Viral Drug prediction with Matrix Decomposition and Two-layer Heterogeneous Graph Inference

Recently, the association prediction between viruses and drugs has drawn more and more attention. A growing number of studies have shown that the problem of antiviral drug resistance is increasing and has become a major problem plaguing the medical community. Moreover, the development cycle of new drugs is long and requires a lot of funds. If new viruses emerge, effective antiviral drugs are urgently needed. Therefore, effective calculation methods are required to predict potential antiviral drugs. In this paper, we developed a computational model of Matrix Decomposition and Heterogeneous Graph based Inference for Drug-Virus Association (MDHGIVDA) to predict potential drug-virus associations. MDHGIVDA integrated virus sequence similarity, drug chemical structure similarity, drug side effect similarity, Gaussian interaction profile kernel similarity for drugs and viruses, new drug-virus associations matrix obtained by matrix decomposition to discover new drug-virus associations. Due to the use of matrix factorization and heterogeneous graphs, our model has a high prediction accuracy compared with the previous four models. In the global and local leave-one-out cross validation (LOOCV), MDHGIVDA obtained area under the receiver operating characteristics curve (AUC) of 0.8528 and AUC of 0.8532, respectively. In addition, in the five-fold cross validation, the AUC and the standard deviation is 0.8299 0.0037, which shows that MDHGIVDA has stability and high prediction accuracy. In the case studies of three important viruses, 18, 14, and 16 out of the top 20 predicted drugs for Zika virus (ZIKV), Severe Acute Respiratory Syndrome Coronavirus 2 ( SARS-COV-2 ), Human Immunodeficiency Virus-1 (HIV-1) were verified respectively by searching the literature on PubMed. These results showed that MDHGIVDA is effective in predicting potential drug-virus associations.

Profiling the baseline performance and limits of machine learning models for adaptive immune receptor repertoire classification

Background: Machine learning (ML) methodology development for classification of immune states in adaptive immune receptor repertoires (AIRR) has seen a recent surge of interest. However, so far, there does not exist a systematic evaluation of scenarios where classical ML methods (such as penalized logistic regression) already perform adequately for AIRR classification. This hinders investigative reorientation to those scenarios where further method development of more sophisticated ML approaches may be required. Results: To identify those scenarios where a baseline method is able to perform well for AIRR classification, we generated a collection of synthetic benchmark datasets encompassing a wide range of dataset architecture-associated and immune state-associated sequence pattern (signal) complexity. We trained ≈1300 ML models with varying assumptions regarding immune signal on ≈850 datasets with a total of ≈210000 repertoires containing ≈42 billion TCRβ CDR3 amino acid sequences, thereby surpassing the sample sizes of current state-of-the-art AIRR ML setups by two orders of magnitude. We found that L1-penalized logistic regression achieved high prediction accuracy even when the immune signal occurs only in 1 out of 50000 AIR sequences. Conclusions: We provide a reference benchmark to guide new AIRR ML classification methodology by: (i) identifying those scenarios characterised by immune signal and dataset complexity, where baseline methods already achieve high prediction accuracy and (ii) facilitating realistic expectations of the performance of AIRR ML models given training dataset properties and assumptions. Our study serves as a template for defining specialized AIRR benchmark datasets for comprehensive benchmarking of AIRR ML methods.

Research on Sports Performance Prediction Based on BP Neural Network

Artificial neural network has the advantages of self-training and fault tolerance, while BP neural network has simple learning algorithms and powerful learning capabilities. The BP neural network algorithm has been widely used in practice. This paper conducts research on sports performance prediction based on 5G and artificial neural network algorithms. This paper uses the BP neural network algorithm as a pretest modelling method to predict the results of the 30th Olympic Men’s 100m Track and Field Championships and is supported by the MATLAB neural network toolbox. According to the experimental results, the scheme proposed in this paper has better performance than the other prediction strategies. In order to explore the feasibility and application of the BP neural network in this kind of prediction, there is a lot of work to be done. The model has a high prediction accuracy and provides a new method for the prediction of sports performance. The results show that the BP neural network algorithm can be used to predict sports performance, with high prediction accuracy and strong generalization ability.

A DNA tetrahedron-based molecular computation device for the logic sensing of dual microRNAs in living cells

Endogenous miRNA expression patterns are specific to cell type and thus offer high prediction accuracy with regard to different cell identities compared to single miRNA analysis.

Modelling postural discomfort perception using CHAID decision tree algorithm

Developing quantitative models that predict discomfort levels of working postures has been an important ergonomics research topic. Such modeling not only has practical applications, but also may serve as a useful research method to improve our understanding of the human postural discomfort perception process. While the existing models have focused on achieving high prediction accuracy, less attention has been given to model interpretability, which is vital for understanding a process through modeling. Research is needed to identify the model types or modeling methods that offer high interpretability as well as good prediction accuracy. The goal of this study was to evaluate the utility of the Chi-square Automatic Interaction Detector (CHAID) decision tree modeling method in developing postural discomfort prediction models. Ten individual-specific decision tree models were developed, which predicted overall upper-body discomfort from local body part discomfort ratings. The prediction models were found to achieve high prediction accuracy and interpretability. (150 words)

Using machine learning to discover shape descriptors for predicting emulsion stability in a microfluidic channel

A novel shape descriptor identified by machine learning captures diverse droplet shapes and achieves high prediction accuracy of droplet instability.

Analysis of Pipe Settlement Based on Numerical Simulation

Settlement of comprehensive pipe is inevitable disturbed by construction and surrounding environment, in order to better understand the settlement rule of comprehensive pipe. Based on a large amount of monitoring data collected from integrated trench, a simulation model is proposed to analyzed comprehensive pipe settlement by using ABAQUS, the case results show that the simulation model performance can analysis the synthesis trench settlement. It has high prediction accuracy and fit for the settlement of integral comprehensive pipe.