In order to address antibiotics resistance, multi-drug resistance, and superbugs phenomena, our research explored the utility of Jamu ingredients on the molecular level to predict new natural antibiotic candidates. Jamu is one of the popular traditional medicines from Indonesia, with different therapeutics usage including curing diseases caused by bacterial infection. We used three types of machine learning methods, such as Random Forest (RF), Support Vector Machine (SVM), Deep Learning (DL), to classify Jamu formulas according to their effectiveness against different types of bacterial diseases. The best accuracy for RF, SVM, and DL models are 89%, 84%, and 80%, respectively. We extracted the potential compounds based on the best model as candidate antibiotics corresponding to five groups of efficacies, e.g., digestive systems, respiratory systems, reproductive systems, skin and soft tissue, and urinary systems. Overall, we mined 111 compounds, and many of them could be validated by published literature, and considering structural similarities with known antibiotics.

Sleep screening is not only a major tool in the assessment of pathophysiology, but also a bridge between the central neuronal systems and behaviour/cognition. Automatic sleep staging is an alternative for the time-consuming gold standard manual scoring procedure. Most of the existing works designed such procedure by using deep neural networks without considering the medical criterion of the sleep staging task. We argue that capturing the stage-specific features which meet the criterion is of significant importance for the automatic sleep staging alternative. In this work we propose an end-to-end sleep staging simulator based on mixed neural networks, i.e., CNN, LSTM, and Transformer. The framework consists of two subnetworks: stage dependent feature mapping network which is constructed by the idea of physiological sleep nature, and an attention-based parallel staging network. Moreover, we adopt a mixed precision training strategy to quantize the model for exploring feasible usage in the clinical settings. Through an experiment with a large EEG database (Sleep Heart Health Study), the proposed method has a competitive stage scoring performance, especially in stages Wake, N2, and N3, with higher precision of 0.92, 0.85, and 0.86, respectively. Our study proves that the quantized model has potential capability for further application in the clinical staging task.

The gut microbiota is crucial for human physiology and susceptibility to diseases. Knowing the AMR phenotype canfacilitate the understanding of the impact of antibiotics administration on the gut microbiota. Nowadays, whole-genome sequencing for antibiotic susceptibility testing (WGS-AST) is widely used in clinical microbiology to predict the AMR phenotype. To release the limitations of the genomic information and improve the WGS-AST prediction, we propose an integrated multi-omics approach, employing a deep generative neural network (VAE: variational auto-encoder). We evaluate the proposed approach by two machine learning techniques (i.e., K-means for clustering and Random Forest for classification). Our evaluation results show that the integrated multi-omics approach achieves relatively better performance than the conventional WGS-AST. Moreover, the integrated multi-omics approach is able to visually reveal AMR phenotype of the gutmicrobiota via antibacterial spectrum. Our work provides evidence that multi-omics information is useful to enhance the WGS-AST prediction.

Recently, deep learning-based methods have been successfully proposed for electrophysiology signal-based sleep staging with promising results. Most existing methods use convolutional layers and recurrent-based architectures to implement a model structure from feature extraction to sequence signal classification. In this study, we propose a method of segmenting electroencephalogram (EEG) and electrooculogram (EOG) data according to frequency bands and construct a Transformer based automatic sleep classification model on top of it. The results show that the classifications of the stage Wake, N3, and REM outperform the state-of-art works, with the Fl-scores of 0.92, 0.85 and 0.91. Our work is the first attempt to explore the feasibility of a truth-involved Transformer-based model with a large-scale sleep database.

Due to the inherent variability such as inter-users anatomical variability and the inter-systems differences, the design of new EEG-based index and a reliable model for sleep stages classification is still the main topic in sleep science. The unsupervised deep learning framework-variational autoencoder (VAE) which can capture the major characteristics of the input by imposing a Gaussian prior distribution on the latent features is suitable in EEG characterization and reconstruction. Although vanilla VAE and convolutional autoencoder (CAE) have been tried, it has yet been discussed that whether a deep structure or a multi-scale structure is more appropriate. In this paper, we constructed a shallow iVAE model, which will capture the multi-scale features of the spectrogram of EEG by replacing the main structure in encoder and decoder with the inception-like structure. By comparing with the vanilla VAE and the CAE, a more accurate reconstruction and a better classification using the latent features of the iVAE can be confirmed.

In this paper, we propose a novel biclustering approach called BiClusO. Biclustering can be applied to various types of bipartite data such as gene-condition or gene-disease relations. For example, we applied BiClusO to bipartite relations between species and volatile organic compounds (VOCs). VOCs, which are emitted by different species, have huge environmental and ecological impacts. The biosynthesis of VOCs depends on different metabolic pathways which can be used to categorize the species. A previous study related to the KNApSAcK VOC database classified microorganisms based on their VOC profiles, which confirmed the consistency between VOC-based and pathogenicity-based classifications. However, due to limited data, classification of all species in terms of VOC profiles was not performed. In this study, we enriched our database with additional data collected from different online sources and journals. Then, by applying BiClusO to species-VOC relational data, we determined that VOC-based classification is consistent with taxonomy-based classification of the species. We also assessed the diversity of VOC pathways across different kingdoms of species.

A biomarker (short for biological marker) is a medical sign of a disease or condition which indicates a normal or abnormal state of a body. The biomarker is a key factor in the analysis of diseases and also for analyzing inter disease relations. In the previous study, we designed and developed a human biomarker (metabolites and proteins) database and the database is currently available online. This work was supported by the Ministry of Education, Japan and NAIST Big Data Project. We have used our previously developed database and collected 486 human biomarkers and their respective diseases. We determined the similarity among NCBI disease classes based on associated biomarker fingerprints. For this purpose, we collected biomarker PubChem IDs and using them downloaded the SDF files in a batch, then with those molecular description files determined their atom pair fingerprints using ChemmineR package. We constructed a network of biomarkers based on Tanimoto similarity between their fingerprints and applied DPclusO algorithm to find clusters consisting of biomarkers with similar chemical structures. We also conducted hierarchical clustering of the biomarkers. We categorized all the diseases in our data into 18 NCBI disease classes. Combining all information, we finally determined inter disease relations based on structural similarity between biomarkers.

Cardiotoxicity caused by drug candidates and chemical compounds that block hERG channels may lead to malignant ventricular arrhythmias and even sudden cardiac death (SCD). Various in-silico models have been built to predict the cardiotoxicity during early stages of drug design. The largest public database of hERG-related compounds by integrating several major databases has been constructed recently, which made it possible to build more sophisticated machine learning models for accurate prediction of cardiotoxicity. Here we developed a novel molecular graph convolution neural network (MGCNN) model, based on the new integrated database. The MGCNN models were built by altering the number of graph convolutional layers (GC) from 1 to 5. A random forest (RF) model input with the extended-connectivity fingerprint (ECFP) of different maximal radii (1 ~ 5) was built to enable a direct comparison with the MCGNN models. We found that the MGCNN model with 2 GCs has the best performance in terms of the ROC-AUC-score (0.84), whereas the RF model input with ECFP has a stable performance (0.77 ~ 0.80) over the preset radii. The machine learning models promise a potential new approach for harnessing the big data to achieve accurate prediction of drug cardiotoxicity.

BACKGROUND: Alkaloids, a class of organic compounds that contain nitrogen bases, are mainly synthesized as secondary metabolites in plants and fungi, and they have a wide range of bioactivities. Although there are thousands of compounds in this class, few of their biosynthesis pathways are fully identified. In this study, we constructed a model to predict their precursors based on a novel kind of neural network called the molecular graph convolutional neural network. Molecular similarity is a crucial metric in the analysis of qualitative structure-activity relationships. However, it is sometimes difficult for current fingerprint representations to emphasize specific features for the target problems efficiently. It is advantageous to allow the model to select the appropriate features according to data-driven decisions for extracting more useful information, which influences a classification or regression problem substantially. RESULTS: In this study, we applied a neural network architecture for undirected graph representation of molecules. By encoding a molecule as an abstract graph and applying "convolution" on the graph and training the weight of the neural network framework, the neural network can optimize feature selection for the training problem. By incorporating the effects from adjacent atoms recursively, graph convolutional neural networks can extract the features of latent atoms that represent chemical features of a molecule efficiently. In order to investigate alkaloid biosynthesis, we trained the network to distinguish the precursors of 566 alkaloids, which are almost all of the alkaloids whose biosynthesis pathways are known, and showed that the model could predict starting substances with an averaged accuracy of 97.5%. CONCLUSION: We have showed that our model can predict more accurately compared to the random forest and general neural network when the variables and fingerprints are not selected, while the performance is comparable when we carefully select 507 variables from 18000 dimensions of descriptors. The prediction of pathways contributes to understanding of alkaloid synthesis mechanisms and the application of graph based neural network models to similar problems in bioinformatics would therefore be beneficial. We applied our model to evaluate the precursors of biosynthesis of 12000 alkaloids found in various organisms and found power-low-like distribution.

Various medicinal plants are available in Bangladesh and these plants are used as traditional medicines for healing and health maintenance. Unani is one of the traditional medicine systems popular among Bangladeshi people because of its high success rate. Disease phenotype is changing constantly. It is Challenging for researchers to get the right medicinal ingredients, for the right disease, within a reasonable time. So we need to analyze the right plants for the right disease based on the existing formulas and to find out the relationship between plant and disease. The predicted plant-disease relations will help the health researcher or pharmacist for finding new drugs for new diseases. In our datasets, we have 409 plants, which are used as ingredients of 609 Unani formulas. Based on 609 formulas, we enlisted and sorted the relationship between diseases and plants. We assigned 609 Unani formulas to 18 National Center for Biotechnology Information (NCBI) disease classes. We then constructed the network of Unani formulas based on their ingredient similarity and applied DPclusO algorithm to find clusters. Clusters are associated with dominant disease and dominant plants by voting thus we established relations between plants and diseases. We predicted associations between 12 diseases and 151 plants. We validated our prediction based on the global set of Unani formulas and obtained 85.57% accuracy

A wearable thermometry for core body temperature (CBT) measurement has both healthcare and clinical applications. On the basis of the mechanism of bioheat transfer, we earlier designed and improved a wearable thermometry using the dual-heat-flux method for CBT measurement. In this study, this thermometry is examined experimentally. We studied a fast-changing CBT measurement (FCCM, 55 min, 12 subjects) inside a thermostatic chamber and performed long-term monitoring of CBT (LTM, 24 h, six subjects). When compared with a reference, the CoreTemp CM-210 by Terumo, FCCM shows 0.07 °C average difference and a 95% CI of [-0.27, 0.12] °C. LTM shows no significant difference in parameters for the inference of circadian rhythm. The FCCM and LTM both simulated scenarios in which this thermometry could be used for intensive monitoring and daily healthcare, respectively. The results suggest that because of its convenient design, this thermometry may be an ideal choice for conventional CBT measurements.

In this paper, we present an unobtrusive cuffless blood pressure (BP) monitoring system based on pulse arrival time (PAT) for facilitating long-term home BP monitoring. The proposed system consists of an electrocardiograph (ECG), a photoplethysmograph (PPG), and a control circuit with a Bluetooth module, all of which are mounted on a common armchair to measure ECG and PPG signals from users while sitting on the armchair in order to calculate continuous PAT. Considering the good linear correlation of systolic BP (SBP) and the nonlinear correlation of diastolic BP (DBP) with PAT, a new BP estimation method was proposed. Ten subjects underwent BP monitoring experiments involving stationary sitting on a chair, lying on a bed, and pedaling using an ergometer in order to assess the accuracy of the estimated BP. A cuff-type BP monitor was used as reference in the experiments. Results showed that the mean difference of the estimated SBP and DBP was within 0.2 ± 5.8 mmHg ( p < 0.00001) and 0.4 ± 5.7 mmHg ( p < 0.00001), respectively, and the mean absolute difference of the estimated SBP and DBP were 4.4 and 4.6 mmHg, respectively, compared to references. Additionally, five subjects participated in data collections consisting of sitting on a chair twice a day for one month. Compared to the reference, the difference did not obviously increase along with time, even though individualized calibration was executed only once at the beginning. These results suggest that the proposed system has quite the potential for long-term home BP monitoring.