RESUMO
In recent times, time-to-event data such as time to failure or death is routinely collected alongside high-throughput covariates. These high-dimensional bioinformatics data often challenge classical survival models, which are either infeasible to fit or produce low prediction accuracy due to overfitting. To address this issue, the focus has shifted towards introducing a novel approaches for feature selection and survival prediction. In this article, we propose a new hybrid feature selection approach that handles high-dimensional bioinformatics datasets for improved survival prediction. This study explores the efficacy of four distinct variable selection techniques: LASSO, RSF-vs, SCAD, and CoxBoost, in the context of non-parametric biomedical survival prediction. Leveraging these methods, we conducted comprehensive variable selection processes. Subsequently, survival analysis models-specifically CoxPH, RSF, and DeepHit NN-were employed to construct predictive models based on the selected variables. Furthermore, we introduce a novel approach wherein only variables consistently selected by a majority of the aforementioned feature selection techniques are considered. This innovative strategy, referred to as the proposed method, aims to enhance the reliability and robustness of variable selection, subsequently improving the predictive performance of the survival analysis models. To evaluate the effectiveness of the proposed method, we compare the performance of the proposed approach with the existing LASSO, RSF-vs, SCAD, and CoxBoost techniques using various performance metrics including integrated brier score (IBS), concordance index (C-Index) and integrated absolute error (IAE) for numerous high-dimensional survival datasets. The real data applications reveal that the proposed method outperforms the competing methods in terms of survival prediction accuracy.
Assuntos
Redes Neurais de Computação , Humanos , Análise de Sobrevida , Estatísticas não Paramétricas , Biologia Computacional/métodosRESUMO
RNA modifications are pivotal in the development of newly synthesized structures, showcasing a vast array of alterations across various RNA classes. Among these, 5-hydroxymethylcytosine (5HMC) stands out, playing a crucial role in gene regulation and epigenetic changes, yet its detection through conventional methods proves cumbersome and costly. To address this, we propose Deep5HMC, a robust learning model leveraging machine learning algorithms and discriminative feature extraction techniques for accurate 5HMC sample identification. Our approach integrates seven feature extraction methods and various machine learning algorithms, including Random Forest, Naive Bayes, Decision Tree, and Support Vector Machine. Through K-fold cross-validation, our model achieved a notable 84.07% accuracy rate, surpassing previous models by 7.59%, signifying its potential in early cancer and cardiovascular disease diagnosis. This study underscores the promise of Deep5HMC in offering insights for improved medical assessment and treatment protocols, marking a significant advancement in RNA modification analysis.
Assuntos
5-Metilcitosina/análogos & derivados , Algoritmos , Redes Neurais de Computação , Teorema de Bayes , Máquina de Vetores de Suporte , RNARESUMO
Probability distributions play a pivotal and significant role in modeling real-life data in every field. For this activity, a series of probability distributions have been introduced and exercised in applied sectors. This paper also contributes a new method for modeling continuous data sets. The proposed family is called the exponent power sine-G family of distributions. Based on the exponent power sine-G method, a new model, namely, the exponent power sine-Weibull model is studied. Several mathematical properties such as quantile function, identifiability property, and rth moment are derived. For the exponent power sine-G method, the maximum likelihood estimators are obtained. Simulation studies are also presented. Finally, the optimality of the exponent power sine-Weibull model is shown by taking two applications from the healthcare sector. Based on seven evaluating criteria, it is demonstrated that the proposed model is the best competing distribution for analyzing healthcare phenomena.
RESUMO
In the model-based approach, researchers assume that the underlying structure, which generates the population of interest, is correctly specified. However, when the working model differs from the underlying true population model, the estimation process becomes quite unreliable due to misspecification bias. Selecting a sample by applying the balancing conditions on some functions of the covariates can reduce such bias. This study aims at suggesting an estimator of population total by applying the balancing conditions on the basis functions of the auxiliary character(s) for the situations where the working model is different from the underlying true model under a ranked set sampling without replacement scheme. Special cases of the misspecified basis function model, i.e. homogeneous, linear, and proportional, are considered and balancing conditions are introduced in each case. Both simulation and bootstrapped studies show that the total estimators under proposed sampling mechanism keep up the superiority over simple random sampling in terms of efficiency and maintaining robustness against model failure.
RESUMO
Background: Lung cancer is the top cause of mortality in males and the second largest cause of cancer-related fatalities in women worldwide. Non-small cell lung cancer (NSCLC) cases are discovered at an advanced stage, raising major challenges in disease management and survival outcomes. This study aimed to investigate the clinical findings and management of stage IIIB and IV NSCLC patients for better decision-making, disease management, and understanding of this fatal disease. Methods: In this cohort study of 340 patients, a total of 140 (41.2%) were diagnosed with advanced-stage NSCLC at a mean age of 64 years. The electronic data of patients from 2015 to 2021 who met the inclusion criteria were retrieved from two tertiary hospitals in Riyadh, Saudi Arabia, and an Excel sheet was used to record the variables. Patients' data including all categorical variables such as gender, stage, metastasis, ALK, EGFR, and ROS, etc., and continuous variables such as age and body mass index (BMI) were retrieved and analyzed. Results: The multivariate Cox-regression model indicated that smoking was the significant risk factor of death for two-thirds of male smokers (37.9%), with a median survival time of 123 days. Disease progression was higher with pleural and brain metastasis, and localized metastasis was the highest in 75% of patients. The intent of treatment was mainly palliative, however, a statistically significant association was found with the simultaneous use of chemotherapy and immunotherapy. Patients' response to first-line treatment revealed a significant improvement if chemotherapy treatment was maintained at the same dose without interruption of dosage. Conclusions: The overall cure and survival rates for NSCLC remain low, particularly in metastatic disease. Therefore, continued research into new drugs and combination therapies is required for better decision-making to expand the clinical benefit to a broader patient population and to improve outcomes in NSCLC.