti-oxidant isoprenoid (carotenoid) that is widely used in the aquaculture, food, pharmaceutical and cosmetic industries. Recently, the potential of this microorganism as a platform cell factory for isoprenoid production has been recognized because of high flux through its native terpene pathway. Addition of mevalonate, the common precursor for isoprenoid biosynthesis, has been shown to be critical to enhance the astaxanthin content in X. dendrorhous. However, addition of mevalonate is unrealistic during industrial isoprenoid production because it is an unstable and costly chemical. Therefore, up-regulating the intracellular mevalonate supply by enhancing the mevalonate synthetic pathway though genetic engineering is a promising strategy to improve isoprenoid production in X. dendrorhous. However, a system to strongly express multiple genes has been poorly developed for X. dendrorhous. RESULTS: Here, we developed a multiple gene expression system using plasmids containing three strong promoters in X. dendrorhous (actin, alcohol dehydrogenase and triose-phosphate isomerase) and their terminators. Using this system, three mevalonate synthetic pathway genes encoding acetoacetyl-CoA thiolase, HMG-CoA synthase and HMG-CoA reductase were overexpressed at the same time. This triple overexpressing strain showed an increase in astaxanthin production compared with each single overexpressing strain. Additionally, this triple overexpression of mevalonate synthetic pathway genes together with genes involved in ß-carotene and astaxanthin synthesis showed a synergetic effect on increasing astaxanthin production. Finally, astaxanthin production was enhanced by 2.1-fold compared with the parental strain without a reduction of cell growth. CONCLUSIONS: We developed a system to strongly overexpress multiple genes in X. dendrorhous. Using this system, the synthetic pathway of mevalonate, a common substrate for isoprenoid biosynthesis, was enhanced, causing an increase in astaxanthin production. Combining this multiple gene overexpression system with a platform strain that overproduces mevalonate has the potential to improve industrial production of various isoprenoids in X. dendrorhous.

Development of bio-based fine chemical production through synthetic bioengineering.

Fine chemicals that are physiologically active, such as pharmaceuticals, cosmetics, nutritional supplements, flavoring agents as well as additives for foods, feed, and fertilizer are produced by enzymatically or through microbial fermentation. The identification of enzymes that catalyze the target reaction makes possible the enzymatic synthesis of the desired fine chemical. The genes encoding these enzymes are then introduced into suitable microbial hosts that are cultured with inexpensive, naturally abundant carbon sources, and other nutrients. Metabolic engineering create efficient microbial cell factories for producing chemicals at higher yields. Molecular genetic techniques are then used to optimize metabolic pathways of genetically and metabolically well-characterized hosts. Synthetic bioengineering represents a novel approach to employ a combination of computer simulation and metabolic analysis to design artificial metabolic pathways suitable for mass production of target chemicals in host strains. In the present review, we summarize recent studies on bio-based fine chemical production and assess the potential of synthetic bioengineering for further improving their productivity.

Evaluation and screening of efficient promoters to improve astaxanthin production in Xanthophyllomyces dendrorhous.

Astaxanthin is a valuable carotenoid that is widely used in the aquaculture, food, pharmaceutical, and cosmetic industries. Xanthophyllomyces dendrorhous is a carotenoid-synthesizing yeast strain that produces astaxanthin as its main pigment. Although metabolic engineering using gene manipulation is a valuable way to improve astaxanthin production, a gene expression system for X. dendrorhous has been poorly developed. In this study, three known promoters of X. dendrorhous, glycerol-3-phosphate dehydrogenase (gpd) promoter (Pgpd), glucose dehydrogenase (gdh) promoter (Pgdh), and actin (act) promoter (Pact), were evaluated for use in the overexpression of target proteins using green fluorescence protein (GFP) as an expression level indicator protein. The actin promoter, Pact, showed the highest expression level of GFP when compared with Pgpd and Pgdh. Additionally, to obtain new promoters for higher expression of target protein in X. dendrorhous, intracellular GFP intensity was evaluated for 13 candidate promoters. An alcohol dehydrogenase promoter, Padh4, showed more efficient expression of GFP rather than Pact. Overexpression of crtE gene encoding rate-limiting enzyme of carotenoid synthesis under the adh4 promoter yielded an increase in intracellular astaxanthin content of about 1.7-fold compared with the control strain. The promoters identified in this study must be useful for improving carotenoids production in X. dendrorhous.

Development of a Machine Learning Model for Classifying Cooking Recipes According to Dietary Styles.

To complement classical methods for identifying Japanese, Chinese, and Western dietary styles, this study aimed to develop a machine learning model. This study utilized 604 features from 8183 cooking recipes based on a Japanese recipe site. The data were randomly divided into training, validation, and test sets for each dietary style at a 60:20:20 ratio. Six machine learning models were developed in this study to effectively classify cooking recipes according to dietary styles. The evaluation indicators were above 0.8 for all models in each dietary style. The top ten features were extracted from each model, and the features common to three or more models were employed as the best predictive features. Five well-predicted features were indicated for the following seasonings: soy sauce, miso (fermented soy beans), and mirin (sweet cooking rice wine) in the Japanese diet; oyster sauce and doubanjiang (chili bean sauce) in the Chinese diet; and olive oil in the Western diet. Predictions by broth were indicated in each diet, such as dashi in the Japanese diet, chicken soup in the Chinese diet, and consommé in the Western diet. The prediction model suggested that seasonings and broths could be used to predict dietary styles.

Prediction of Metabolic Flux Distribution by Flux Sampling: As a Case Study, Acetate Production from Glucose in Escherichia coli.

Omics data was acquired, and the development and research of metabolic simulation and analysis methods using them were also actively carried out. However, it was a laborious task to acquire such data each time the medium composition, culture conditions, and target organism changed. Therefore, in this study, we aimed to extract and estimate important variables and necessary numbers for predicting metabolic flux distribution as the state of cell metabolism by flux sampling using a genome-scale metabolic model (GSM) and its analysis. Acetic acid production from glucose in Escherichia coli with GSM iJO1366 was used as a case study. Flux sampling obtained by OptGP using 1000 pattern constraints on substrate, product, and growth fluxes produced a wider sample than the default case. The analysis also suggested that the fluxes of iron ions, O2, CO2, and NH4+, were important for predicting the metabolic flux distribution. Additionally, the comparison with the literature value of 13C-MFA using CO2 emission flux as an example of an important flux suggested that the important flux obtained by this method was valid for the prediction of flux distribution. In this way, the method of this research was useful for extracting variables that were important for predicting flux distribution, and as a result, the possibility of contributing to the reduction of measurement variables in experiments was suggested.

Different Recognition of Protein Features Depending on Deep Learning Models: A Case Study of Aromatic Decarboxylase UbiD.

The number of unannotated protein sequences is explosively increasing due to genome sequence technology. A more comprehensive understanding of protein functions for protein annotation requires the discovery of new features that cannot be captured from conventional methods. Deep learning can extract important features from input data and predict protein functions based on the features. Here, protein feature vectors generated by 3 deep learning models are analyzed using Integrated Gradients to explore important features of amino acid sites. As a case study, prediction and feature extraction models for UbiD enzymes were built using these models. The important amino acid residues extracted from the models were different from secondary structures, conserved regions and active sites of known UbiD information. Interestingly, the different amino acid residues within UbiD sequences were regarded as important factors depending on the type of models and sequences. The Transformer models focused on more specific regions than the other models. These results suggest that each deep learning model understands protein features with different aspects from existing knowledge and has the potential to discover new laws of protein functions. This study will help to extract new protein features for the other protein annotations.

EnzymeNet: residual neural networks model for Enzyme Commission number prediction.

Motivation: Enzymes are key targets to biosynthesize functional substances in metabolic engineering. Therefore, various machine learning models have been developed to predict Enzyme Commission (EC) numbers, one of the enzyme annotations. However, the previously reported models might predict the sequences with numerous consecutive identical amino acids, which are found within unannotated sequences, as enzymes. Results: Here, we propose EnzymeNet for prediction of complete EC numbers using residual neural networks. EnzymeNet can exclude the exceptional sequences described above. Several EnzymeNet models were built and optimized to explore the best conditions for removing such sequences. As a result, the models exhibited higher prediction accuracy with macro F1 score up to 0.850 than previously reported models. Moreover, even the enzyme sequences with low similarity to training data, which were difficult to predict using the reported models, could be predicted extensively using EnzymeNet models. The robustness of EnzymeNet models will lead to discover novel enzymes for biosynthesis of functional compounds using microorganisms. Availability and implementation: The source code of EnzymeNet models is freely available at https://github.com/nwatanbe/enzymenet.

Predicting Cardiovascular Disease Mortality: Leveraging Machine Learning for Comprehensive Assessment of Health and Nutrition Variables.

Cardiovascular disease (CVD) is one of the primary causes of death around the world. This study aimed to identify risk factors associated with CVD mortality using data from the National Health and Nutrition Examination Survey (NHANES). We created three models focusing on dietary data, non-diet-related health data, and a combination of both. Machine learning (ML) models, particularly the random forest algorithm, demonstrated robust consistency across health, nutrition, and mixed categories in predicting death from CVD. Shapley additive explanation (SHAP) values showed age, systolic blood pressure, and several other health factors as crucial variables, while fiber, calcium, and vitamin E, among others, were significant nutritional variables. Our research emphasizes the importance of comprehensive health evaluation and dietary intake in predicting CVD mortality. The inclusion of nutrition variables improved the performance of our models, underscoring the utility of dietary intake in ML-based data analysis. Further investigation using large datasets with recurring dietary recalls is necessary to enhance the effectiveness and interpretability of such models.

Ramen Consumption and Gut Microbiota Diversity in Japanese Women: Cross-Sectional Data from the NEXIS Cohort Study.

A cross-sectional study involving 224 healthy Japanese adult females explored the relationship between ramen intake, gut microbiota diversity, and blood biochemistry. Using a stepwise regression model, ramen intake was inversely associated with gut microbiome alpha diversity after adjusting for related factors, including diets, Age, BMI, and stool habits (ß = -0.018; r = -0.15 for Shannon index). The intake group of ramen was inversely associated with dietary nutrients and dietary fiber compared with the no-intake group of ramen. Sugar intake, Dorea as a short-chain fatty acid (SCFA)-producing gut microbiota, and γ-glutamyl transferase as a liver function marker were directly associated with ramen intake after adjustment for related factors including diets, gut microbiota, and blood chemistry using a stepwise logistic regression model, whereas Dorea is inconsistently less abundant in the ramen group. In conclusion, the increased ramen was associated with decreased gut bacterial diversity accompanying a perturbation of Dorea through the dietary nutrients, gut microbiota, and blood chemistry, while the methodological limitations existed in a cross-sectional study. People with frequent ramen eating habits need to take measures to consume various nutrients to maintain and improve their health, and dietary management can be applied to the dietary feature in ramen consumption.

Predicting Overweight and Obesity Status Among Malaysian Working Adults With Machine Learning or Logistic Regression: Retrospective Comparison Study.

BACKGROUND: Overweight or obesity is a primary health concern that leads to a significant burden of noncommunicable disease and threatens national productivity and economic growth. Given the complexity of the etiology of overweight or obesity, machine learning (ML) algorithms offer a promising alternative approach in disentangling interdependent factors for predicting overweight or obesity status. OBJECTIVE: This study examined the performance of 3 ML algorithms in comparison with logistic regression (LR) to predict overweight or obesity status among working adults in Malaysia. METHODS: Using data from 16,860 participants (mean age 34.2, SD 9.0 years; n=6904, 41% male; n=7048, 41.8% with overweight or obesity) in the Malaysia's Healthiest Workplace by AIA Vitality 2019 survey, predictor variables, including sociodemographic characteristics, job characteristics, health and weight perceptions, and lifestyle-related factors, were modeled using the extreme gradient boosting (XGBoost), random forest (RF), and support vector machine (SVM) algorithms, as well as LR, to predict overweight or obesity status based on a BMI cutoff of 25 kg/m2. RESULTS: The area under the receiver operating characteristic curve was 0.81 (95% CI 0.79-0.82), 0.80 (95% CI 0.79-0.81), 0.80 (95% CI 0.78-0.81), and 0.78 (95% CI 0.77-0.80) for the XGBoost, RF, SVM, and LR models, respectively. Weight satisfaction was the top predictor, and ethnicity, age, and gender were also consistent predictor variables of overweight or obesity status in all models. CONCLUSIONS: Based on multi-domain online workplace survey data, this study produced predictive models that identified overweight or obesity status with moderate to high accuracy. The performance of both ML-based and logistic regression models were comparable when predicting obesity among working adults in Malaysia.

Comprehensive Machine Learning Prediction of Extensive Enzymatic Reactions.

New enzyme functions exist within the increasing number of unannotated protein sequences. Novel enzyme discovery is necessary to expand the pathways that can be accessed by metabolic engineering for the biosynthesis of functional compounds. Accordingly, various machine learning models have been developed to predict enzymatic reactions. However, the ability to predict unknown reactions that are not included in the training data has not been clarified. In order to cover uncertain and unknown reactions, a wider range of reaction types must be demonstrated by the models. Here, we establish 16 expanded enzymatic reaction prediction models developed using various machine learning algorithms, including deep neural network. Improvements in prediction performances over that of our previous study indicate that the updated methods are more effective for the prediction of enzymatic reactions. Overall, the deep neural network model trained with combined substrate-enzyme-product information exhibits the highest prediction accuracy with Macro F1 scores up to 0.966 and with robust prediction of unknown enzymatic reactions that are not included in the training data. This model can predict more extensive enzymatic reactions in comparison to previously reported models. This study will facilitate the discovery of new enzymes for the production of useful substances.