AnoChem: Prediction of chemical structural abnormalities based on machine learning models.

De novo drug design aims to rationally discover novel and potent compounds while reducing experimental costs during the drug development stage. Despite the numerous generative models that have been developed, few successful cases of drug design utilizing generative models have been reported. One of the most common challenges is designing compounds that are not synthesizable or realistic. Therefore, methods capable of accurately assessing the chemical structures proposed by generative models for drug design are needed. In this study, we present AnoChem, a computational framework based on deep learning designed to assess the likelihood of a generated molecule being real. AnoChem achieves an area under the receiver operating characteristic curve score of 0.900 for distinguishing between real and generated molecules. We utilized AnoChem to evaluate and compare the performances of several generative models, using other metrics, namely SAscore and Fréschet ChemNet distance (FCD). AnoChem demonstrates a strong correlation with these metrics, validating its effectiveness as a reliable tool for assessing generative models. The source code for AnoChem is available at https://github.com/CSB-L/AnoChem.

Intrahepatic IgA complex induces polarization of cancer-associated fibroblasts to matrix phenotypes in the tumor microenvironment of HCC.

BACKGROUND AND AIMS: Cancer-associated fibroblasts (CAFs) play key roles in the tumor microenvironment. IgA contributes to inflammation and dismantling antitumor immunity in the human liver. In this study, we aimed to elucidate the effects of the IgA complex on CAFs in Pil Soo Sung the tumor microenvironment of HCC. APPROACH AND RESULTS: CAF dynamics in HCC tumor microenvironment were analyzed through single-cell RNA sequencing of HCC samples. CAFs isolated from 50 HCC samples were treated with mock or serum-derived IgA dimers in vitro. Progression-free survival of patients with advanced HCC treated with atezolizumab and bevacizumab was significantly longer in those with low serum IgA levels ( p <0.05). Single-cell analysis showed that subcluster proportions in the CAF-fibroblast activation protein-α matrix were significantly increased in patients with high serum IgA levels. Flow cytometry revealed a significant increase in the mean fluorescence intensity of fibroblast activation protein in the CD68 + cells from patients with high serum IgA levels ( p <0.001). We confirmed CD71 (IgA receptor) expression in CAFs, and IgA-treated CAFs exhibited higher programmed death-ligand 1 expression levels than those in mock-treated CAFs ( p <0.05). Coculture with CAFs attenuated the cytotoxic function of activated CD8 + T cells. Interestingly, activated CD8 + T cells cocultured with IgA-treated CAFs exhibited increased programmed death-1 expression levels than those cocultured with mock-treated CAFs ( p <0.05). CONCLUSIONS: Intrahepatic IgA induced polarization of HCC-CAFs into more malignant matrix phenotypes and attenuates cytotoxic T-cell function. Our study highlighted their potential roles in tumor progression and immune suppression.

Investigation of the General Molecular Mechanisms of Gallic Acid via Analyses of Its Transcriptome Profile.

Gallic acid (GA), a phenolic compound naturally found in many plants, exhibits potential preventive and therapeutic roles. However, the underlying molecular mechanisms of its diverse biological activities remain unclear. Here, we investigated possible mechanisms of GA function through a transcriptome-based analysis using LINCS L1000, a publicly available data resource. We compared the changes in the gene expression profiles induced by GA with those induced by FDA-approved drugs in three cancer cell lines (A549, PC3, and MCF7). The top 10 drugs exhibiting high similarity with GA in their expression patterns were identified by calculating the connectivity score in the three cell lines. We specified the known target proteins of these drugs, which could be potential targets of GA, and identified 19 potential targets. Next, we retrieved evidence in the literature that GA likely binds directly to DNA polymerase ß and ribonucleoside-diphosphate reductase. Although our results align with previous studies suggesting a direct and/or indirect connection between GA and the target proteins, further experimental investigations are required to fully understand the exact molecular mechanisms of GA. Our study provides insights into the therapeutic mechanisms of GA, introducing a new approach to characterizing therapeutic natural compounds using transcriptome-based analyses.

Genome-wide identification of overexpression and downregulation gene targets based on the sum of covariances of the outgoing reaction fluxes.

In metabolic engineering, predicting gene overexpression targets remains challenging because both endogenous and heterologous genes in a large metabolic space can be candidates, in contrast to gene knockout targets that are confined to endogenous genes. We report the development of iBridge that identifies positive and negative metabolites exerting positive and negative impacts on product formation, respectively, based on the sum of covariances of their outgoing (consuming) reaction fluxes for a target chemical. Then, "bridge" reactions converting negative metabolites to positive metabolites are identified as overexpression targets, while the opposites as downregulation targets. Using iBridge, overexpression and downregulation targets are suggested for the production of 298 chemicals and validated for 36 chemicals experimentally demonstrated in previous studies. Finally, iBridge is employed to engineer Escherichia coli strains capable of producing 10.3 g/L of D-panthenol, a compound not previously produced, as well as putrescine and 4-hydroxyphenyllactate at enhanced titers, 63.7 and 8.3 g/L, respectively.

Targeted and high-throughput gene knockdown in diverse bacteria using synthetic sRNAs.

Synthetic sRNAs allow knockdown of target genes at translational level, but have been restricted to a limited number of bacteria. Here, we report the development of a broad-host-range synthetic sRNA (BHR-sRNA) platform employing the RoxS scaffold and the Hfq chaperone from Bacillus subtilis. BHR-sRNA is tested in 16 bacterial species including commensal, probiotic, pathogenic, and industrial bacteria, with >50% of target gene knockdown achieved in 12 bacterial species. For medical applications, virulence factors in Staphylococcus epidermidis and Klebsiella pneumoniae are knocked down to mitigate their virulence-associated phenotypes. For metabolic engineering applications, high performance Corynebacterium glutamicum strains capable of producing valerolactam (bulk chemical) and methyl anthranilate (fine chemical) are developed by combinatorial knockdown of target genes. A genome-scale sRNA library covering 2959 C. glutamicum genes is constructed for high-throughput colorimetric screening of indigoidine (natural colorant) overproducers. The BHR-sRNA platform will expedite engineering of diverse bacteria of both industrial and medical interest.

Computational prediction of interactions between Paxlovid and prescription drugs.

Pfizer's Paxlovid has recently been approved for the emergency use authorization (EUA) from the US Food and Drug Administration (FDA) for the treatment of mild-to-moderate COVID-19. Drug interactions can be a serious medical problem for COVID-19 patients with underlying medical conditions, such as hypertension and diabetes, who have likely been taking other drugs. Here, we use deep learning to predict potential drug-drug interactions between Paxlovid components (nirmatrelvir and ritonavir) and 2,248 prescription drugs for treating various diseases.

PredAOT: a computational framework for prediction of acute oral toxicity based on multiple random forest models.

BACKGROUND: Acute oral toxicity of drug candidates can lead to drug development failure; thus, predicting the acute oral toxicity of small compounds is important for successful drug development. However, evaluation of the acute oral toxicity of small compounds considered in the early stages of drug discovery is limited because of cost and time. Here, we developed a computational framework, PredAOT, that predicts the acute oral toxicity of small compounds in mice and rats. METHODS: PredAOT is based on multiple random forest models for the accurate prediction of acute oral toxicity. A total of 6226 and 6238 compounds evaluated in mice and rats, respectively, were used to train the models. RESULTS: PredAOT has the advantage of predicting acute oral toxicity in mice and rats simultaneously, and its prediction performance is similar to or better than that of existing tools. CONCLUSION: PredAOT will be a useful tool for the quick and accurate prediction of the acute oral toxicity of small compounds in mice and rats during drug development.

Identification and New Indication of Melanin-Concentrating Hormone Receptor 1 (MCHR1) Antagonist Derived from Machine Learning and Transcriptome-Based Drug Repositioning Approaches.

Melanin-concentrating hormone receptor 1 (MCHR1) has been a target for appetite suppressants, which are helpful in treating obesity. However, it is challenging to develop an MCHR1 antagonist because its binding site is similar to that of the human Ether-à-go-go-Related Gene (hERG) channel, whose inhibition may cause cardiotoxicity. Most drugs developed as MCHR1 antagonists have failed in clinical development due to cardiotoxicity caused by hERG inhibition. Machine learning-based prediction models can overcome these difficulties and provide new opportunities for drug discovery. In this study, we identified KRX-104130 with potent MCHR1 antagonistic activity and no cardiotoxicity through virtual screening using two MCHR1 binding affinity prediction models and an hERG-induced cardiotoxicity prediction model. In addition, we explored other possibilities for expanding the new indications for KRX-104130 using a transcriptome-based drug repositioning approach. KRX-104130 increased the expression of low-density lipoprotein receptor (LDLR), which induced cholesterol reduction in the gene expression analysis. This was confirmed by comparison with gene expression in a nonalcoholic steatohepatitis (NASH) patient group. In a NASH mouse model, the administration of KRX-104130 showed a protective effect by reducing hepatic lipid accumulation, liver injury, and histopathological changes, indicating a promising prospect for the therapeutic effect of NASH as a new indication for MCHR1 antagonists.

PredMS: a random forest model for predicting metabolic stability of drug candidates in human liver microsomes.

MOTIVATION: Poor metabolic stability leads to drug development failure. Therefore, it is essential to evaluate the metabolic stability of small compounds for successful drug discovery and development. However, evaluating metabolic stability in vitro and in vivo is expensive, time-consuming and laborious. In addition, only a few free software programs are available for metabolic stability data and prediction. Therefore, in this study, we aimed to develop a prediction model that predicts the metabolic stability of small compounds. RESULTS: We developed a computational model, PredMS, which predicts the metabolic stability of small compounds as stable or unstable in human liver microsomes. PredMS is based on a random forest model using an in-house database of metabolic stability data of 1917 compounds. To validate the prediction performance of PredMS, we generated external test data of 61 compounds. PredMS achieved an accuracy of 0.74, Matthew's correlation coefficient of 0.48, sensitivity of 0.70, specificity of 0.86, positive predictive value of 0.94 and negative predictive value of 0.46 on the external test dataset. PredMS will be a useful tool to predict the metabolic stability of small compounds in the early stages of drug discovery and development. AVAILABILITY AND IMPLEMENTATION: The source code for PredMS is available at https://bitbucket.org/krictai/predms, and the PredMS web server is available at https://predms.netlify.app. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Identification of new target proteins of a Urotensin-II receptor antagonist using transcriptome-based drug repositioning approach.

Drug repositioning research using transcriptome data has recently attracted attention. In this study, we attempted to identify new target proteins of the urotensin-II receptor antagonist, KR-37524 (4-(3-bromo-4-(piperidin-4-yloxy)benzyl)-N-(3-(dimethylamino)phenyl)piperazine-1-carboxamide dihydrochloride), using a transcriptome-based drug repositioning approach. To do this, we obtained KR-37524-induced gene expression profile changes in four cell lines (A375, A549, MCF7, and PC3), and compared them with the approved drug-induced gene expression profile changes available in the LINCS L1000 database to identify approved drugs with similar gene expression profile changes. Here, the similarity between the two gene expression profile changes was calculated using the connectivity score. We then selected proteins that are known targets of the top three approved drugs with the highest connectivity score in each cell line (12 drugs in total) as potential targets of KR-37524. Seven potential target proteins were experimentally confirmed using an in vitro binding assay. Through this analysis, we identified that neurologically regulated serotonin transporter proteins are new target proteins of KR-37524. These results indicate that the transcriptome-based drug repositioning approach can be used to identify new target proteins of a given compound, and we provide a standalone software developed in this study that will serve as a useful tool for drug repositioning.

A deep learning approach to evaluate the feasibility of enzymatic reactions generated by retrobiosynthesis.

Retrobiosynthesis allows the designing of novel biosynthetic pathways for the production of chemicals and materials through metabolic engineering, but generates a large number of reactions beyond the experimental feasibility. Thus, an effective method that can reduce a large number of the initially predicted enzymatic reactions has been needed. Here, we present Deep learning-based Reaction Feasibility Checker (DeepRFC) to classify the feasibility of a given enzymatic reaction with high performance and speed. DeepRFC is designed to receive Simplified Molecular-Input Line-Entry System (SMILES) strings of a reactant pair, which is defined as a substrate and a product of a reaction, as an input, and evaluates whether the input reaction is feasible. A deep neural network is selected for DeepRFC as it leads to better classification performance than five other representative machine learning methods examined. For validation, the performance of DeepRFC is compared with another in-house reaction feasibility checker that uses the concept of reaction similarity. Finally, the use of DeepRFC is demonstrated for the retrobiosynthesis-based design of novel one-carbon assimilation pathways. DeepRFC will allow retrobiosynthesis to be more practical for metabolic engineering applications by efficiently screening a large number of retrobiosynthesis-derived enzymatic reactions. DeepRFC is freely available at https://bitbucket.org/kaistsystemsbiology/deeprfc.

LightBBB: computational prediction model of blood-brain-barrier penetration based on LightGBM.

MOTIVATION: Identification of blood-brain barrier (BBB) permeability of a compound is a major challenge in neurotherapeutic drug discovery. Conventional approaches for BBB permeability measurement are expensive, time-consuming and labor-intensive. BBB permeability is associated with diverse chemical properties of compounds. However, BBB permeability prediction models have been developed using small datasets and limited features, which are usually not practical due to their low coverage of chemical diversity of compounds. Aim of this study is to develop a BBB permeability prediction model using a large dataset for practical applications. This model can be used for facilitated compound screening in the early stage of brain drug discovery. RESULTS: A dataset of 7162 compounds with BBB permeability (5453 BBB+ and 1709 BBB-) was compiled from the literature, where BBB+ and BBB- denote BBB-permeable and non-permeable compounds, respectively. We trained a machine learning model based on Light Gradient Boosting Machine (LightGBM) algorithm and achieved an overall accuracy of 89%, an area under the curve (AUC) of 0.93, specificity of 0.77 and sensitivity of 0.93, when 10-fold cross-validation was performed. The model was further evaluated using 74 central nerve system compounds (39 BBB+ and 35 BBB-) obtained from the literature and showed an accuracy of 90%, sensitivity of 0.85 and specificity of 0.94. Our model outperforms over existing BBB permeability prediction models. AVAILABILITYAND IMPLEMENTATION: The prediction server is available at http://ssbio.cau.ac.kr/software/bbb.

KR-39038, a Novel GRK5 Inhibitor, Attenuates Cardiac Hypertrophy and Improves Cardiac Function in Heart Failure.

G protein-coupled receptor kinase 5 (GRK5) has been considered as a potential target for the treatment of heart failure as it has been reported to be an important regulator of pathological cardiac hypertrophy. To discover novel scaffolds that selectively inhibit GRK5, we have identified a novel small molecule inhibitor of GRK5, KR-39038 [7-((3-((4-((3-aminopropyl)amino)butyl)amino)propyl) amino)-2-(2-chlorophenyl)-6-fluoroquinazolin-4(3H)-one]. KR-39038 exhibited potent inhibitory activity (IC50 value=0.02 µM) against GRK5 and significantly inhibited angiotensin II-induced cellular hypertrophy and HDAC5 phosphorylation in neonatal cardiomyocytes. In the pressure overload-induced cardiac hypertrophy mouse model, the daily oral administration of KR-39038 (30 mg/kg) for 14 days showed a 43% reduction in the left ventricular weight. Besides, KR-39038 treatment (10 and 30 mg/kg/ day, p.o.) showed significant preservation of cardiac function and attenuation of myocardial remodeling in a rat model of chronic heart failure following coronary artery ligation. These results suggest that potent GRK5 inhibitor could effectively attenuate both cardiac hypertrophy and dysfunction in experimental heart failure, and KR-39038 may be useful as an effective GRK5 inhibitor for pharmaceutical applications.

DeepHIT: a deep learning framework for prediction of hERG-induced cardiotoxicity.

MOTIVATION: Blockade of the human ether-à-go-go-related gene (hERG) channel by small compounds causes a prolonged QT interval that can lead to severe cardiotoxicity and is a major cause of the many failures in drug development. Thus, evaluating the hERG-blocking activity of small compounds is important for successful drug development. To this end, various computational prediction tools have been developed, but their prediction performances in terms of sensitivity and negative predictive value (NPV) need to be improved to reduce false negative predictions. RESULTS: We propose a computational framework, DeepHIT, which predicts hERG blockers and non-blockers for input compounds. For the development of DeepHIT, we generated a large-scale gold-standard dataset, which includes 6632 hERG blockers and 7808 hERG non-blockers. DeepHIT is designed to contain three deep learning models to improve sensitivity and NPV, which, in turn, produce fewer false negative predictions. DeepHIT outperforms currently available tools in terms of accuracy (0.773), MCC (0.476), sensitivity (0.833) and NPV (0.643) on an external test dataset. We also developed an in silico chemical transformation module that generates virtual compounds from a seed compound, based on the known chemical transformation patterns. As a proof-of-concept study, we identified novel urotensin II receptor (UT) antagonists without hERG-blocking activity derived from a seed compound of a previously reported UT antagonist (KR-36676) with a strong hERG-blocking activity. In summary, DeepHIT will serve as a useful tool to predict hERG-induced cardiotoxicity of small compounds in the early stages of drug discovery and development. AVAILABILITY AND IMPLEMENTATION: https://bitbucket.org/krictai/deephit and https://bitbucket.org/krictai/chemtrans. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Genomic and metabolic analysis of Komagataeibacter xylinus DSM 2325 producing bacterial cellulose nanofiber.

Bacterial cellulose nanofiber (CNF) is a polymer with a wide range of potential industrial applications. Several Komagataeibacter species, including Komagataeibacter xylinus as a model organism, produce CNF. However, the industrial application of CNF has been hampered by inefficient CNF production, necessitating metabolic engineering for the enhanced CNF production. Here, we present complete genome sequence and a genome-scale metabolic model KxyMBEL1810 of K. xylinus DSM 2325 for metabolic engineering applications. Genome analysis of this bacterium revealed that a set of genes associated with CNF biosynthesis and regulation were present in this bacterium, which were also conserved in another six representative Komagataeibacter species having complete genome information. To better understand the metabolic characteristics of K. xylinus DSM 2325, KxyMBEL1810 was reconstructed using genome annotation data, relevant computational resources and experimental growth data generated in this study. Random sampling and correlation analysis of the KxyMBEL1810 predicted pgi and gnd genes as novel overexpression targets for the enhanced CNF production. Among engineered K. xylinus strains individually overexpressing heterologous pgi and gnd genes, either from Escherichia coli or Corynebacterium glutamicum, batch fermentation of a strain overexpressing the E. coli pgi gene produced 3.15 g/L of CNF in a complex medium containing glucose, which was the best CNF concentration achieved in this study, and 115.8% higher than that (1.46 g/L) obtained from the control strain. Genome sequence data and KxyMBEL1810 generated in this study should be useful resources for metabolic engineering of K. xylinus for the enhanced CNF production.

Deep learning enables high-quality and high-throughput prediction of enzyme commission numbers.

High-quality and high-throughput prediction of enzyme commission (EC) numbers is essential for accurate understanding of enzyme functions, which have many implications in pathologies and industrial biotechnology. Several EC number prediction tools are currently available, but their prediction performance needs to be further improved to precisely and efficiently process an ever-increasing volume of protein sequence data. Here, we report DeepEC, a deep learning-based computational framework that predicts EC numbers for protein sequences with high precision and in a high-throughput manner. DeepEC takes a protein sequence as input and predicts EC numbers as output. DeepEC uses 3 convolutional neural networks (CNNs) as a major engine for the prediction of EC numbers, and also implements homology analysis for EC numbers that cannot be classified by the CNNs. Comparative analyses against 5 representative EC number prediction tools show that DeepEC allows the most precise prediction of EC numbers, and is the fastest and the lightest in terms of the disk space required. Furthermore, DeepEC is the most sensitive in detecting the effects of mutated domains/binding site residues of protein sequences. DeepEC can be used as an independent tool, and also as a third-party software component in combination with other computational platforms that examine metabolic reactions.

Expanded synthetic small regulatory RNA expression platforms for rapid and multiplex gene expression knockdown.

Synthetic small regulatory RNA (sRNA) can efficiently downregulate target gene expression at translational level in metabolic engineering, but cannot be used in engineered strain already having incompatible plasmid(s). To address this problem and make the sRNA gene expression modulation platform universally applicable, we report the development and applications of expanded synthetic sRNA expression platforms for rapid, multiplexed and genome-scale target gene knockdown in engineered Escherichia coli. As proof-of-concept, high performance strains capable of producing L-proline (54.1 g l-1) and L-threonine (22.9 g l-1) are rapidly developed by combinatorial knockdown of up to three genes via one-step co-transformation of sRNA expression vectors. Furthermore, a genome-scale sRNA library targeting 1,858 E. coli genes is employed to construct crude violacein (5.19 g l-1) and indigo (135 mg l-1) producers by high-throughput colorimetric screening. These examples demonstrate that the expanded sRNA expression vectors developed here enables rapid development of chemical overproducers regardless of plasmid compatibility.

Revisiting Statistical Design and Analysis in Scientific Research.

Statistics is essential to design experiments and interpret experimental results. Inappropriate use of the statistical analysis, however, often leads to a wrong conclusion. This concept article revisits basic concepts of statistics and provides a brief guideline of applying the statistical analysis for scientific research from designing experiments to analyzing and presenting the data.

Deep learning improves prediction of drug-drug and drug-food interactions.

Drug interactions, including drug-drug interactions (DDIs) and drug-food constituent interactions (DFIs), can trigger unexpected pharmacological effects, including adverse drug events (ADEs), with causal mechanisms often unknown. Several computational methods have been developed to better understand drug interactions, especially for DDIs. However, these methods do not provide sufficient details beyond the chance of DDI occurrence, or require detailed drug information often unavailable for DDI prediction. Here, we report development of a computational framework DeepDDI that uses names of drug-drug or drug-food constituent pairs and their structural information as inputs to accurately generate 86 important DDI types as outputs of human-readable sentences. DeepDDI uses deep neural network with its optimized prediction performance and predicts 86 DDI types with a mean accuracy of 92.4% using the DrugBank gold standard DDI dataset covering 192,284 DDIs contributed by 191,878 drug pairs. DeepDDI is used to suggest potential causal mechanisms for the reported ADEs of 9,284 drug pairs, and also predict alternative drug candidates for 62,707 drug pairs having negative health effects. Furthermore, DeepDDI is applied to 3,288,157 drug-food constituent pairs (2,159 approved drugs and 1,523 well-characterized food constituents) to predict DFIs. The effects of 256 food constituents on pharmacological effects of interacting drugs and bioactivities of 149 food constituents are predicted. These results suggest that DeepDDI can provide important information on drug prescription and even dietary suggestions while taking certain drugs and also guidelines during drug development.

Occurrence, evolution, and functions of DNA phosphorothioate epigenetics in bacteria.

The chemical diversity of physiological DNA modifications has expanded with the identification of phosphorothioate (PT) modification in which the nonbridging oxygen in the sugar-phosphate backbone of DNA is replaced by sulfur. Together with DndFGH as cognate restriction enzymes, DNA PT modification, which is catalyzed by the DndABCDE proteins, functions as a bacterial restriction-modification (R-M) system that protects cells against invading foreign DNA. However, the occurrence of dnd systems across a large number of bacterial genomes and their functions other than R-M are poorly understood. Here, a genomic survey revealed the prevalence of bacterial dnd systems: 1,349 bacterial dnd systems were observed to occur sporadically across diverse phylogenetic groups, and nearly half of these occur in the form of a solitary dndBCDE gene cluster that lacks the dndFGH restriction counterparts. A phylogenetic analysis of 734 complete PT R-M pairs revealed the coevolution of M and R components, despite the observation that several PT R-M pairs appeared to be assembled from M and R parts acquired from distantly related organisms. Concurrent epigenomic analysis, transcriptome analysis, and metabolome characterization showed that a solitary PT modification contributed to the overall cellular redox state, the loss of which perturbed the cellular redox balance and induced Pseudomonas fluorescens to reconfigure its metabolism to fend off oxidative stress. An in vitro transcriptional assay revealed altered transcriptional efficiency in the presence of PT DNA modification, implicating its function in epigenetic regulation. These data suggest the versatility of PT in addition to its involvement in R-M protection.