Current Prediction Algorithms Research Articles

The Alzheimer's Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge: Results after 1 Year Follow-up

Accurate prediction of progression in subjects at risk of Alzheimer's disease is crucial for enrolling the right subjects in clinical trials. However, a prospective comparison of state-of-the-art algorithms for predicting disease onset and progression is currently lacking. We present the findings of "The Alzheimer's Disease Prediction Of Longitudinal Evolution" (TADPOLE) Challenge, which compared the performance of 92 algorithms from 33 international teams at predicting the future trajectory of 219 individuals at risk of Alzheimer's disease. Challenge participants were required to make a prediction, for each month of a 5-year future time period, of three key outcomes: clinical diagnosis, Alzheimer's Disease Assessment Scale Cognitive Subdomain (ADAS-Cog13), and total volume of the ventricles. The methods used by challenge participants included multivariate linear regression, machine learning methods such as support vector machines and deep neural networks, as well as disease progression models. No single submission was best at predicting all three outcomes. For clinical diagnosis and ventricle volume prediction, the best algorithms strongly outperform simple baselines in predictive ability. However, for ADAS-Cog13 no single submitted prediction method was significantly better than random guesswork. Two ensemble methods based on taking the mean and median over all predictions, obtained top scores on almost all tasks. Better than average performance at diagnosis prediction was generally associated with the additional inclusion of features from cerebrospinal fluid (CSF) samples and diffusion tensor imaging (DTI). On the other hand, better performance at ventricle volume prediction was associated with inclusion of summary statistics, such as the slope or maxima/minima of patient-specific biomarkers. On a limited, cross-sectional subset of the data emulating clinical trials, performance of the best algorithms at predicting clinical diagnosis decreased only slightly (2 percentage points) compared to the full longitudinal dataset. The submission system remains open via the website https://tadpole.grand-challenge.org, while TADPOLE SHARE (https://tadpole-share.github.io/) collates code for submissions. TADPOLE's unique results suggest that current prediction algorithms provide sufficient accuracy to exploit biomarkers related to clinical diagnosis and ventricle volume, for cohort refinement in clinical trials for Alzheimer's disease. However, results call into question the usage of cognitive test scores for patient selection and as a primary endpoint in clinical trials.

Healthcare Data-Based Prediction Algorithm for Potential Knee Joint Injury of Football Players.

It is important to predict the potential harm to the knee joint in order to prevent football players from inflicting numerous injuries to the knee during activity. Numerous professionals have been drawn to this subject, and many viable prediction systems have been developed. Prediction of potential knee joint injury is critical to effectively avoid knee joint injury during exercise. The current prediction algorithms are mainly implemented through expert interviews, medical reports, and historical documents. The algorithms have problems with low prediction accuracy or precision values. There is a need to understand more knee injury factors and improve the prediction accuracy; hence, the intelligent prediction algorithm for potential injury of knee joints of football players is proposed in this paper. Firstly, the characteristics of the knee joint injury and the injury factors of the football players are gathered and analyzed. Then, the damage is predicted by the similarity measurement. The experimental results show that the proposed algorithm has higher prediction accuracy and shorter time. According to the findings of a survey that collected healthcare data, several key factors contribute to football knee injuries. To a degree, this algorithm can predict the likelihood of a football player's knee injury.

Familial Hypercholesterolemia-Risk-Score: A New Score Predicting Cardiovascular Events and Cardiovascular Mortality in Familial Hypercholesterolemia.

Objective: Familial hypercholesterolemia (FH) is associated with a high risk of premature atherosclerotic cardiovascular disease (ASCVD). However, this risk is highly heterogeneous and current risk prediction algorithms for FH suffer from limitations. The primary objective of this study was to develop a score predicting incident ASCVD events over 10 years in a large multinational FH cohort. The secondary objective was to investigate the prediction of major adverse cardiovascular events and cardiovascular mortality using this score. Approach and Results: We prospectively followed 3881 patients with adult heterozygous FH with no prior history of ASCVD (32 361 person-years of follow-up) from 5 registries in Europe and North America. The FH-Risk-Score incorporates 7 clinical variables: sex, age, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, hypertension, smoking, and lipoprotein (a) (Lp(a)) with a Harrell C-index for 10-year ASCVD event of 0.75, which was superior to the SAFEHEART-RE (Spanish Familial Hypercholesterolemia Cohort; 0.69). Subjects with an elevated FH-Risk-Score had decreases in 10-year ASCVD-free survival, 10-year major adverse cardiovascular event-free survival, and 30-year survival for CV mortality compared with the low-risk group, with hazard ratios of 5.52 (3.94-7.73), 4.64 (2.66-8.11), and 10.73 (2.51-45.79), respectively. The FH-Risk-Score showed a similar performance in subjects with and without an FH-causing mutation. Conclusions: The FH-Risk-Score is a stronger predictor of future ASCVD than the SAFEHEART-RE and was developed in FH subjects with no prior cardiovascular event. Furthermore, the FH-Risk-Score is the first score to predict CV death and could offer personalized cardiovascular risk assessment and treatment for patients with FH. Future studies are required to validate the FH-Risk-Score in different ethnic groups.

Convolutional Neural Networks for Automatic Virtual Metrology

To ensure stable manufacturing and high yield of production, factories (e.g., semiconductor or TFT-LCD fabs) conduct quality inspection on workpieces. They tend to adopt sampling inspection in consideration of reducing cost and cycle time, yet that fails to achieve real-time and online total inspection because of the sampling strategy and metrology delay. Automatic Virtual Metrology (AVM) is the best solution to tackle the problem mentioned above, due to the fact that it can convert sampling inspection with metrology delay into on-line and real-time total inspection. However, with the advancement of science and technology, the processes become more and more sophisticated, and the requirement for the accuracy of virtual metrology becomes higher. The current AVM prediction algorithm is the traditional machine learning method, Back-Propagation Neural Networks (BPNN). However, even if the amount of data in this method increases, the performance improvement has its limits, and it requires a strict and time-consuming feature selection process. To improve the prediction accuracy, this work proposes the deep learning method, Convolutional Neural Networks (CNN), for the AVM server. The accuracy of CNN improves as the amount of data grows. In other words, if there are sufficient data, the current accuracy limit of machine learning can be enhanced. Experimental results reveal that CNN can automatically extract highly informative features from the data and improves the original AVM accuracy.

Fixed switching frequency scheme for current predictive control of DFIG

This paper suggests a new current predictive control algorithm for the wind-driven doubly-fed induction generator (DFIG) based on space vector modulation (SVM) with fixed switching frequency. Combining the current predictive controller and SVM has the benefits of a predictive controller, and fixing the switching frequency improves the output waveform quality. However, if classical predictive control is used, the computational burden will still be high. The proposed algorithm uses SVM to fix the switching frequency of the rotor side converter. Additionally, by using an incremental algorithm, the proposed technique prevents examining all inputs over the prediction horizon. As a result, in the proposed current predictive algorithm the computational time is significantly reduced. The detailed state-space model of DFIG and rotor side converter are used to execute the predictive algorithm. The proposed algorithm uses the rotor current to predict the next behavior of the DFIG, and then, by using SVM, the duty cycles of 2-level voltage source inverter are obtained. Finally, the proposed controller's responses are compared to those of the SVM-based field-oriented control strategy. Simulation results of the proposed controller on a two-level voltage source inverter under a balanced three-phase power system illustrate the satisfactory active and reactive power tracking and improved quality of the inverter outputs.

An Improved Deadbeat Control Strategy Based on Repetitive Prediction Against Grid Frequency Fluctuation for Active Power Filter

In order to improve the harmonic compensation performance of active power filter (APF) in distribution network, based on deadbeat control theory, the command current prediction algorithm and current tracking control strategy are optimized in this article. Firstly, the command current repetitive prediction in abc coordinate system is transferred to dq for improving its accuracy in lead compensation, and the equivalent for fractional delay beat is achieved by Lagrange Interpolation Polynomial to solve the problem of inaccurate prediction caused by grid frequency fluctuation. Then, considering the inherent half-sampling-period delay of sinusoidal PWM (SPWM), an improved deadbeat control strategy for current tracking is proposed by estimating the output current of next sampling period. Because the output current in next sampling period is replaced by that in current sampling period with traditional deadbeat control strategy, this estimation could make up for the defect of low control precision caused by that replacement. After that, adding error repetitive correction into the improved deadbeat control channel to reduce the periodic tracking error of output current. Finally, the stability and accuracy of the improved control system are analyzed theoretically, and its feasibility and effectiveness are verified by the simulation and hardware-in-the-loop (HIL) experiments.

A Network-Adaptive Prediction Algorithm for Haptic Data Under Network Impairments

Real-time tele-haptic applications require capturing, compressing, transmitting, and displaying haptic information, which includes tactile and kinesthetic information. To achieve a high quality of service (QoS), real-time haptic data stream synchronization between local and remote environments is required. However, transmission of data over a computer network is often affected by network impairments, such as network delay, jitter, and packet loss, thus leading to system instability and poor performance. Current prediction algorithms for networked haptics comprise perceptual data reduction, traffic prioritization approaches, congestion control approaches, and radio resource allocation. However, the mentioned prediction algorithms either do not consider packet loss and time-varying delays (i.e., jitter) in their experimental setup, or only consider packet loss or delays. In real-world network environments, both packet loss and delays often occur simultaneously. In this work, a network adaptive Trust Strategy Prediction (TSP) algorithm was modified to work under both network impairments. The objective of the TSP is to maintain real-time haptic synchronization (haptic data stream synchronization) between the haptic interactive environments, by compensating network impairments using selective and specific prediction strategies, according to changes in the network’s characteristics. The experimental results demonstrate that TSP offers greater accuracy and smaller inconsistencies in terms of the predicted position, compared to the dead reckoning prediction and velocity estimation, which is often employed with filtering techniques.

COVID-19 cases prediction by using hybrid machine learning and beetle antennae search approach

The main objective of this paper is to further improve the current time-series prediction (forecasting) algorithms based on hybrids between machine learning and nature-inspired algorithms. After the recent COVID-19 outbreak, almost all countries were forced to impose strict measures and regulations in order to control the virus spread. Predicting the number of new cases is crucial when evaluating which measures should be implemented. The improved forecasting approach was then used to predict the number of the COVID-19 cases. The proposed prediction model represents a hybridized approach between machine learning, adaptive neuro-fuzzy inference system and enhanced beetle antennae search swarm intelligence metaheuristics. The enhanced beetle antennae search is utilized to determine the parameters of the adaptive neuro-fuzzy inference system and to improve the overall performance of the prediction model. First, an enhanced beetle antennae search algorithm has been implemented that overcomes deficiencies of its original version. The enhanced algorithm was tested and validated against a wider set of benchmark functions and proved that it substantially outperforms original implementation. Afterwards, the proposed hybrid method for COVID-19 cases prediction was then evaluated using the World Health Organization’s official data on the COVID-19 outbreak in China. The proposed method has been compared against several existing state-of-the-art approaches that were tested on the same datasets. The proposed CESBAS-ANFIS achieved R2 score of 0.9763, which is relatively high when compared to the R2 value of 0.9645, achieved by FPASSA-ANFIS. To further evaluate the robustness of the proposed method, it has also been validated against two different datasets of weekly influenza confirmed cases in China and the USA. Simulation results and the comparative analysis show that the proposed hybrid method managed to outscore other sophisticated approaches that were tested on the same datasets and proved to be a useful tool for time-series prediction.

The Irish dual-energy X-ray absorptiometry (DXA) Health Informatics Prediction (HIP) for Osteoporosis Project

PurposeThe purpose of the Irish dual-energy X-ray absorptiometry (DXA) Health Informatics Prediction (HIP) for Osteoporosis Project is to create a large retrospective cohort of adults in Ireland to examine the...

Link Prediction Adversarial Attack Via Iterative Gradient Attack

Increasing deep neural networks are applied in solving graph evolved tasks, such as node classification and link prediction. However, the vulnerability of deep models can be revealed using carefully crafted adversarial examples generated by various adversarial attack methods. To explore this security problem, we define the link prediction adversarial attack problem and put forward a novel iterative gradient attack (IGA) strategy using the gradient information in the trained graph autoencoder (GAE) model. Not surprisingly, GAE can be fooled by an adversarial graph with a few links perturbed on the clean one. The results on comprehensive experiments of different real-world graphs indicate that most deep models and even the state-of-the-art link prediction algorithms cannot escape the adversarial attack, such as GAE. We can benefit the attack as an efficient privacy protection tool from the link prediction of unknown violations. On the other hand, the adversarial attack is a robust evaluation metric for current link prediction algorithms of their defensibility.

A large peptidome dataset improves HLA class I epitope prediction across most of the human population.

Prediction of HLA epitopes is important for the development of cancer immunotherapies and vaccines. However, current prediction algorithms have limited predictive power, in part because they were not trained on high-quality epitope datasets covering a broad range of HLA alleles. To enable prediction of endogenous HLA class I–associated peptides across a large fraction of the human population, we used mass spectrometry to profile >185,000 peptides eluted from 95 HLA-A, B, C and G mono-allelic cell lines. We identified canonical peptide motifs per HLA allele, unique and shared binding submotifs across alleles, and distinct motifs associated with different peptide lengths. By integrating these data with transcript abundance and peptide processing, we developed HLAthena, providing allele-and-length-specific and pan-allele-pan-length prediction models for endogenous peptide presentation. These models predicted endogenous HLA class I–associated ligands with 1.5-fold improvement in positive predictive value compared with existing tools and correctly identified >75% of HLA-bound peptides that were observed experimentally in 11 patient-derived tumor cell lines.

Research on Motion Prediction Algorithm Based on Nash Equilibrium

The Nash equilibrium method is studied in this paper, and a new motion prediction algorithm based on Nash equilibrium is proposed. This method includes 3 main contents: 1. Analyze and abstract the influencing factors of object motion. 2. Analyze the effects of each factor and list the Nash equilibrium equation. 3. Find the optimal solution of the system of equations. Use Nash equilibrium method to attain the purpose of overall considerate all influencing factors of motion prediction. This algorithm can solve the problem that the current motion prediction algorithm can’t take all the influence factors into account. To verify this method, we introduce this method to the ocean drift motion prediction and prove the accuracy of the method through three sets of comparison experiments.

IMMU-24. A COMPREHENSIVE IN SILICO APPROACH TO DISCOVERING TUMOR REJECTION ANTIGENS IN MALIGNANT BRAIN TUMORS

Abstract BACKGROUND Proteins that can serve as effective tumor rejection antigens within brain tumors are poorly characterized. Current prediction algorithms relying on identification of mutated epitopes as neoantigens are limited in low mutational burden tumors. We have developed a multi-faceted computer algorighm for identifying tumor rejection antigens called the Open Reading Frame Antigen Network (ORAN). This pipeline provides a comprehensive solution for predicting brain tumor immunogenic targetable epitopes/transcripts. METHODS Human and murine RNASeq and WES data were QCed. Patients individual HLA-I and HLA-II haplotypes were predicted by highly customized Optitype and Phlat Algorithms. SNPs and Indels were called from tumors WES and read through matched RNASeq data. 19,131 transcripts expression were counted per TOIL algorithm. Tumor associate antigens(TAA) of individual patient or murine tumors were selected by setting a cutoff of Transcripts Per Million (TPM) > 1 on individual patient’s tumor, while RNA Seq data from 7000 normal tissues was used to identify tumor unique transcripts. Actual sequence of HLA and SNPed Consenses TAA(CTAA) were called. Only expressed mutations and personalized TAAs were used for antigenic epitope predictions. All neoepitopes were screened against a human reference proteomic library to guarantee that epitopes were not shared by other expressed isoforms or genes. In silico validation were preformed to cross validate predictions made by ORAN. RESULTS In medullobastoma (N=121 samples), ORAN gives an average of 15.6 MHC class I restricted neoepitopes,11.9 epitopes encoded by Indels and with 33.2 MHC class II restricted neoepitopes and 16.2 Indel antigens per patient. The TAAs of each patient reaches average 256 antigenic epitopes per patient. ORAN also predicts the exact HLA and neo-antigens from a validated neoantigens vaccine dataset (Gros A Nat Med 2016). CONCLUSION ORAN accurately identifies validated neoantigens and provides a comprehensive list of potential tumor rejection antigens within human and murine brain tumors.

High-Precision Power Load Forecasting Using Real-time Temperature Information and Deep Learning Method

Load forecasting is an important part of power system planning, which directly affects the safety and reliability of power grid operation. Real-time and high-precision load forecasting results are the key to improve the efficiency of the entire power grid. In order to solve the problem of low prediction accuracy of existing algorithms, based on the deep analysis of the strong correlation between temperature and electricity consumption, Long Short Memory Network (LSTM) is constructed, which implements the deep mining of the characteristics of historical electricity consumption data and the deep self-learning of the correlation between electricity consumption and temperature, and realizes the power load forecasting. Compared with traditional load forecasting techniques, the prediction results were significantly improved. In addition, the experiments utilizing the Google Tensor-flow platform were carried out to further study the impact of the combination of different activation functions on the prediction performance of the LSTM algorithm. The verification results shown that the prediction accuracy was significantly improved by using the ELU activation function than other activation functions. The developed algorithm could more effectively solve the low precision problem that is common in current prediction algorithms.

Novel Methylated DNA Markers Discriminate Advanced Neoplasia in Pancreatic Cysts: Marker Discovery, Tissue Validation, and Cyst Fluid Testing.

Pancreatic cystic lesions (PCLs) may be precancerous. Those likely to harbor high-grade dysplasia (HGD) or pancreatic cancer (PC) are targets for surgical resection. Current algorithms to predict advanced neoplasia (HGD/PC) in PCLs lack diagnostic accuracy. In pancreatic tissue and cyst fluid (CF) from PCLs, we sought to identify and validate novel methylated DNA markers (MDMs) that discriminate HGD/PC from low-grade dysplasia (LGD) or no dysplasia (ND). From an unbiased whole-methylome discovery approach using predefined selection criteria followed by multistep validation on case (HGD or PC) and control (ND or LGD) tissues, we identified discriminant MDMs. Top candidate MDMs were then assayed by quantitative methylation-specific polymerase chain reaction on archival CF from surgically resected PCLs. Of 25 discriminant MDMs identified in tissue, 13 were selected for validation in 134 CF samples (21 cases [8 HGD, 13 PC], 113 controls [45 ND, 68 LGD]). A tree-based algorithm using 2 CF-MDMs (TBX15, BMP3) achieved sensitivity and specificity above 90%. Discrimination was significantly better by this CF-MDM panel than by mutant KRAS or carcinoembryonic antigen, with areas under the receiver operating characteristic curve of 0.93 (95% confidence interval: 0.86-0.99), 0.71 (0.57-0.85), and 0.72 (0.60-0.84), respectively. Cutoffs for the MDM panel applied to an independent CF validation set (31 cases, 56 controls) yielded similarly high discrimination, areas under the receiver operating characteristic curve = 0.86 (95% confidence interval: 0.77-0.94, P = 0.2). Novel MDMs discovered and validated in tissue accurately identify PCLs harboring HGD/PC. A panel of 2 MDMs assayed in CF yielded results with potential to enhance current risk prediction algorithms. Prospective studies are indicated to optimize and further evaluate CF-MDMs for clinical use.

A Comparative Study of Customer Complaint Prediction Model of Time Series, Multiple Linear Regression and BP Neural Network

The current prediction algorithm is mainly the “time series”, the “multiple linear regression” and the “BP neural network”. This article studies and compares the three algorithms in the field of customer complaint prediction. By using SPSS, this article predicts taking “Jibei Electric Power Customer Complaints” as the target. By comparing and analyzing the actual prediction results, the “BP neural network” algorithm is described as the most suitable complaint prediction for customer.

The Role of Mass Spectrometry in the Advancement of HLA Epitope Prediction

A challenge in developing personalized cancer immunotherapies is the prediction of putative cancer‐specific antigens. Currently, predictive algorithms are used to infer binding of peptides to human leukocyte antigen (HLA) heterodimers to aid in the selection of putative epitope targets. One drawback of current epitope prediction algorithms is that they are trained on datasets containing biochemical HLA‐peptide binding data that may not completely capture the rules associated with endogenous processing and presentation. The field of MS has made great improvements in instrumentation speed and sensitivity, chromatographic resolution, and proteogenomic database search strategies to facilitate the identification of HLA‐ligands from a variety of cell types and tumor tissues. As such, these advances have enabled MS profiling of HLA‐binding peptides to be a tractable, orthogonal approach to lower throughput biochemical assays for generating comprehensive datasets to train epitope prediction algorithms. Today, we will highlight recent progress made in the field of HLA‐ligand profiling enabled by MS and its impact on epitope prediction strategies.This abstract is from the Experimental Biology 2019 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.

Meta-Circuit machine: Inferencing human collaborative relationships in heterogeneous information networks

Human collaborative relationship inference is a meaningful task for online social networks and is called link prediction in network science. Real-world networks contain multiple types of interacting components and can be modeled naturally as heterogeneous information networks (HINs). The current link prediction algorithms in HINs fail to effectively extract training samples from snapshots of HINs; moreover, they underutilise the differences between nodes and between meta-paths. Therefore, we propose a meta-circuit machine (MCM) that can learn and fuse node and meta-path features efficiently, and we use these features to inference the collaborative relationships in question-and-answer and bibliographic networks. We first utilise meta-circuit random walks to obtain training samples in which the basic idea is to perform biased meta-path random walks on the input and target network successively and then connect them. Then, a meta-circuit recurrent neural network (mcRNN) is designed for link prediction, which represents each node and meta-path by a dense vector and leverages an RNN to fuse the features of node sequences. Experiments on two real-world networks demonstrate the effectiveness of our framework. This study promotes the investigation of potential evolutionary mechanisms for collaborative relationships and offers practical guidance for designing more effective recommendation systems for online social networks.

Computational analysis in designing T cell epitopes enriched peptides of Ebola glycoprotein exhibiting strong binding interaction with HLA molecules

Computational approach has shown remarkable progress in epitope mapping, paving the way to finding vaccine candidates against different viruses. In the current study, prediction algorithms and molecular docking were applied to select peptides containing multiple Ebola glycoprotein epitopes showing interaction with different HLA molecules. Six peptides containing overlapping multiple HLA I (CD8+) and II (CD4+) restricted T cell epitopes were generated via consensus approach applying six different prediction tools. Four (P1, P2, P5 and P6) out of six peptides were selected after screening for absence of undesirable responses and presence of B cell epitopes. Peptide-HLA interaction analysis based on Autodock Vina and CABS-dock showed strong binding of these four peptides with eighteen HLA molecules. HLA coverage analysis from each prediction tool showed that these peptides were able to bind to diverse HLA-A, HLA-B, HLA-DP, HLA-DQ and HLA-DR alleles. Population coverage analysis of peptides for expected immune response in four different continents (Africa, America, Asia and Europe) have shown average population coverage viz, P1 (95%), P2 (96%), P5 (91%) and P6 (94%). Further, these peptides were found to be nearly 100% conserved in Zaire Ebola virus while LANETTQALQLF (P5) was found to be 100% conserved in Zaire, Sudan, Bundibugyo and Tai Forest species. Therefore, these peptides capable of inducing T and B cell response and being presented by a wide range of HLA molecules have a strong potential to be part of diagnostic and preventive tools against Ebola virus disease.

Analysis of Current Predictive Control Algorithm for Permanent Magnet Synchronous Motor Based on Three-Level Inverters

In some high-precision control applications, such as defense industry and computerized numerical control machine tools, fast and stable electromagnetic torque response is required to ensure the high dynamic performance of the system, while traditional PI control often cannot meet its requirements. For this purpose, a predictive control algorithm based on the deadbeat control algorithm is proposed in order to improve the performance of the motor current loop. In order to solve the problem that the conventional deadbeat control algorithm has a large dependence on system parameters and low robustness, this paper proposes an improved deadbeat control scheme for the permanent magnet synchronous motor based on the three-level inverters. The scheme is based on the second theorem of Lyapunov stability. The improved deadbeat control algorithm can achieve a good output waveform when the switching frequency is not high and the response speed is fast. The robustness of the system is improved, and there are good characteristics in reduced torque ripple. Compared to the traditional PI regulators, the improved deadbeat control can quickly track the current commands without overshoot and oscillation and suppress torque ripple. The simulation and experimental results show that the improved deadbeat control proposed in this paper has good dynamic and static performance.