Deep Learning Model for Pathogen Classification Using Feature Fusion and Data Augmentation

Implementation of Antimicrobial Stewardship Programs in Saudi Arabia: A Systematic Review

BACKGROUND

Antimicrobial resistance has highlighted the need for effective infectious disease strategies. Antimicrobial stewardship programs (ASPs) may reduce antibiotic resistance, adverse reactions, and treatment failures. This systematic review examines ASPs in Saudi Arabia, assessing their efficacy, challenges, and outcomes to improve antimicrobial use and patient care.

METHODS

Searches were carried out in the Ovid, MEDLINE, Embase, PsycInfo, Scopus, and Web of Science Core Collection databases for studies published from 2007 to July 2024, in Saudi Arabia, following the PRISMA guidelines. Studies that assessed ASPs' implementation, effectiveness, and outcomes in hospital settings were included.

RESULTS

Out of the 6080 titles identified, 14 studies met the inclusion criteria, covering different regions of the country, including Riyadh, Jeddah, Dhahran, Makkah, Al-Kharj, and a multi-regional study in Qassim and Riyadh. Various interventions were implemented by the ASPs, such as educational programs, audit and feedback, switching from intravenous to oral administration, and enhanced policies. These interventions collectively led to a decrease in the overall antimicrobial consumption and cost, and a reduction in cases with multidrug-resistant bacteria.

CONCLUSIONS

The findings of this review highlight the positive impact of ASPs in Saudi Arabia. However, addressing challenges such as data limitations and training gaps is essential to enhance their effectiveness. Expanding education and refining implementation strategies are crucial for ensuring their long-term success.

Rapid bacterial identification through volatile organic compound analysis and deep learning

BACKGROUND

The increasing antimicrobial resistance caused by the improper use of antibiotics poses a significant challenge to humanity. Rapid and accurate identification of microbial species in clinical settings is crucial for precise medication and reducing the development of antimicrobial resistance. This study aimed to explore a method for automatic identification of bacteria using Volatile Organic Compounds (VOCs) analysis and deep learning algorithms.

RESULTS

AlexNet, where augmentation is applied, produces the best results. The average accuracy rate for single bacterial culture classification reached 99.24% using cross-validation, and the accuracy rates for identifying the three bacteria in randomly mixed cultures were SA:98.6%, EC:98.58% and PA:98.99%, respectively.

CONCLUSION

This work provides a new approach to quickly identify bacterial microorganisms. Using this method can automatically identify bacteria in GC-IMS detection results, helping clinical doctors quickly detect bacterial species, accurately prescribe medication, thereby controlling epidemics, and minimizing the negative impact of bacterial resistance on society.

Interpretable CRISPR/Cas9 off-target activities with mismatches and indels prediction using BERT

Off-target effects of CRISPR/Cas9 can lead to suboptimal genome editing outcomes. Numerous deep learning-based approaches have achieved excellent performance for off-target prediction; however, few can predict the off-target activities with both mismatches and indels between single guide RNA (sgRNA) and target DNA sequence pair. In addition, data imbalance is a common pitfall for off-target prediction. Moreover, due to the complexity of genomic contexts, generating an interpretable model also remains challenged. To address these issues, firstly we developed a BERT-based model called CRISPR-BERT for enhancing the prediction of off-target activities with both mismatches and indels. Secondly, we proposed an adaptive batch-wise class balancing strategy to combat the noise exists in imbalanced off-target data. Finally, we applied a visualization approach for investigating the generalizable nucleotide position-dependent patterns of sgRNA-DNA pair for off-target activity. In our comprehensive comparison to existing methods on five mismatches-only datasets and two mismatches-and-indels datasets, CRISPR-BERT achieved the best performance in terms of AUROC and PRAUC. Besides, the visualization analysis demonstrated how implicit knowledge learned by CRISPR-BERT facilitates off-target prediction, which shows potential in model interpretability. Collectively, CRISPR-BERT provides an accurate and interpretable framework for off-target prediction, further contributes to sgRNA optimization in practical use for improved target specificity in CRISPR/Cas9 genome editing. The source code is available at https://github.com/BrokenStringx/CRISPR-BERT.

Identifying key soil characteristics for Francisella tularensis classification with optimized Machine learning models

Francisella tularensis (Ft) poses a significant threat to both animal and human populations, given its potential as a bioweapon. Current research on the classification of this pathogen and its relationship with soil physical-chemical characteristics often relies on traditional statistical methods. In this study, we leverage advanced machine learning models to enhance the prediction of epidemiological models for soil-based microbes. Our model employs a two-stage feature ranking process to identify crucial soil attributes and hyperparameter optimization for accurate pathogen classification using a unique soil attribute dataset. Optimization involves various classification algorithms, including Support Vector Machines (SVM), Ensemble Models (EM), and Neural Networks (NN), utilizing Bayesian and Random search techniques. Results indicate the significance of soil features such as clay, nitrogen, soluble salts, silt, organic matter, and zinc , while identifying the least significant ones as potassium, calcium, copper, sodium, iron, and phosphorus. Bayesian optimization yields the best results, achieving an accuracy of 86.5% for SVM, 81.8% for EM, and 83.8% for NN. Notably, SVM emerges as the top-performing classifier, with an accuracy of 86.5% for both Bayesian and Random Search optimizations. The insights gained from employing machine learning techniques enhance our understanding of the environmental factors influencing Ft's persistence in soil. This, in turn, reduces the risk of false classifications, contributing to better pandemic control and mitigating socio-economic impacts on communities.

HNetGO: protein function prediction via heterogeneous network transformer

Protein function annotation is one of the most important research topics for revealing the essence of life at molecular level in the post-genome era. Current research shows that integrating multisource data can effectively improve the performance of protein function prediction models. However, the heavy reliance on complex feature engineering and model integration methods limits the development of existing methods. Besides, models based on deep learning only use labeled data in a certain dataset to extract sequence features, thus ignoring a large amount of existing unlabeled sequence data. Here, we propose an end-to-end protein function annotation model named HNetGO, which innovatively uses heterogeneous network to integrate protein sequence similarity and protein-protein interaction network information and combines the pretraining model to extract the semantic features of the protein sequence. In addition, we design an attention-based graph neural network model, which can effectively extract node-level features from heterogeneous networks and predict protein function by measuring the similarity between protein nodes and gene ontology term nodes. Comparative experiments on the human dataset show that HNetGO achieves state-of-the-art performance on cellular component and molecular function branches.

Deep ensemble approach for pathogen classification in large-scale images using patch-based training and hyper-parameter optimization

Pathogenic bacteria present a major threat to human health, causing various infections and illnesses, and in some cases, even death. The accurate identification of these bacteria is crucial, but it can be challenging due to the similarities between different species and genera. This is where automated classification using convolutional neural network (CNN) models can help, as it can provide more accurate, authentic, and standardized results.In this study, we aimed to create a larger and balanced dataset by image patching and applied different variations of CNN models, including training from scratch, fine-tuning, and weight adjustment, and data augmentation through random rotation, reflection, and translation. The results showed that the best results were achieved through augmentation and fine-tuning of deep models. We also modified existing architectures, such as InceptionV3 and MobileNetV2, to better capture complex features. The robustness of the proposed ensemble model was evaluated using two data splits (7:2:1 and 6:2:2) to see how performance changed as the training data was increased from 10 to 20%. In both cases, the model exhibited exceptional performance. For the 7:2:1 split, the model achieved an accuracy of 99.91%, F-Score of 98.95%, precision of 98.98%, recall of 98.96%, and MCC of 98.92%. For the 6:2:2 split, the model yielded an accuracy of 99.94%, F-Score of 99.28%, precision of 99.31%, recall of 98.96%, and MCC of 99.26%. This demonstrates that automatic classification using the ensemble model can be a valuable tool for diagnostic staff and microbiologists in accurately identifying pathogenic bacteria, which in turn can help control epidemics and minimize their social and economic impact.

PredMHC: An Effective Predictor of Major Histocompatibility Complex Using Mixed Features

The major histocompatibility complex (MHC) is a large locus on vertebrate DNA that contains a tightly linked set of polymorphic genes encoding cell surface proteins essential for the adaptive immune system. The groups of proteins encoded in the MHC play an important role in the adaptive immune system. Therefore, the accurate identification of the MHC is necessary to understand its role in the adaptive immune system. An effective predictor called PredMHC is established in this study to identify the MHC from protein sequences. Firstly, PredMHC encoded a protein sequence with mixed features including 188D, APAAC, KSCTriad, CKSAAGP, and PAAC. Secondly, three classifiers including SGD, SMO, and random forest were trained on the mixed features of the protein sequence. Finally, the prediction result was obtained by the voting of the three classifiers. The experimental results of the 10-fold cross-validation test in the training dataset showed that PredMHC can obtain 91.69% accuracy. Experimental results on comparison with other features, classifiers, and existing methods showed the effectiveness of PredMHC in predicting the MHC.

Multi-scale deep learning for the imbalanced multi-label protein subcellular localization prediction based on immunohistochemistry images

MOTIVATION

The development of microscopic imaging techniques enables us to study protein subcellular locations from the tissue level down to the cell level, contributing to the rapid development of image-based protein subcellular location prediction approaches. However, existing methods suffer from intrinsic limitations, such as poor feature representation ability, data imbalanced issue, and multi-label classification problem, greatly impacting the model performance and generalization.

RESULTS

In this study, we propose MSTLoc, a novel multi-scale end-to-end deep learning model to identify protein subcellular locations in the imbalanced multi-label immunohistochemistry (IHC) images dataset. In our MSTLoc, we deploy a deep convolution neural network to extract multi-scale features from the IHC images, aggregate the high-level features and low-level features via feature fusion to sufficiently exploit the dependencies amongst various subcellular locations, and utilize Vision Transformer (ViT) to model the relationship amongst the features and enhance the feature representation ability. We demonstrate that the proposed MSTLoc achieves better performance than current state-of-the-art models in multi-label subcellular location prediction. Through feature visualization and interpretation analysis, we demonstrate that as compared with the hand-crafted features, the multi-scale deep features learnt from our model exhibit better ability in capturing discriminative patterns underlying protein subcellular locations, and the features from different scales are complementary for the improvement in performance. Finally, case study results indicate that our MSTLoc can successfully identify some biomarkers from proteins that are closely involved with cancer development. For the convenient use of our method, we establish a user-friendly webserver available at http://server.wei-group.net/ MSTLoc.

AVAILABILITY AND IMPLEMENTATION

http://server.wei-group.net/ MSTLoc.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Risk prediction of diabetes and pre-diabetes based on physical examination data

Diabetes is a metabolic disorder caused by insufficient insulin secretion and insulin secretion disorders. From health to diabetes, there are generally three stages: health, pre-diabetes and type 2 diabetes. Early diagnosis of diabetes is the most effective way to prevent and control diabetes and its complications. In this work, we collected the physical examination data from Beijing Physical Examination Center from January 2006 to December 2017, and divided the population into three groups according to the WHO (1999) Diabetes Diagnostic Standards: normal fasting plasma glucose (NFG) (FPG < 6.1 mmol/L), mildly impaired fasting plasma glucose (IFG) (6.1 mmol/L ≤ FPG < 7.0 mmol/L) and type 2 diabetes (T2DM) (FPG > 7.0 mmol/L). Finally, we obtained1,221,598 NFG samples, 285,965 IFG samples and 387,076 T2DM samples, with a total of 15 physical examination indexes. Furthermore, taking eXtreme Gradient Boosting (XGBoost), random forest (RF), Logistic Regression (LR), and Fully connected neural network (FCN) as classifiers, four models were constructed to distinguish NFG, IFG and T2DM. The comparison results show that XGBoost has the best performance, with AUC (macro) of 0.7874 and AUC (micro) of 0.8633. In addition, based on the XGBoost classifier, three binary classification models were also established to discriminate NFG from IFG, NFG from T2DM, IFG from T2DM. On the independent dataset, the AUCs were 0.7808, 0.8687, 0.7067, respectively. Finally, we analyzed the importance of the features and identified the risk factors associated with diabetes.

dPromoter-XGBoost: Detecting promoters and strength by combining multiple descriptors and feature selection using XGBoost

Promoters play an irreplaceable role in biological processes and genetics, which are responsible for stimulating the transcription and expression of specific genes. Promoter abnormalities have been found in some diseases, and the level of promoter-binding transcription factors can be used as a marker before a disease occurs. Hence, detecting promoters from DNA sequences has important biological significance, particular, distinguishing strong promoters can help to elucidate differences in gene expression and the mechanisms of specific diseases. With the introduction of third-generation sequencing, it is difficult to match the speed of sequencing to the speed of labeling promoters experimentally. Many computing models have been designed to fill this gap and identify unlabeled DNA. However, their feature representation methods are very singular, which cannot reflect the information contained in the original samples. With the aim of avoiding information loss, we propose a computational model based on multiple descriptors and feature selection to jointly express samples. It is worth mentioning that a new feature descriptor called K-mer word vector is defined. The promoter model of multiple feature descriptors dominated by K-mer word vector achieves similar performance to existing methods, the sensitivity of 85.72% can distinguish the promoter more effectively than other methods. Furthermore, the performance of the promoter strength has surpassed published methods, and accuracy of 77.00% greatly improves the ability to distinguish between strong and weak promoters.

A smart LED therapy device with an automatic facial acne vulgaris diagnosis based on deep learning and internet of things application

In low-level laser therapy, providing an optimal dosage and proposing a proper diagnosis before dermatological treatment are essential to reduce the side effects and potential dangers. In this article, a smart LED therapy system for automatic facial acne vulgaris diagnosis based on deep learning and Internet of Things application is proposed. The main goals of this study were to (1) develop an LED therapy device with different power densities and LED grid control; (2) propose a deep learning model based on modified ResNet50 and YOLOv2 for an automatic acne diagnosis; and (3) develop a smartphone application for facial photography image capture and LED therapy parameter configuration. Furthermore, a healthcare Internet of Things (H-IoT) platform for the connectivity between smartphone apps, the cloud server, and the LED therapy device is proposed to improve the efficiency of the treatment process. Experiments were conducted on test data sets divided by a cross-validation method to verify the feasibility of the proposed LED therapy system with automatic facial acne detection. The obtained results evidenced the practical application of the proposed LED therapy system for automatic acne diagnosis and H-IoT-based solutions.