Establishment of a Combined Diagnostic Model of Abdominal Aortic Aneurysm with Random Forest and Artificial Neural Network

Abstract

Background Abdominal aortic aneurysm (AAA), a disease with high mortality, is limited by the current diagnostic methods in the early screening. This study aimed to construct a diagnostic model for AAA by using a novel machine learning method, i.e., an ensemble of the random forest (RF) algorithm and an artificial neural network (ANN) (RF-ANN), to identify potential AAA-associated genetic biomarkers. Methods Through a search of the Gene Expression Omnibus (GEO) database, two large-sample gene expression datasets (GSE57691 and GSE47472) were identified and downloaded. The differentially expressed genes (DEGs) between the AAA and normal control samples were identified, followed by Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis using the Database for Annotation, Visualization, and Integrated Discovery (DAVID). Then, RF-ANN was used to identify the key genes from the DEGs, and an AAA diagnostic model was established. Finally, the diagnostic performance of the model was assessed using the area under the receiver operating characteristic curve (AUC) with GSE47472 as a test dataset. Results Using GSE57691, we obtained 2486 DEGs, 52 biological process annotations, 17 cellular component annotations, 17 molecular function annotations, and 13 significantly enriched KEGG pathways. Out of these DEGs, we further identified 74 key candidate feature genes by using the RF machine learning algorithm. The weight of each key gene was calculated by the ANN with GSE57691 as a training dataset to construct an AAA diagnostic model. A transcription factor (TF) regulatory network of key genes was constructed. Finally, GSE47472 was used to validate the model. The AUC value was 0.786, indicating that the model had a highly satisfactory diagnostic performance. Conclusion Potential AAA-associated gene biomarkers were identified, and a diagnostic model of AAA was established. This study may provide a valuable reference for early clinical diagnosis and the search for therapeutic targets of AAA.

Automated Design of Affine Maximizer Mechanisms in Dynamic Settings

“…Due to the difficulty of analytically finding optimal mechanisms, a number of works have instead attempted to treat static mechanism design as a computational optimization problem, starting with Conitzer and Sandholm (2002) and Sandholm (2003), and learn good mechanisms from samples, starting with Likhodedov and Sandholm (2004). One line of work makes use of static affine maximizers and achieves good performance in multi-item multi-bidder auctions (Sandholm and Likhodedov 2015;Curry, Sandholm, and Dickerson 2023;Duan et al 2023). There is also a line of learning theory research on choosing the AMA parameters given samples from the valuation distribution Vitercik 2016, 2018;Balcan et al 2021).…”

Advancing sustainable air quality through calibration of miniature air quality monitors with SRA-SVR combined model

“…Stepwise regression is an analytical method that introduces (or removes) variables into (or from) a regression equation one by one, according to the significance of the effect of the explanatory variables on the explained variables. In the case of multivariate screening of important variables, the use of stepwise regression analysis is a better Frontiers in Environmental Science frontiersin.org way of avoiding multicollinearity between explanatory variables and eliminates the need for a heavy variable screening process (Duan et al, 2022).…”