Deep-HPI-pred: An R-Shiny applet for network-based classification and prediction of Host-Pathogen protein-protein interactions

Host-pathogen interactions (HPIs) are vital in numerous biological activities and are intrinsically linked to the onset and progression of infectious diseases. HPIs are pivotal in the entire lifecycle of diseases: from the onset of pathogen introduction, navigating through the mechanisms that bypass host cellular defenses, to its subsequent proliferation inside the host. At the heart of these stages lies the synergy of proteins from both the host and the pathogen. By understanding these interlinking protein dynamics, we can gain crucial insights into how diseases progress and pave the way for stronger plant defenses and the swift formulation of countermeasures. In the framework of current study, we developed a web-based R/Shiny app, Deep-HPI-pred, that uses network-driven feature learning method to predict the yet unmapped interactions between pathogen and host proteins. Leveraging citrus and CLas bacteria training datasets as case study, we spotlight the effectiveness of Deep-HPI-pred in discerning Protein-protein interaction (PPIs) between them. Deep-HPI-pred use Multilayer Perceptron (MLP) models for HPI prediction, which is based on a comprehensive evaluation of topological features and neural network architectures. When subjected to independent validation datasets, the predicted models consistently surpassed a Matthews correlation coefficient (MCC) of 0.80 in host-pathogen interactions. Remarkably, the use of Eigenvector Centrality as the leading topological feature further enhanced this performance. Further, Deep-HPI-pred also offers relevant gene ontology (GO) term information for each pathogen and host protein within the system. This protein annotation data contributes an additional layer to our understanding of the intricate dynamics within host-pathogen interactions. In the additional benchmarking studies, the Deep-HPI-pred model has proven its robustness by consistently delivering reliable results across different host-pathogen systems, including plant-pathogens (accuracy of 98.4% and 97.9%), human-virus (accuracy of 94.3%), and animal-bacteria (accuracy of 96.6%) interactomes. These results not only demonstrate the model's versatility but also pave the way for gaining comprehensive insights into the molecular underpinnings of complex host-pathogen interactions. Taken together, the Deep-HPI-pred applet offers a unified web service for both identifying and illustrating interaction networks. Deep-HPI-pred applet is freely accessible at its homepage: https://cbi.gxu.edu.cn/shiny-apps/Deep-HPI-pred/ and at github: https://github.com/tahirulqamar/Deep-HPI-pred.

A dual data stream hybrid neural network for classifying pathological images of lung adenocarcinoma

Lung cancer has seriously threatened human health due to its high lethality and morbidity. Lung adenocarcinoma, in particular, is one of the most common subtypes of lung cancer. Pathological diagnosis is regarded as the gold standard for cancer diagnosis. However, the traditional manual screening of lung cancer pathology images is time consuming and error prone. Computer-aided diagnostic systems have emerged to solve this problem. Current research methods are unable to fully exploit the beneficial features inherent within patches, and they are characterized by high model complexity and significant computational effort. In this study, a deep learning framework called Multi-Scale Network (MSNet) is proposed for the automatic detection of lung adenocarcinoma pathology images. MSNet is designed to efficiently harness the valuable features within data patches, while simultaneously reducing model complexity, computational demands, and storage space requirements. The MSNet framework employs a dual data stream input method. In this input method, MSNet combines Swin Transformer and MLP-Mixer models to address global information between patches and the local information within each patch. Subsequently, MSNet uses the Multilayer Perceptron (MLP) module to fuse local and global features and perform classification to output the final detection results. In addition, a dataset of lung adenocarcinoma pathology images containing three categories is created for training and testing the MSNet framework. Experimental results show that the diagnostic accuracy of MSNet for lung adenocarcinoma pathology images is 96.55 %. In summary, MSNet has high classification performance and shows effectiveness and potential in the classification of lung adenocarcinoma pathology images.

Predicting reservoir sedimentation using multilayer perceptron - Artificial neural network model with measured and forecasted hydrometeorological data in Gibe-III reservoir, Omo-Gibe River basin, Ethiopia

The estimation and prediction of the amount of sediment accumulated in reservoirs are imperative for sustainable reservoir sedimentation planning and management and to minimize reservoir storage capacity loss. The main objective of this study was to estimate and predict reservoir sedimentation using multilayer perceptron-artificial neural network (MLP-ANN) and random forest regressor (RFR) models in the Gibe-III reservoir, Omo-Gibe River basin. The hydrological and meteorological parameters considered for the estimation and prediction of reservoir sedimentation include annual rainfall, annual water inflow, minimum reservoir level, and reservoir storage capacity. The MLP-ANN and RFR models were employed to estimate and predict the amount of sediment accumulated in the Gibe-III reservoir using time series data from 2014 to 2022. ANN-architecture N4-100-100-1 with a coefficient of determination (R2) of 0.97 for the (80, 20) train-test approach was chosen because it showed better performance both in training and testing (validation) the model. The MLP-ANN and RFR models' performance evaluation was conducted using MAE, MSE, RMSE, and R2. The models' evaluation result revealed that the MLP-ANN model outperformed the RFR model. Regarding the train data simulation of MLP-ANN and RFR shown R2 (0.99) and RMSE (0.77); and R2 (0.97) and RMSE (1.80), respectively. On the other hand, the test data simulation of MLP-ANN and RFR demonstrated R2 (0.98) and RMSE (1.32); and R2 (0.96) and RMSE (2.64), respectively. The MLP-ANN model simulation output indicates that the amount of sediment accumulation in the Gibe-III reservoir will increase in the future, reaching 110 MT in 2030-2031, 130 MT in 2050-2051, and above 137 MTin 2071-2072.

Computational modeling of light processing in the habenula and dorsal raphe based on laser ablation of functionally-defined cells

BACKGROUND

The habenula is a major regulator of serotonergic neurons in the dorsal raphe, and thus of brain state. The functional connectivity between these regions is incompletely characterized. Here, we use the ability of changes in irradiance to trigger reproducible changes in activity in the habenula and dorsal raphe of zebrafish larvae, combined with two-photon laser ablation of specific neurons, to establish causal relationships.

RESULTS

Neurons in the habenula can show an excitatory response to the onset or offset of light, while neurons in the anterior dorsal raphe display an inhibitory response to light, as assessed by calcium imaging. The raphe response changed in a complex way following ablations in the dorsal habenula (dHb) and ventral habenula (vHb). After ablation of the ON cells in the vHb (V-ON), the raphe displayed no response to light. After ablation of the OFF cells in the vHb (V-OFF), the raphe displayed an excitatory response to darkness. After ablation of the ON cells in the dHb (D-ON), the raphe displayed an excitatory response to light. We sought to develop in silico models that could recapitulate the response of raphe neurons as a function of the ON and OFF cells of the habenula. Early attempts at mechanistic modeling using ordinary differential equation (ODE) failed to capture observed raphe responses accurately. However, a simple two-layer fully connected neural network (NN) model was successful at recapitulating the diversity of observed phenotypes with root-mean-squared error values ranging from 0.012 to 0.043. The NN model also estimated the raphe response to ablation of D-off cells, which can be verified via future experiments.

CONCLUSION

Lesioning specific cells in different regions of habenula led to qualitatively different responses to light in the dorsal raphe. A simple neural network is capable of mimicking experimental observations. This work illustrates the ability of computational modeling to integrate complex observations into a simple compact formalism for generating testable hypotheses, and for guiding the design of biological experiments.

Multilayer perceptron and support vector regression models for feline parturition date prediction

A crucial challenge in feline obstetric care is the accurate prediction of the parturition date during late pregnancy. The classic simple linear regression (SLR) model, which employed the fetal biparietal diameter (BPD) as the single input feature, was frequently applied for such prediction with limited accuracy. Since Multilayer Perceptron (MLP) and Support Vector Regression (SVR) are now two of the most potent scientific regression models, this study, for the first time, introduced such models as the new promising tools for feline parturition date prediction. The following features were candidate inputs for our models: biparietal diameter (BPD), litter size, and maternal weight. We observed and compared the performance results for each model. As the best-performed model, MLP delivered the highest coefficient score (0.972 ± 0.006), lowest mean absolute error score (1.110 ± 0.060), and lowest mean squared error score (1.540 ± 0.141), respectively. For the first time in this study, BPD, litter size, and maternal weight were considered the essential features for the innovative MLP and SVR modeling. With the optimized model parameters and the described analytical platform, further verification of these advanced models in feline obstetric practices is feasible.

Prediction of compressive strength of concrete based on improved artificial bee colony-multilayer perceptron algorithm

There are many factors that affect the compressive strength of concrete. The relationship between compressive strength and these factors is a complex nonlinear problem. Empirical formulas commonly used to predict the compressive strength of concrete are based on summarizing experimental data of several different mix proportions and curing periods, and their generality is poor. This article proposes an improved artificial bee colony algorithm (IABC) and a multilayer perceptron (MLP) coupled model for predicting the compressive strength of concrete. To address the shortcomings of the basic artificial bee colony algorithm, such as easily falling into local optima and slow convergence speed, this article introduces a Gaussian mutation operator into the basic artificial bee colony algorithm to optimize the initial honey source position and designs an MLP neural network model based on the improved artificial bee colony algorithm (IABC-MLP). Compared with traditional strength prediction models, the ABC-MLP model can better capture the nonlinear relationship of the compressive strength of concrete and achieve higher prediction accuracy when considering the compound effect of multiple factors. The IABC-MLP model built in this study is compared with the ABC-MLP and particle swarm optimization (PSO) coupling algorithms. The research shows that IABC can significantly improve the training and prediction accuracy of MLP. Compared with the ABC-MLP and PSO-MLP coupling models, the training accuracy of the IABC-MLP model is increased by 1.6% and 4.5%, respectively. This model is also compared with common individual learning algorithms such as MLP, decision tree (DT), support vector machine regression (SVR), and random forest algorithms (RF). Based on the comparison of prediction results, the proposed method shows excellent performance in all indicators and demonstrates the superiority of heuristic algorithms in predicting the compressive strength of concrete.

CSSNet: Cascaded spatial shift network for multi-organ segmentation

Multi-organ segmentation is vital for clinical diagnosis and treatment. Although CNN and its extensions are popular in organ segmentation, they suffer from the local receptive field. In contrast, MultiLayer-Perceptron-based models (e.g., MLP-Mixer) have a global receptive field. However, these MLP-based models employ fully connected layers with many parameters and tend to overfit on sample-deficient medical image datasets. Therefore, we propose a Cascaded Spatial Shift Network, CSSNet, for multi-organ segmentation. Specifically, we design a novel cascaded spatial shift block to reduce the number of model parameters and aggregate feature segments in a cascaded way for efficient and effective feature extraction. Then, we propose a feature refinement network to aggregate multi-scale features with location information, and enhance the multi-scale features along the channel and spatial axis to obtain a high-quality feature map. Finally, we employ a self-attention-based fusion strategy to focus on the discriminative feature information for better multi-organ segmentation performance. Experimental results on the Synapse (multiply organs) and LiTS (liver & tumor) datasets demonstrate that our CSSNet achieves promising segmentation performance compared with CNN, MLP, and Transformer models. The source code will be available at https://github.com/zkyseu/CSSNet.

Coastal groundwater quality prediction using objective-weighted WQI and machine learning approach

The water quality index (WQI) is a globally accepted guideline to indicate the water quality standard of any groundwater resource. Water levels in existing groundwater sources are declining in several coastal zones. Therefore, for monitoring water quality and improving water management, the prediction and identification of groundwater status by an effective technique with higher accuracy is urgently needed. Therefore, this research aims to find an effective model for WQI prediction by comparing entropy and critic weight-based WQI (ENW-WQI and CRITIC-WQI) with multi-layer perceptron artificial neural network (MLP-ANN) technique and also to identify contaminated zones using GIS. Initially, 1000 water sampling datasets with concentrations of several water quality parameters of different coastal blocks of eastern India during 2018 to 2022 are considered for the estimation of ENW-WQI and CRITIC-WQI. It shows 65% and 67% of the samples are excellent to good for drinking. ENW-WQI and CRITIC-WQI-based MLP-ANN models have been established considering different data portioning and hidden neuron numbers. Input variables and appropriate dataset partitioning with hidden neurons for models obtained from correlation and trial-error analysis. Spatial distribution maps are also produced for calculated WQIs using inverse distance weighted interpolation approaches. Three fitting models are obtained: ENW-WQI-MLP-ANN, CRITIC-WQI-MLP-ANN-I and CRITIC-WQI-MLP-ANN-II. CRITIC-WQI-MLP-ANN-II model (data ratio 85:15, network structure 6-12-1, R2 = 0.986, NSE = 0.98, and error rate 0.49%) provides the best accuracy in WQI prediction. The GIS-based WQI maps record several areas related to drinking water quality. The results of this research can help in planning the provision of safe drinking water in the future.

Hierarchical attention-guided multiscale aggregation network for infrared small target detection

All man-made flying objects in the sky, ships in the ocean can be regarded as small infrared targets, and the method of tracking them has been received widespread attention in recent years. In search of a further efficient method for infrared small target recognition, we propose a hierarchical attention-guided multiscale aggregation network (HAMANet) in this thesis. The proposed HAMANet mainly consists of a compound guide multilayer perceptron (CG-MLP) block embedded in the backbone net, a spatial-interactive attention module (SiAM), a pixel-interactive attention module (PiAM) and a contextual fusion module (CFM). The CG-MLP marked the width-axis, height-axis, and channel-axis, which can result in a better segmentation effect while reducing computational complexity. SiAM improves global semantic information exchange by increasing the connections between different channels, while PiAM changes the extraction of local key information features by enhancing information exchange at the pixel level. CFM fuses low-level positional information and high-level channel information of the target through coding to improve network stability and target feature utilization. Compared with other state-of-the-art methods on public infrared small target datasets, the results show that our proposed HAMANet has high detection accuracy and a low false-alarm rate.

Semantic representation and comparative analysis of physical activity sensor observations using MOX2-5 sensor in real and synthetic datasets: a proof-of-concept-study

The widespread use of devices like mobile phones and wearables allows for automatic monitoring of human daily activities, generating vast datasets that offer insights into long-term human behavior. A structured and controlled data collection process is essential to unlock the full potential of this information. While wearable sensors for physical activity monitoring have gained significant traction in healthcare, sports science, and fitness applications, securing diverse and comprehensive datasets for research and algorithm development poses a notable challenge. In this proof-of-concept study, we underscore the significance of semantic representation in enhancing data interoperability and facilitating advanced analytics for physical activity sensor observations. Our approach focuses on enhancing the usability of physical activity datasets by employing a medical-grade (CE certified) sensor to generate synthetic datasets. Additionally, we provide insights into ethical considerations related to synthetic datasets. The study conducts a comparative analysis between real and synthetic activity datasets, assessing their effectiveness in mitigating model bias and promoting fairness in predictive analysis. We have created an ontology for semantically representing observations from physical activity sensors and conducted predictive analysis on data collected using MOX2-5 activity sensors. Until now, there has been a lack of publicly available datasets for physical activity collected with MOX2-5 activity monitoring medical grade (CE certified) device. The MOX2-5 captures and transmits high-resolution data, including activity intensity, weight-bearing, sedentary, standing, low, moderate, and vigorous physical activity, as well as steps per minute. Our dataset consists of physical activity data collected from 16 adults (Male: 12; Female: 4) over a period of 30-45 days (approximately 1.5 months), yielding a relatively small volume of 539 records. To address this limitation, we employ various synthetic data generation methods, such as Gaussian Capula (GC), Conditional Tabular General Adversarial Network (CTGAN), and Tabular General Adversarial Network (TABGAN), to augment the dataset with synthetic data. For both the authentic and synthetic datasets, we have developed a Multilayer Perceptron (MLP) classification model for accurately classifying daily physical activity levels. The findings underscore the effectiveness of semantic ontology in semantic search, knowledge representation, data integration, reasoning, and capturing meaningful relationships between data. The analysis supports the hypothesis that the efficiency of predictive models improves as the volume of additional synthetic training data increases. Ontology and Generative AI hold the potential to expedite advancements in behavioral monitoring research. The data presented, encompassing both real MOX2-5 and its synthetic counterpart, serves as a valuable resource for developing robust methods in activity type classification. Furthermore, it opens avenues for exploration into research directions related to synthetic data, including model efficiency, detection of generated data, and considerations regarding data privacy.

A multilayer perceptron neural network approach for optimizing solar irradiance forecasting in Central Africa with meteorological insights

Promoting renewable energy sources, particularly in the solar industry, has the potential to address the energy shortfall in Central Africa. Nevertheless, a difficulty occurs due to the erratic characteristics of solar irradiance data, which is influenced by climatic fluctuations and challenging to regulate. The current investigation focuses on predicting solar irradiance on an inclined surface, taking into consideration the impact of climatic variables such as temperature, wind speed, humidity, and air pressure. The used methodology for this objective is Artificial Neural Network (ANN), and the inquiry is carried out in the metropolitan region of Douala. The data collection device used in this research is the meteorological station located at the IUT of Douala. This station was built as a component of the Douala sustainable city effort, in partnership with the CUD and the IRD. Data was collected at 30-min intervals for a duration of around 2 years, namely from January 17, 2019, to October 30, 2020. The aforementioned data has been saved in a database that underwent pre-processing in Excel and later employed MATLAB for the creation of the artificial neural network model. 80% of the available data was utilized for training the network, 15% was allotted for validation, and the remaining 5% was used for testing. Different combinations of input data were evaluated to ascertain their individual degrees of accuracy. The logistic Sigmoid function, with 50 hidden layer neurons, yielded a correlation coefficient of 98.883% between the observed and estimated sun irradiation. This function is suggested for evaluating the intensities of solar radiation at the place being researched and at other sites that have similar climatic conditions.

A deep learning-based theoretical protocol to identify potentially isoform-selective PI3Kα inhibitors

Phosphoinositide 3-kinase alpha (PI3Kα) is one of the most frequently dysregulated kinases known for their pivotal role in many oncogenic diseases. While the side effects linked to existing drugs against PI3Kα-induced cancers provide an avenue for further research, the significant structural conservation among PI3Ks makes it extremely difficult to develop new isoform-selective PI3Kα inhibitors. Embracing this challenge, we herein designed a hybrid protocol by integrating machine learning (ML) with in silico drug-designing strategies. A deep learning classification model was developed and trained on the physicochemical descriptors data of known PI3Kα inhibitors and used as a screening filter for a database of small molecules. This approach led us to the prediction of 662 compounds showcasing appropriate features to be considered as PI3Kα inhibitors. Subsequently, a multiphase molecular docking was applied to further characterize the predicted hits in terms of their binding affinities and binding modes in the targeted cavity of the PI3Kα. As a result, a total of 12 compounds were identified whereas the best poses highlighted the efficiency of these ligands in maintaining interactions with the crucial residues of the protein to be targeted for the inhibition of associated activity. Notably, potential activity of compound 12 in counteracting PI3Kα function was found in a previous in vitro study. Following the drug-likeness and pharmacokinetic characterizations, six compounds (compounds 1, 2, 3, 6, 7, and 11) with suitable ADME-T profiles and promising bioavailability were selected. The mechanistic studies in dynamic mode further endorsed the potential of identified hits in blocking the ATP-binding site of the receptor with higher binding affinities than the native inhibitor, alpelisib (BYL-719), particularly the compounds 1, 2, and 11. These outcomes support the reliability of the developed classification model and the devised computational strategy for identifying new isoform-selective drug candidates for PI3Kα inhibition.

Exploring recurrence quantification analysis and fractal dimension algorithms for diagnosis of encephalopathy

Electroencephalography (EEG) is a crucial non-invasive medical tool for diagnosing neurological disorder called encephalopathy. There is a requirement for powerful signal processing algorithms as EEG patterns in encephalopathies are not specific to a particular etiology. As visual examination and linear methods of EEG analysis are not sufficient to get the subtle information regarding various neuro pathologies, non-linear analysis methods can be employed for exploring the dynamic, complex and chaotic nature of EEG signals. This work aims identifying and differentiating the patterns specific to cerebral dysfunctions associated with Encephalopathy using Recurrence Quantification Analysis and Fractal Dimension algorithms. This study analysed six RQA features, namely, recurrence rate, determinism, laminarity, diagonal length, diagonal entropy and trapping time and comparing them with fractal dimensions, namely, Higuchi's and Katz's fractal dimension. Fractal dimensions were found to be lower for encephalopathy cases showing decreased complexity when compared to that of normal healthy subjects. On the other hand, RQA features were found to be higher for encephalopathy cases indicating higher recurrence and more periodic patterns in EEGs of encephalopathy compared to that of normal healthy controls. The feature reduction was then performed using Principal Component Analysis and fed to three promising classifiers: SVM, Random Forest and Multi-layer Perceptron. The resultant system provides a practically realizable pipeline for the diagnosis of encephalopathy.

Insights the dominant contribution of biomass burning to methanol-soluble PM2.5 bounded oxidation potential based on multilayer perceptron neural network analysis in Xi'an, China

Atmospheric fine particulate matter (PM2.5) is associated with cardiorespiratory morbidity and mortality due to its ability to generate reactive oxygen species (ROS). Ambient PM2.5 samples were collected during heating and nonheating seasons in Xi'an, China, and the ROS-generation potential of PM2.5 was quantified using the dithiothreitol (DTT) assay. Additionally, positive matrix factorization combined with multilayer perceptron was employed to apportion sources contributing to the oxidation potential of PM2.5. Both the mass concentration of PM2.5 and the volume-based DTT activity (DTTv) were higher during the heating season than during the nonheating season. The primary contributors to DTTv were combustion (biomass and coal) sources during the heating season (>52 %), whereas secondary formation dominated DTT activity during the nonheating season (35.7 %). In addition, the secondary reaction process promoted the generation of intrinsic oxidation potential (OP) of sources. Among all the sources investigated (traffic source, industrial emission, mineral dust, biomass burning, secondary formation and coal combustion), the inherent oxidation potential of biomass burning was the highest, whereas that of mineral dust was the lowest. Our study indicates that anthropogenic sources, especially biomass burning, should be prioritized in PM2.5 toxicity control strategies.

Predicting the future land use and land cover changes for Saroor Nagar Watershed, Telangana, India, using open-source GIS

The dynamics of land use and land cover are profoundly affected by the growth, mobility, and demand of people. Thematic maps of land use and land cover (LULC) help planners account for conservation, concurrent uses, and land-use compressions by providing a reference for analysis, resource management, and prediction. The purpose of this research is to identify the transition of land-use changes in the Saroor Nagar Watershed between 2008 and 2014 using the Modules for Land Use Change Evaluation (MOLUSCE) plugin (MLP-ANN) model and to forecast and establish potential land-use changes for the years 2020 and 2026. To predict how these factors affected LULC from 2008 to 2014, MLP-ANN was trained with maps of DEM, slope, distance from the road, and distance to a waterbody. The projected and accurate LULC maps for 2020 have a Kappa value of 0.70 and a correctness percentage of 81.8%, indicating a high degree of accuracy. Changes in LULC are then predicted for the year 2026 using MLP-ANN, which shows a 17.4% increase in built-up area at the expense of vegetation and barren land. The results contribute to the development of sustainable plans for land use and resource management.

Protein language models can capture protein quaternary state

BACKGROUND

Determining a protein's quaternary state, i.e. the number of monomers in a functional unit, is a critical step in protein characterization. Many proteins form multimers for their activity, and over 50% are estimated to naturally form homomultimers. Experimental quaternary state determination can be challenging and require extensive work. To complement these efforts, a number of computational tools have been developed for quaternary state prediction, often utilizing experimentally validated structural information. Recently, dramatic advances have been made in the field of deep learning for predicting protein structure and other characteristics. Protein language models, such as ESM-2, that apply computational natural-language models to proteins successfully capture secondary structure, protein cell localization and other characteristics, from a single sequence. Here we hypothesize that information about the protein quaternary state may be contained within protein sequences as well, allowing us to benefit from these novel approaches in the context of quaternary state prediction.

RESULTS

We generated ESM-2 embeddings for a large dataset of proteins with quaternary state labels from the curated QSbio dataset. We trained a model for quaternary state classification and assessed it on a non-overlapping set of distinct folds (ECOD family level). Our model, named QUEEN (QUaternary state prediction using dEEp learNing), performs worse than approaches that include information from solved crystal structures. However, it successfully learned to distinguish multimers from monomers, and predicts the specific quaternary state with moderate success, better than simple sequence similarity-based annotation transfer. Our results demonstrate that complex, quaternary state related information is included in such embeddings.

CONCLUSIONS

QUEEN is the first to investigate the power of embeddings for the prediction of the quaternary state of proteins. As such, it lays out strengths as well as limitations of a sequence-based protein language model approach, compared to structure-based approaches. Since it does not require any structural information and is fast, we anticipate that it will be of wide use both for in-depth investigation of specific systems, as well as for studies of large sets of protein sequences. A simple colab implementation is available at: https://colab.

RESEARCH

google.com/github/Furman-Lab/QUEEN/blob/main/QUEEN_prediction_notebook.ipynb .

Prediction model for hepatocellular carcinoma recurrence after hepatectomy: Machine learning-based development and interpretation study

Background

Identifying patients with hepatocellular carcinoma (HCC) at high risk of recurrence after hepatectomy can help to implement timely interventional treatment. This study aimed to develop a machine learning (ML) model to predict the recurrence risk of HCC patients after hepatectomy.

Methods

We retrospectively collected 315 HCC patients who underwent radical hepatectomy at the Third Affiliated Hospital of Sun Yat-sen University from April 2013 to October 2017, and randomly divided them into the training and validation sets at a ratio of 7:3. According to the postoperative recurrence of HCC patients, the patients were divided into recurrence group and non-recurrence group, and univariate and multivariate logistic regression were performed for the two groups. We applied six machine learning algorithms to construct the prediction models and performed internal validation by 10-fold cross-validation. Shapley additive explanations (SHAP) method was applied to interpret the machine learning model. We also built a web calculator based on the best machine learning model to personalize the assessment of the recurrence risk of HCC patients after hepatectomy.

Results

A total of 13 variables were included in the machine learning models. The multilayer perceptron (MLP) machine learning model was proved to achieve optimal predictive value in test set (AUC = 0.680). The SHAP method displayed that γ-glutamyl transpeptidase (GGT), fibrinogen, neutrophil, aspartate aminotransferase (AST) and total bilirubin (TB) were the top 5 important factors for recurrence risk of HCC patients after hepatectomy. In addition, we further demonstrated the reliability of the model by analyzing two patients. Finally, we successfully constructed an online web prediction calculator based on the MLP machine learning model.

Conclusion

MLP was an optimal machine learning model for predicting the recurrence risk of HCC patients after hepatectomy. This predictive model can help identify HCC patients at high recurrence risk after hepatectomy to provide early and personalized treatment.

Diagnosis of Alzheimer's disease by joining dual attention CNN and MLP based on structural MRIs, clinical and genetic data

Alzheimer's disease (AD) is an irreversible central nervous degenerative disease, while mild cognitive impairment (MCI) is a precursor state of AD. Accurate early diagnosis of AD is conducive to the prevention and early intervention treatment of AD. Although some computational methods have been developed for AD diagnosis, most employ only neuroimaging, ignoring other data (e.g., genetic, clinical) that may have potential disease information. In addition, the results of some methods lack interpretability. In this work, we proposed a novel method (called DANMLP) of joining dual attention convolutional neural network (CNN) and multilayer perceptron (MLP) for computer-aided AD diagnosis by integrating multi-modality data of the structural magnetic resonance imaging (sMRI), clinical data (i.e., demographics, neuropsychology), and APOE genetic data. Our DANMLP consists of four primary components: (1) the Patch-CNN for extracting the image characteristics from each local patch, (2) the position self-attention block for capturing the dependencies between features within a patch, (3) the channel self-attention block for capturing dependencies of inter-patch features, (4) two MLP networks for extracting the clinical features and outputting the AD classification results, respectively. Compared with other state-of-the-art methods in the 5CV test, DANMLP achieves 93% and 82.4% classification accuracy for the AD vs. MCI and MCI vs. NC tasks on the ADNI database, which is 0.2%∼15.2% and 3.4%∼26.8% higher than that of other five methods, respectively. The individualized visualization of focal areas can also help clinicians in the early diagnosis of AD. These results indicate that DANMLP can be effectively used for diagnosing AD and MCI patients.

A structured combination of ensemble classifier and filter-based feature selection to improve breast cancer diagnosis

INTRODUCTION

Advances in technology have led to the emergence of computerized diagnostic systems as intelligent medical assistants. Machine learning approaches cannot replace professional humans, but they can change the treatment of diseases such as cancer and be used as medical assistants.

BACKGROUND

Breast cancer treatment can be very effective, especially when the disease is detected in the early stages. Feature selection and classification are common data mining techniques in machine learning that can provide breast cancer diagnosis with high speed, low cost and high precision.

METHODOLOGY

This paper proposes a new intelligent approach using an integrated filter-evolutionary search-based feature selection and an optimized ensemble classifier for breast cancer diagnosis. The selected features mainly relate to the viable solution as the selected features are successfully used in the breast cancer disease classification process. The proposed feature selection method selects the most informative features from the original feature set by integrating adaptive thresholder information gain-based feature selection and evolutionary gravity-search-based feature selection. Meanwhile, classification model is done by proposing a new intelligent multi-layer perceptron neural network-based ensemble classifier.

RESULTS

The simulation results show that the proposed method provides better performance compared to the state-of-the-art algorithms in terms of various criteria such as accuracy, sensitivity and specificity. Specifically, the proposed method achieves an average accuracy of 99.42% on WBCD, WDBC and WPBC datasets from Wisconsin database with only 56.7% of features.

CONCLUSION

Systems based on intelligent medical assistants configured with machine learning approaches are an important step toward helping doctors to detect breast cancer early.

Using artificial intelligence to predict adverse outcomes in emergency department patients with hyperglycemic crises in real time

BACKGROUND

Hyperglycemic crises are associated with high morbidity and mortality. Previous studies have proposed methods to predict adverse outcomes of patients in hyperglycemic crises; however, artificial intelligence (AI) has never been used to predict adverse outcomes. We implemented an AI model integrated with the hospital information system (HIS) to clarify whether AI could predict adverse outcomes.

METHODS

We included 2,666 patients with hyperglycemic crises from emergency departments (ED) between 2009 and 2018. The patients were randomized into a 70%/30% split for AI model training and testing. Twenty-two feature variables from the electronic medical records were collected. The performance of the multilayer perceptron (MLP), logistic regression, random forest, Light Gradient Boosting Machine (LightGBM), support vector machine (SVM), and K-nearest neighbor (KNN) algorithms was compared. We selected the best algorithm to construct an AI model to predict sepsis or septic shock, intensive care unit (ICU) admission, and all-cause mortality within 1 month. The outcomes between the non-AI and AI groups were compared after implementing the HIS and predicting the hyperglycemic crisis death (PHD) score.

RESULTS

The MLP had the best performance in predicting the three adverse outcomes, compared with the random forest, logistic regression, SVM, KNN, and LightGBM models. The areas under the curves (AUCs) using the MLP model were 0.852 for sepsis or septic shock, 0.743 for ICU admission, and 0.796 for all-cause mortality. Furthermore, we integrated the AI predictive model with the HIS to assist decision making in real time. No significant differences in ICU admission or all-cause mortality were detected between the non-AI and AI groups. The AI model performed better than the PHD score for predicting all-cause mortality (AUC 0.796 vs. 0.693).

CONCLUSIONS

A real-time AI predictive model is a promising method for predicting adverse outcomes in ED patients with hyperglycemic crises. Further studies recruiting more patients are warranted.

KSFinder-a knowledge graph model for link prediction of novel phosphorylated substrates of kinases

Background

Aberrant protein kinase regulation leading to abnormal substrate phosphorylation is associated with several human diseases. Despite the promise of therapies targeting kinases, many human kinases remain understudied. Most existing computational tools predicting phosphorylation cover less than 50% of known human kinases. They utilize local feature selection based on protein sequences, motifs, domains, structures, and/or functions, and do not consider the heterogeneous relationships of the proteins. In this work, we present KSFinder, a tool that predicts kinase-substrate links by capturing the inherent association of proteins in a network comprising 85% of the known human kinases. We also postulate the potential role of two understudied kinases based on their substrate predictions from KSFinder.

Methods

KSFinder learns the semantic relationships in a phosphoproteome knowledge graph using a knowledge graph embedding algorithm and represents the nodes in low-dimensional vectors. A multilayer perceptron (MLP) classifier is trained to discern kinase-substrate links using the embedded vectors. KSFinder uses a strategic negative generation approach that eliminates biases in entity representation and combines data from experimentally validated non-interacting protein pairs, proteins from different subcellular locations, and random sampling. We assess KSFinder's generalization capability on four different datasets and compare its performance with other state-of-the-art prediction models. We employ KSFinder to predict substrates of 68 "dark" kinases considered understudied by the Illuminating the Druggable Genome program and use our text-mining tool, RLIMS-P along with manual curation, to search for literature evidence for the predictions. In a case study, we performed functional enrichment analysis for two dark kinases - HIPK3 and CAMKK1 using their predicted substrates.

Results

KSFinder shows improved performance over other kinase-substrate prediction models and generalized prediction ability on different datasets. We identified literature evidence for 17 novel predictions involving an understudied kinase. All of these 17 predictions had a probability score ≥0.7 (nine at >0.9, six at 0.8-0.9, and two at 0.7-0.8). The evaluation of 93,593 negative predictions (probability ≤0.3) identified four false negatives. The top enriched biological processes of HIPK3 substrates relate to the regulation of extracellular matrix and epigenetic gene expression, while CAMKK1 substrates include lipid storage regulation and glucose homeostasis.

Conclusions

KSFinder outperforms the current kinase-substrate prediction tools with higher kinase coverage. The strategically developed negatives provide a superior generalization ability for KSFinder. We predicted substrates of 432 kinases, 68 of which are understudied, and hypothesized the potential functions of two dark kinases using their predicted substrates.

PredDRBP-MLP: Prediction of DNA-binding proteins and RNA-binding proteins by multilayer perceptron

Proteins interact with many molecules in order to maintain the vital activities in cells. Proteins that interact with DNA are called DNA-binding proteins (DBP), and proteins that interact with RNA are called RNA-binding proteins (RBP). Since DBPs and RBPs are involved in critical biological processes, their classification is quite important. Although the convolutional neural network and bidirectional long-short-term memory hybrid model (CNN-BiLSTM) is very popular in DBP and RBP classification, it has problems such as requirement of high processing power and long training time. Therefore, a multilayer perceptron (MLP) based predictor, PredDRBP-MLP (Predictor of DNA-Binding Proteins and RNA-Binding Proteins - Multilayer Perceptron) was developed in this study. PredDRBP-MLP is an artificial learning model that performs multi-class classification of DBPs, RBPs and non-nucleic acid-binding proteins (NNABP). PredDRBP-MLP achieved quite successful results on the independent dataset, specifically in the NNABP class, compared to the existing predictors, in addition to requiring lower processing power and being able to train quicker compared to CNN-BiLSTM based predictors. In NNABP class, PredDRBP-MLP predictor achieved 0.578 precision, 0.522 recall and 0.549 F1-score, while other multi-class predictor achieved 0.486 precision, 0.183 recall and 0.266 F1-score. A desktop application was developed for PredDRBP-MLP. The application is freely accessible at https://sourceforge.net/projects/preddrbp-mlp.

Automatic breast cancer diagnosis based on hybrid dimensionality reduction technique and ensemble classification

INTRODUCTION

Feature selection in the face of high-dimensional data can reduce overfitting and learning time, and at the same time improve the accuracy and efficiency of the system. Since there are many irrelevant and redundant features in breast cancer diagnosis, removing such features leads to more accurate prediction and reduced decision time when dealing with large-scale data. Meanwhile, ensemble classifiers are powerful techniques to improve the prediction performance of classification models, where several individual classifier models are combined to achieve higher accuracy.

METHODS

In this paper, an ensemble classifier algorithm based on multilayer perceptron neural network is proposed for the classification task, in which the parameters (e.g., number of hidden layers, number of neurons in each hidden layer, and weights of links) are adjusted based on an evolutionary approach. Meanwhile, this paper uses a hybrid dimensionality reduction technique based on principal component analysis and information gain to address this problem.

RESULTS

The effectiveness of the proposed algorithm was evaluated based on the Wisconsin breast cancer database. In particular, the proposed algorithm provides an average of 17% better accuracy compared to the best results obtained from the existing state-of-the-art methods.

CONCLUSION

Experimental results show that the proposed algorithm can be used as an intelligent medical assistant system for breast cancer diagnosis.

Modeling dissolved oxygen concentration using machine learning techniques with dimensionality reduction approach

Oxygen is crucial to keep the life cycle balance in any aspect. Aquatic life is highly influenced by the levels of dissolved oxygen (DO). This calls for not just constant monitoring of the DO in aquatic systems, but to generate an accurate prediction model for future levels of the DO. This study aims to propose an accurate prediction model for DO concentrations. The performance of the Random Forest (RF) and multilayer perceptron (MLP) algorithms was evaluated in generating the regression models. Moreover, the effect of dimensionality reduction of the data by the wrapper feature Selection method on the performance of the models was evaluated. The results showed that the RF regressor excelled MLP in performance with both the dataset of all variables and the dataset of reduced variables with the best performance achieved by the RF regressor by considering Pearson correlation coefficient (0.8052), Mean absolute error (0.8911), and root mean square error (1.2805) when trained by the dataset of reduced variables. As for the accuracy of the models, the estimation error deviation of both models declined significantly when trained by the reduced variables. When the accuracy of the prediction was increased by 0.95% by the RF regressor, the accuracy of the MLP was incremented by 5.7% when trained by the dataset of reduced variables. The results demonstrated the positive impact of the dimensionality reduction on the accuracy of both models. However, RF can be considered a robust regressor in predicting DO concentrations.

Dynamics and future prediction of LULC on Pare River basin of Arunachal Pradesh using machine learning techniques

Anthropogenic disturbances caused by increasing population densities are a significant concern as they accelerate climate change. Thus, regular monitoring of land use/land cover (LULC) is essential to mitigate these effects. Pare River basin of Arunachala Pradesh situated in the foothills of Eastern Himalayas was selected for this study. Landsat-5 TM and Landsat-8 OLI data from 2000 (T₁), 2015 (T₂), and 2020 (T₃) were used to prepare the LULC map. A support vector machine (SVM) classifier in the Google Earth Engine (GEE) environment was utilized for classification of LULC, while the TerrSet software environment was used for change analysis and projection using the CA-MC model. The SVM classifier produced overall all classification accuracies of 0.91, 0.85, and 0.91 with kappa values of 0.88, 0.82, and 0.89 for T₁, T₂, and T₃, respectively. The CA-MC model, which combines Markov chain and hybrid cellular automata, was calibrated with various predictor variables, including natural, proximity, and demographic variables along with T₁ and T₂ LULC and validated using T₃ LULC. The MLP was used for calibration, and an accuracy rate of above 0.70 was employed to generate transition potential maps (TPMs). The TPMs were used to project future LULC for 2030, 2040, and 2050. Validation analysis produced satisfactory results, with K_no, K_location, K_quality, and K_standard values of 0.96, 0.95, 0.95, and 0.93, respectively. Receiver operating characteristics (ROC) analysis showed an excellent area under the curve (AUC) value of 0.87. The findings of this study provide important insights to decision-makers and stakeholders in addressing the impacts of LULC changes.

Neural network algorithms predict new diffusion MRI data for multi-compartmental analysis of brain microstructure in a clinical setting

High angular resolution diffusion imaging (HARDI) is a promising method for advanced analysis of brain microstructure. However, comprehensive HARDI analysis requires multiple acquisitions of diffusion images (multi-shell HARDI), which is time consuming and often impractical in clinical settings. This study aimed to establish neural network models that can predict new diffusion datasets from clinically feasible brain diffusion MRI for multi-shell HARDI. The development included 2 algorithms: multi-layer perceptron (MLP) and convolutional neural network (CNN). Both followed a voxel-based approach for model training (70%), validation (15%), and testing (15%). The investigations involved 2 multi-shell HARDI datasets: 1) 11 healthy subjects from the Human Connectome Project (HCP); and 2) 10 local subjects with multiple sclerosis (MS). To assess outcomes, we conducted neurite orientation dispersion and density imaging using both predicted and original data and compared their orientation dispersion index (ODI) and neurite density index (NDI) in different brain tissues with 2 measures: peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM). Results showed that both models achieved robust predictions, which provided competitive ODI and NDI, especially in brain white matter. The CNN outperformed MLP with the HCP data on both PSNR (p < 0.001) and SSIM (p < 0.01). With the MS data, the models performed similarly. Overall, the optimized neural networks can help generate non-acquired brain diffusion MRI, which will make advanced HARDI analysis possible in clinical practice following further validation. Enabling detailed characterization of brain microstructure will allow enhanced understanding of brain function in both health and disease.

A method for real-time mechanical characterisation of microcapsules

Characterising the mechanical properties of flowing microcapsules is important from both fundamental and applied points of view. In the present study, we develop a novel multilayer perceptron (MLP)-based machine learning (ML) approach, for real-time simultaneous predictions of the membrane mechanical law type, shear and area-dilatation moduli of microcapsules, from their camera-recorded steady profiles in tube flow. By MLP, we mean a neural network where many perceptrons are organised into layers. A perceptron is a basic element that conducts input-output mapping operation. We test the performance of the present approach using both simulation and experimental data. We find that with a reasonably high prediction accuracy, our method can reach an unprecedented low prediction latency of less than 1 millisecond on a personal computer. That is the overall computational time, without using parallel computing, from a single experimental image to multiple capsule mechanical parameters. It is faster than a recently proposed convolutional neural network-based approach by two orders of magnitude, for it only deals with the one-dimensional capsule boundary instead of the entire two-dimensional capsule image. Our new approach may serve as the foundation of a promising tool for real-time mechanical characterisation and online active sorting of deformable microcapsules and biological cells in microfluidic devices.

Application of regression and artificial neural network analysis of Red-Green-Blue image components in prediction of chlorophyll content in microalgae

This study presented a novel methodology to predict microalgae chlorophyll content from colour models using linear regression and artificial neural network. The analysis was performed using SPSS software. Type of extractant solvents and image indexes were used as the input data for the artificial neural network calculation. The findings revealed that the regression model was highly significant, with high R² of 0.58 and RSME of 3.16, making it a useful tool for predicting the chlorophyll concentration. Simultaneously, artificial neural network model with R² of 0.66 and low RMSE of 2.36 proved to be more accurate than regression model. The model which fitted to the experimental data indicated that acetone was a suitable extraction solvent. In comparison to the cyan-magenta-yellow-black model in image analysis, the red-greenblue model offered a better correlation. In short, the estimation of chlorophyll concentration using prediction models are rapid, more efficient, and less expensive.

Machine learning-guided prediction of potential engineering targets for microbial production of lycopene

The process of designing streamlined workflows for developing microbial strains using classical methods from vast amounts of biological big data has reached its limits. With the continuous increase in the amount of biological big data, data-driven machine learning approaches are being used to overcome the limits of classical approaches for strain development. Here, machine learning-guided engineering of Deinococcus radiodurans R1 for high-yield production of lycopene was demonstrated. The multilayer perceptron models were first trained using the mRNA expression levels of the key genes along with lycopene titers and yields obtained from 17 strains. Then, the potential overexpression targets from 2,047 possible combinations were predicted by the multilayer perceptron combined with a genetic algorithm. Through the machine learning-aided fine-tuning of the predicted genes, the final-engineered LY04 strain resulted in an 8-fold increase in the lycopene production, up to 1.25 g/L from glycerol, and a 6-fold increase in the lycopene yield.

An improved butterfly optimization algorithm for training the feed-forward artificial neural networks

Artificial neural network (ANN) which is an information processing technique developed by modeling the nervous system of the human brain is one of the most powerful learning methods today. One of the factors that make ANN successful is its training algorithm. In this paper, an improved butterfly optimization algorithm (IBOA) based on the butterfly optimization algorithm was proposed for training the feed-forward artificial neural networks. The IBOA algorithm has the chaotic property which helps optimization algorithms to explore the search space more dynamically and globally. In the experiments, ten chaotic maps were used. The success of the IBOA algorithm was tested on 13 benchmark functions which are well known to those working on global optimization and are frequently used for testing and analysis of optimization algorithms. The Tent-mapped IBOA algorithm outperformed the other algorithms in most of the benchmark functions. Moreover, the success of the IBOA-MLP algorithm also has been tested on five classification datasets (xor, balloon, iris, breast cancer, and heart) and the IBOA-MLP algorithm was compared with four algorithms in the literature. According to the statistical performance metrics (sensitivity, specificity, precision, F1-score, and Friedman test), the IBOA-MLP outperformed the other algorithms and proved to be successful in training the feed-forward artificial neural networks.

Dengue outbreaks prediction in Bangladesh perspective using distinct multilayer perceptron NN and decision tree

Dengue fever is a disease that has been outbreak worldwide in the last few years. Dengue is a fatal disease; sometimes, it may cause life-threatening complications and even death. Dengue is considered to be one of the critical diseases which is spreading in more than 110 countries. Nearly 45,000 case reports have been found around Bangladesh in the last year. Dengue fever has become a major health hazard in Bangladesh. Hence, early detection would mitigate major casualties of Dengue disease. Distinct studies have been performed concerning Dengue disease; however, no effective study, particularly from Bangladesh's perspective, it seemed that reveals Dengue outbreaks prediction method. In this scenario, this research work aims to analyse the Dengue disease and build an apposite model to predict dengue outbreaks. This paper also aims to find the best technique to early predicts Dengue disease. The real-time data of the patients admitted to different hospitals in Bangladesh is accumulated to achieve the goal of the current research. Then different multilayer perceptron neural networks and a Decision tree are used for Dengue outbreaks prediction. Twenty-five parameters are analysed to find these parameters' infection rates in this work. A comparative study of the developed models' performances is also accomplished to obtain a better Dengue outbreaks prediction model. The results evidence that the Levenberg-Marquardt is the best technique with 97.3% accuracy and 2.7% error in Dengue disease prediction. On the other hand, the Decision tree may have the second choice to assess Dengue disease.

Development and validation of artificial intelligence models for preoperative prediction of inferior mesenteric artery lymph nodes metastasis in left colon and rectal cancer

BACKGROUND

Dissection of lymph nodes at the roots of the inferior mesenteric artery (IMAN) should be offered only to selected patients at a major risk of developing IMAN involvement. The aim of this study is to present the first artificial intelligence (AI) models to predict IMAN metastasis risk in the left colon and rectal cancer patients.

METHODS

A total of 2891 patients with descending colon including splenic flexure, sigmoid colon and rectal cancer undergoing major primary surgery and IMAN dissection were included as a study cohort, which was then split into a training set (67%) and a testing set (33%). Feature selection was conducted using the least absolute shrinkage and selection operator (LASSO) regression model. Seven AI algorithms, namely Support Vector Machine (SVM), Logistic Regression (LR), Extreme Gradient Boosting (XGB), Light Gradient Boosting (LGB), Decision Tree Classifier (DTC), Random Forest (RF) classifier, and Multilayer Perceptron (MLP), as well as traditional multivariate LR model were employed to construct predictive models. The optimal hyperparameters were determined with 5 fold cross-validation. The predictive performance of models and the expert surgeon was assessed and compared in the testing set independently.

RESULTS

The IMAN involvement incidence was 4.6%. The optimal set of features selected by LASSO included 10 characteristics: neoadjuvant treatment, age, synchronous liver metastasis, synchronous lung metastasis, signet ring adenocarcinoma, neural invasion, lymphovascular invasion, CA199, endoscopic obstruction, T stage evaluated by MRI. The most accurate model derived from MLP showed excellent prediction power with area under the receiver operating characteristic curve (AUROC) of 0.873 and produced 81.0% recognition sensitivity and 82.5% specificity in the testing set independently. In contrast, the judgment of IMAN metastasis by expert surgeon yield rather imprecise and unreliable results with a significantly lower AUROC of 0.509. Additionally, the proposed MLP had the highest net benefits and the largest reduction of unnecessary IMAN dissection without the cost of additional involved IMAN missed.

CONCLUSION

MLP model was able to maintain its prediction accuracy in the testing set better than other models and expert surgeons. Our MLP model could be used to help identify IMA nodal metastasis and to select candidates for individual IMAN dissection.

Design principles and mechanistic explanation

In this essay I propose that what design principles in systems biology and systems neuroscience do is to present abstract characterizations of mechanisms, and thereby facilitate mechanistic explanation. To show this, one design principle in systems neuroscience, i.e., the multilayer perceptron, is examined. However, Braillard (2010) contends that design principles provide a sort of non-mechanistic explanation due to two related reasons: they are very general and describe non-causal dependence relationships. In response to this, I argue that, on the one hand, all mechanisms are more or less general (or abstract), and on the other, many (if not all) design principles are causal systems.

Driving effects and transfer prediction of heavy metal(loid)s in contaminated courtyard gardens using redundancy analysis and multilayer perceptron

The distribution and migration of heavy metal(loid)s in the soil-vegetable systems of courtyard gardens near mining areas have rarely been investigated, leading to potential food safety risks for residents. Moreover, the existing research is mainly focused on the total content of heavy metal(loid)s (tMetals) rather than the bioavailable contents (aMetals). In this study, 26 and 28 pairs of soil and vegetable samples were collected from the courtyard gardens near the Realgar mine in Baiyun Town and the lead-zinc (Pb-Zn) mine in Shuikoushan Town, respectively. The tMetal and aMetal of cadmium (Cd), mercury (Hg), arsenic (As), Pb, chromium (Cr), nickel (Ni), copper (Cu), Zn, manganese (Mn), iron (Fe), and calcium (Ca) in the samples were analyzed in this study. The results showed that courtyard gardens were polluted by various heavy metal(loid)s at varying degrees. The bioavailabilities of different metals varied significantly, among which Cd has the highest bioavailability (> 30%). In the transfer process of heavy metal(loid)s, the transfer rate (Tf) was ranked as soil-roots (1.50) > stems-leaves (1.07) > roots-stems (0.46) > stems-fruits (0.33). Redundancy analysis was used to evaluate the driving effects, and the results revealed that aCa, aZn, and aFe in soil could inhibit the absorption of aCd by plant roots. Soil organic matter was the inhibiting factor regarding the transfer of aAs and aCu, whereas it was also the promoting factor for transferring aPb, aNi, and aCr. Furthermore, the multilayer perceptron (MLP) could effectively predict the Tf of heavy metal(loid)s based on the aMetal. The R² values of the MLP were ranked as follows: 0.91 for As, 0.88 for Zn, 0.85 for Hg, 0.83 for Cu, 0.79 for Cr, 0.66 for Cd, 0.65 for Pb, and 0.52 for Ni. This study emphasizes the aMetal-based ecological characteristics and prediction ability. The study results are significant for guiding residents to strategize appropriate crop planting and ensure the safe production and consumption of vegetables.

DGMP: Identifying Cancer Driver Genes by Jointing DGCN and MLP from Multi-omics Genomic Data

Identification of cancer driver genes plays an important role in precision oncology research, which is helpful to understand cancer initiation and progression. However, most existing computational methods mainly used the protein-protein interaction (PPI) networks, or treated the directed gene regulatory networks (GRNs) as the undirected gene-gene association networks to identify the cancer driver genes, which will lose the unique structure regulatory information in the directed GRNs, and then affect the outcome of the cancer driver gene identification. Here, based on the multi-omics pan-cancer data (i.e., gene expression, mutation, copy number variation, and DNA methylation), we propose a novel method (called DGMP) to identify cancer driver genes by jointing directed graph convolutional network (DGCN) and multilayer perceptron (MLP). DGMP learns the multi-omics features of genes as well as the topological structure features in GRN with the DGCN model and uses MLP to weigh more on gene features for mitigating the bias toward the graph topological features in the DGCN learning process. The results on three GRNs show that DGMP outperforms other existing state-of-the-art methods. The ablation experimental results on the DawnNet network indicate that introducing MLP into DGCN can offset the performance degradation of DGCN, and jointing MLP and DGCN can effectively improve the performance of identifying cancer driver genes. DGMP can identify not only the highly mutated cancer driver genes but also the driver genes harboring other kinds of alterations (e.g., differential expression and aberrant DNA methylation) or genes involved in GRNs with other cancer genes. The source code of DGMP can be freely downloaded from https://github.com/NWPU-903PR/DGMP.

Prediction and verification of the effect of psoriasis on coronary heart disease based on artificial neural network

Background and objectives

Psoriasis is an independent risk factor for coronary heart disease. It is important for predicting the complications of coronary heart disease in patients with psoriasis.

Methods

In this study, related cases were collected from the case system of Qingdao University, and commonly used laboratory indicators were extracted. Artificial neural network (ANN) and logistics regression analysis were used to learn to distinguish psoriasis patients, coronary heart disease patients, and psoriasis patients with coronary heart disease. We identified independent risk factors for coronary heart disease in psoriasis patients that exacerbate coronary heart disease symptoms in patients with psoriasis.

Findings

Analysis shows that the accuracy of the ANN model was higher than 79%. It was determined that age, chlorinated, phosphorus, magnesium, low-density lipoprotein, triglycerides, high density lipoprotein and total cholesterol are independent risk factors for coronary heart disease in patients with psoriasis. Similarly, gender, age, chlorinated, magnesium, triglycerides, and high density lipoprotein are risk factors that exacerbate coronary heart disease symptoms in patients with psoriasis.

Interpretation

The presented approach is a valuable tool for identifying psoriasis patients, coronary heart disease patients, and psoriasis patients with coronary heart disease. It can also serve as a support tool clinicians in the diagnostic process, by providing an outstanding support in the diagnostics prevention of coronary heart disease in psoriasis.

Building Models of Functional Interactions Among Brain Domains that Encode Varying Information Complexity: A Schizophrenia Case Study

Revealing associations among various structural and functional patterns of the brain can yield highly informative results about the healthy and disordered brain. Studies using neuroimaging data have more recently begun to utilize the information within as well as across various functional and anatomical domains (i.e., groups of brain networks). However, most whole-brain approaches assume similar complexity of interactions throughout the brain. Here we investigate the hypothesis that interactions between brain networks capture varying amounts of complexity, and that we can better capture this information by varying the complexity of the model subspace structure based on available training data. To do this, we employ a Bayesian optimization-based framework known as the Tree Parzen Estimator (TPE) to identify, exploit and analyze patterns of variation in the information encoded by temporal information extracted from functional magnetic resonance imaging (fMRI) subdomains of the brain. Using a repeated cross-validation procedure on a schizophrenia classification task, we demonstrate evidence that interactions between specific functional subdomains are better characterized by more sophisticated model architectures compared to less complicated ones required by the others for optimally contributing towards classification and understanding the brain's functional interactions. We show that functional subdomains known to be involved in schizophrenia require more complex architectures to optimally unravel discriminatory information about the disorder. Our study points to the need for adaptive, hierarchical learning frameworks that cater differently to the features from different subdomains, not only for a better prediction but also for enabling the identification of features predicting the outcome of interest.

CM-SegNet: A deep learning-based automatic segmentation approach for medical images by combining convolution and multilayer perceptron

Accurate segmentation of lesions in medical images is of great significance for clinical diagnosis and evaluation. The low contrast between lesions and surrounding tissues increases the difficulty of automatic segmentation, while the efficiency of manual segmentation is low. In order to increase the generalization performance of segmentation model, we proposed a deep learning-based automatic segmentation model called CM-SegNet for segmenting medical images of different modalities. It was designed according to the multiscale input and encoding-decoding thoughts, and composed of multilayer perceptron and convolution modules. This model achieved communication of different channels and different spatial locations of each patch, and considers the edge related feature information between adjacent patches. Thus, it could fully extract global and local image information for the segmentation task. Meanwhile, this model met the effective segmentation of different structural lesion regions in different slices of three-dimensional medical images. In this experiment, the proposed CM-SegNet was trained, validated, and tested using six medical image datasets of different modalities and 5-fold cross validation method. The results showed that the CM-SegNet model had better segmentation performance and shorter training time for different medical images than the previous methods, suggesting it is faster and more accurate in automatic segmentation and has great potential application in clinic.

A multi-layer perceptron-based approach for early detection of BSR disease in oil palm trees using hyperspectral images

Basal Stem Rot (BSR) disease caused by Ganoderma boninense is identified as the biggest threat in oil palm industry in Malaysia, resulting in significant yield losses. Effective BSR disease detection is important for plantation management to ensure stable palm oil production. Existing method is done by experience personnel, via visual inspection it is very time consuming. Rapid development of unmanned aerial vehicle (UAV) and machine learning has the potential to address this issue with higher efficiency. This paper proposed a new framework to automate BSR disease detection with UAV images to improve time efficiency and automate detection process. The proposed method has two steps, first hyperspectral image (HSI) pre-processing, followed by artificial neural network disease detection. Multilayer-Perceptron model is introduced to learn spectral features from different infection stages. The model is trained with ground truth collected by trained surveyors. The HSI sample size consists of 2 healthy trees, 5 Stage A (mild infection), 5 Stage B (moderate infection), and 3 Stage C (severe infection). Performance is examined with support vector machine (SVM), 1 dimensional convolutional network (1D CNN), and several vegetation indices, namely Normalized Difference Vegetation Index (NDVI), Normalized Difference Red Edge (NDRE), Optimised Soil-Adjusted Vegetation Index (OSAVI), and Merris Terrestrial Chlorophyll Index (MTCI). All machine learning algorithms can segregate infection stages, MLP modal had a highest overall accuracy 86.67%, compared to SVM and 1D CNN at 66.67% and 73.33%. Whereas for vegetation index, it can only detect Stage C tree, and not able to differentiate between Healthy, Stage A and Stage B tree. In term of computational cost, MLP modal had balance performance with moderate training time, but faster inference time. It demonstrates effectiveness on BSR disease detection, even at early infection stage.

Modelling Covid-19 infections in Zambia using data mining techniques

The outbreak of Covid-19 pandemic has been declared a global health crisis by the World Health Organization since its emergence. Several researchers have proposed a number of techniques to understand how the pandemic affects the populations. Reported among these techniques are data mining models which have been successfully applied in a wide range of situations before the advent of Covid-19 pandemic. In this work, the researchers have applied a number of existing data mining methods (classifiers) available in the Waikato Environment for Knowledge Analysis (WEKA) machine learning library. WEKA was used to gain a better understanding on how the epidemic spread within Zambia. The classifiers used are J48 decision tree, Multilayer Perceptron and Naïve Bayes among others. The predictions of these techniques are compared against simpler classifiers and those reported in related works.

COVID and nutrition: A machine learning perspective

A self-report questionnaire survey was conducted online to collect big data from over 16000 Iranian families (who were the residents of 1000 urban and rural areas of Iran). The resulting data storage contained over 1 M records of data and over 1G records of automatically inferred information. Based on this data storage, a series of machine learning experiments was conducted to investigate the relationship between nutrition and the risk of contracting COVID-19. With highly accurate scores, the findings strongly suggest that foods and water sources containing certain natural bioactive and phytochemical agents may help to reduce the risk of apparent COVID-19 infection.

Estimation of metal elements content in soil using x-ray fluorescence based on multilayer perceptron

X-ray fluorescence (XRF) is widely used to rapidly detect heavy metals in soil. Spectra processing has been an important research topic to improve accuracy. In this study, 80 soil samples were analyzed by XRF under indoor conditions, where different preprocessing and quantitative analysis methods were compared in terms of prediction accuracy. Denoising algorithms were used to preprocess the soil spectra before establishing prediction models for As, Pb, Cu, Cr, and Cd in soil. The influence of denoising methods on the prediction effects of different models was compared and discussed. The results on five heavy metal spectra show that the proper spectral preprocessing method can effectively improve the prediction performance of the model. The multilayer perceptron model provides promising analysis and modeling for the five metal elements. The determination coefficients (R²) of the models were 0.857, 0.976, 0.977, 0.995, and 0.886, respectively. The proposed method provides the potential to support accurate quantitation by XRF analysis.

PTML modeling for peptide discovery: in silico design of non-hemolytic peptides with antihypertensive activity

Hypertension is a medical condition that affects millions of people worldwide. Despite the high efficacy of the current antihypertensive drugs, they are associated with serious side effects. Peptides constitute attractive options for chemical therapy against hypertension, and computational models can accelerate the design of antihypertensive peptides. Yet, to the best of our knowledge, all the in silico models predict only the antihypertensive activity of peptides while neglecting their inherent toxic potential to red blood cells. In this work, we report the first sequence-based model that combines perturbation theory and machine learning through multilayer perceptron networks (SB-PTML-MLP) to enable the simultaneous screening of antihypertensive activity and hemotoxicity of peptides. We have interpreted the molecular descriptors present in the model from a physicochemical and structural point of view. By strictly following such interpretations as guidelines, we performed two tasks. First, we selected amino acids with favorable contributions to both the increase of the antihypertensive activity and the diminution of hemotoxicity. Then, we assembled those suitable amino acids, virtually designing peptides that were predicted by the SB-PTML-MLP model as antihypertensive agents exhibiting low hemotoxicity. The potentiality of the SB-PTML-MLP model as a tool for designing potent and safe antihypertensive peptides was confirmed by predictions performed by online computational tools reported in the scientific literature. The methodology presented here can be extended to other pharmacological applications of peptides.

An Improved Animal Migration Optimization Algorithm to Train the Feed-Forward Artificial Neural Networks

The most important and demanding part of the artificial neural network is the training process which involves finding the most suitable values for the weights in the network architecture, a challenging optimization problem. Gradient approaches and the meta-heuristic approaches are two methods extensively used to optimize the weights in the network. Gradient approaches have serious disadvantages including getting stuck in local optima, inadequate exploration, etc. To overcome these disadvantages, meta-heuristic approaches are preferred in training the artificial neural network instead of gradient methods. Therefore, in this study, an improved animal migration optimization algorithm with the Lévy flight feature was proposed to train the multilayer perceptron. The proposed hybrid algorithm is named IAMO-MLP. The main contributions of this article are that the IAMO algorithm was developed, the IAMO-MLP algorithm can successfully escape from local optima, and the initial positions did not affect the performance of the IAMO-MLP algorithm. The enhanced algorithm was tested and validated against a wider set of benchmark functions and indicated that it substantially outperformed the original implementation. Afterward, the IAMO-MLP was compared with ten algorithms on five classification problems (xor, balloon, iris, breast cancer, and heart) and one real-world problem in terms of mean squared error, classification accuracy, and nonparametric statistical Friedman test. According to the results, the IAMO was successful in training the multilayer perceptron.

Effect of climatology parameters on air pollution during COVID-19 pandemic in Jordan

This study aims to explore the real-time impact of the COVID-19 pandemic on measured air pollution in the three largest cities of Jordan (Amman, Irbid and Zarqa). It is hypothesized that a sharp decrease in the emitted amounts of particulate matter (PM₁₀), CO, NO₂ and SO₂ during COVID-19 pandemic will be obtained, this corresponds with the reduced traffic due to mandated business closures. To achieve this exploration a paired sample t-test is used to compare the concentration of these four pollutants in the three cities over the period from 15 March to 30 June during the years from 2016 to 2020. It is found that there is a significant difference between the emitted concentrations mean values of CO, PM₁₀, SO₂ and NO₂ during the period of study. This was indicated by the values of p for each species, which was less than 5 % for all these pollutants. The maximum reduction in SO2 and NO2 concentration during the lockdown period was in Zarqa. Irbid city witnessed the highest percentage reduction in CO and PM10. Furthermore, the correlation test, independent variable importance of multilayer perceptron and global sensitivity analysis using Sobol analysis showed that metrological data (Humidity, wind speed, average temperature and pressure) have a direct relationship with concentrations of CO, PM10, SO2 and NO2 in Amman, Irbid and Zarqa before and after COVID-19 pandemic.

Multilayer perceptron neural network model development for mechanical ventilator parameters prediction by real time system learning

Background and objective

In pandemic situations like COVID 19, real time monitoring of patient condition and continuous delivery of inspired oxygen can be made possible only through artificial intelligence-based system modeling. Even now manual control of mechanical ventilator parameters is continuing despite the ever-increasing number of patients in critical epidemic conditions. Here a suggestive multi-layer perceptron neural network model is developed to predict the level of inspired oxygen delivered by the mechanical ventilator along with mode and positive end expiratory pressure (PEEP) changes for reducing the effort of health care professionals.

Methods

The artificial neural network model is developed by Python programming using real time data. Parameter identification for model inputs and outputs is done by in corporating consistent real time patient data including periodical arterial blood gas analysis, continuous pulse oximetry readings and mechanical ventilator settings using statistical pairwise analysis using R programming.

Results

Mean square error values and R values of the model are calculated and found to be an average of 0.093 and 0.81 respectively for various data sets. Accuracy loss will be in good fit with validation loss for a comparable number of epochs.

Conclusions

Comparison of the model output is undertaken with physician’s prediction using statistical analysis and shows an accuracy error of 4.11 percentages which is permissible for a good predictive system.

A novel stacking technique for prediction of diabetes

BACKGROUND

Machine Learning (ML) represents a rapidly growing technology that supplies the most effective solutions for solving complex problems. The application of ML techniques in healthcare is gaining more attention because of ML-associated automatic pattern identification mechanisms. Diabetes is characterized by hyperglycemia resulting from improper insulin secretion and/or insulin utilization.

METHODS

The PIMA Indian diabetes dataset was obtained from the University of California/Irvine (UCI) machine learning repository for experimental purposes. The study was carried out in three stages: (1) a correlation technique was developed for feature selection; (2) the AdaBoost technique was implemented on selected features for classification; and (3) a novel stacking technique with multi-layer perceptron, support vector machine, and logistic regression (MLP, SVM, and LR, respectively) was designed and developed for the selected features.

RESULTS

The proposed stacking technique integrated the intelligent models and led to an improvement in model performance, thereby overcoming the issue of generating multiple decision stumps by AdaBoost. The proposed novel stacking technique outperformed other models when compared with AdaBoost in terms of performance metrics. The proposed models were then implemented on other datasets, such as the Cleveland heart disease and Wisconsin breast cancer diagnostic datasets, to illustrate their broader applications.

CONCLUSION

Stacking can outperform other models when compared with the other reported techniques that were implemented using the PIMA Indian diabetes dataset.

A multilayer perceptron neural network approach for the solution of hyperbolic telegraph equations

Neural networks have been extensively used for solving differential equations in the past, but they rely mostly on computationally expensive gradient-based numerical optimization procedure for solving differential equations. In this work, we are introducing a faster way to train neural networks for solving differential equations based on extreme learning machine algorithm. This algorithm is much faster as compared to traditional approaches, and it also provides highly accurate results. Reliability of the approach is tested by solving various cases of the hyperbolic telegraph equations. Solutions so obtained are compared to the results existing in the literature for analysing the accuracy of the proposed approach.

Spatial and temporal analysis of human infection with the avian influenza A (H7N9) virus in China and research on a risk assessment agent-based model

OBJECTIVES

From 2013 to 2017, the avian influenza A (H7N9) virus frequently infected people in China, which seriously affected the public health of society. This study aimed to analyze the spatial characteristics of human infection with the H7N9 virus in China and assess the risk areas of the epidemic.

METHODS

Using kernel density estimation, standard deviation ellipse analysis, spatial and temporal scanning cluster analysis, and Pearson correlation analysis, the spatial characteristics and possible risk factors of the epidemic were studied. Meteorological factors, time (month), and environmental factors were combined to establish an epidemic risk assessment proxy model to assess the risk range of an epidemic.

RESULTS

The epidemic situation was significantly correlated with atmospheric pressure, temperature, and daily precipitation (P < 0.05), and there were six temporal and spatial clusters. The fitting accuracy of the epidemic risk assessment agent-based model for lower-risk, low-risk, medium-risk, and high-risk was 0.795, 0.672, 0.853, 0.825, respectively.

CONCLUSIONS

This H7N9 epidemic was found to have more outbreaks in winter and spring. It gradually spread to the inland areas of China. This model reflects the risk areas of human infection with the H7N9 virus.

Empirical strategy for stretching probability distribution in neural-network-based regression

In regression analysis under artificial neural networks, the prediction performance depends on determining the appropriate weights between layers. As randomly initialized weights are updated during back-propagation using the gradient descent procedure under a given loss function, the loss function structure can affect the performance significantly. In this study, we considered the distribution error, i.e., the inconsistency of two distributions (those of the predicted values and label), as the prediction error, and proposed weighted empirical stretching (WES) as a novel loss function to increase the overlap area of the two distributions. The function depends on the distribution of a given label, thus, it is applicable to any distribution shape. Moreover, it contains a scaling hyperparameter (β) such that the appropriate parameter value maximizes the common section of the two distributions. To test the function capability, we generated ideal distributed curves (unimodal, skewed unimodal, bimodal, and skewed bimodal) as the labels, and used the Fourier-extracted input data from the curves under a feedforward neural network. In general, WES outperformed loss functions in wide use, and the performance was robust to the various noise levels. The improved results in RMSE for the extreme domain (i.e., both tail regions of the distribution) are expected to be utilized for prediction of abnormal events in non-linear complex systems such as natural disaster and financial crisis.