Deep-learning based spatio-temporal generative model on assessing state-of-health for Li-ion batteries with partially-cycled profiles.

Accurately estimating the state-of-health (SOH) of lithium-ion batteries is emerging as a hot topic because of the rapid increase in electric appliance usage. However, versatile applicability to various battery compositions and diverse cycling conditions, and prediction only with partial data still remain challenges. In this paper, a Deep-learning-based Graphical approach to Estimation of Lithium-ion batteries SOH (D-GELS) was developed to predict the SOH covering three cathode materials, LiFePO4, LiNiCoAlO2, and LiNiCOMnO2. D-GELS shows an accurate performance for SOH prediction, less than 0.012 of RMSE, was predicted regardless of cathode materials, and its applicability was confirmed. Furthermore, D-GELS was capable of predicting the SOH using partially-cycled data, since less than 0.046 of RMSE was observed even with 50% of the image missing. When using partially-cycled profiles, significant economic benefits can be seen in used battery management, as the number of assessed batteries increases greatly, leading to cost savings.

Deep-learning optimized DEOCSU suite provides an iterable pipeline for accurate ChIP-exo peak calling.

Recognizing binding sites of DNA-binding proteins is a key factor for elucidating transcriptional regulation in organisms. ChIP-exo enables researchers to delineate genome-wide binding landscapes of DNA-binding proteins with near single base-pair resolution. However, the peak calling step hinders ChIP-exo application since the published algorithms tend to generate false-positive and false-negative predictions. Here, we report the development of DEOCSU (DEep-learning Optimized ChIP-exo peak calling SUite), a novel machine learning-based ChIP-exo peak calling suite. DEOCSU entails the deep convolutional neural network model which was trained with curated ChIP-exo peak data to distinguish the visualized data of bona fide peaks from false ones. Performance validation of the trained deep-learning model indicated its high accuracy, high precision and high recall of over 95%. Applying the new suite to both in-house and publicly available ChIP-exo datasets obtained from bacteria, eukaryotes and archaea revealed an accurate prediction of peaks containing canonical motifs, highlighting the versatility and efficiency of DEOCSU. Furthermore, DEOCSU can be executed on a cloud computing platform or the local environment. With visualization software included in the suite, adjustable options such as the threshold of peak probability, and iterable updating of the pre-trained model, DEOCSU can be optimized for users' specific needs.

A Machine Learning Approach Reveals a Microbiota Signature for Infection with Mycobacterium avium subsp. paratuberculosis in Cattle.

Although Mycobacterium avium subsp. paratuberculosis (MAP) has threatened public health and the livestock industry, the current diagnostic tools (e.g., fecal PCR and enzyme-linked immunosorbent assay [ELISA]) for MAP infection have some limitations, such as inconsistent results due to intermittent bacterial shedding or low sensitivity during the early stage of infection. Therefore, this study aimed to develop a novel biomarker focusing on elucidating the gut microbial signature of MAP-positive ruminants, since the clinical signs of MAP infection are closely related to dysbiosis. 16S rRNA-based gut microbial community analysis revealed both a decrease in microbial diversity and the emergence of several distinct taxa following MAP infection. To determine the discriminant taxa diagnostic of MAP infection, machine learning-based feature selection and predictive model construction were applied to taxon abundance data or their transformed derivatives. The selected taxa, such as Clostridioides (formerly Clostridium) difficile, were used to build models using a support vector machine, linear support vector classification, k-nearest neighbor, and random forest with 10-fold cross-validation. The receiver operating characteristic-area under the curve (ROC-AUC) analysis of the models revealed their high accuracy, up to approximately 96%. Collectively, taxonomic signatures of cattle gut microbiotas according to MAP infection status could be identified by feature selection tools and applied to establish a predictive model for the infection state. IMPORTANCE Due to the limitations, such as intermittent bacterial shedding or poor sensitivity, of the current diagnostic tools for Johne's disease, novel biomarkers are urgently needed to aid control of the disease. Here, we explored the fecal microbiota of Johne's disease-affected cattle and tried to discover distinct microbial characteristics which have the potential to be novel noninvasive biomarkers. Through 16S rRNA sequencing and machine learning approaches, a dozen taxa were selected as taxonomic signatures to discriminate the disease state. In addition, when constructing predictive models using relative abundance data of the corresponding taxa, the models showed high accuracy for classification, even including animals with subclinical infection. Thus, our study suggested novel noninvasive microbiological biomarkers that are robustly expressed regardless of subclinical infection and the applicability of machine learning for diagnosis of Johne's disease.