Search

BackgroundInsomnia (ISM) is one of the non-traditional drivers of atherosclerosis (AS) and an important risk factor for AS-related cardiovascular disease. Our study aimed to explore the shared pathways and diagnostic biomarkers of ISM-related AS using integrated bioinformatics analysis.MethodsWe download the datasets from the Gene Expression Omnibus database and the GeneCards database. Weighted gene co-expression network analysis and gene differential expression analysis were applied to screen the AS-related gene set. The shared genes of ISM and AS were obtained by intersecting with ISM-related genes. Subsequently, candidate diagnostic biomarkers were identified by constructing protein–protein interaction networks and machine learning algorithms, and a nomogram was constructed. Moreover, to explore potential mechanisms, a comprehensive analysis of shared genes was carried out, including enrichment analysis, protein interactions, immune cell infiltration, and single-cell sequencing analysis.ResultsWe successfully screened 61 genes shared by ISM and AS, of which 3 genes (IL10RA, CCR1, and SPI1) were identified as diagnostic biomarkers. A nomogram with excellent predictive value was constructed (the area under curve of the model constructed by the biomarkers was 0.931, and the validation set was 0.745). In addition, the shared genes were mainly enriched in immune and inflammatory response regulation pathways. The biomarkers were associated with a variety of immune cells, especially myeloid immune cells.ConclusionWe constructed a diagnostic nomogram based on IL10RA, CCR1, and SPI1 and explored the inflammatory-immune mechanisms, which indicated new insights for early diagnosis and treatment of ISM-related AS.

Abstract In representation learning domain, the mainstream methods for model ensemble include “implicit” ensemble approaches, such as using techniques like dropout, and “explicit” ensemble methods, such as voting or weighted averaging based on multiple model outputs. Compared to implicit ensemble techniques, explicit ensemble methods allow for more flexibility in combining models with different structures to obtain different perspectives on representations. However, the representations obtained from different models do not guarantee a linear relationship, and simply linearly combining multiple model outputs may result in a degraded performance. Meanwhile, employing non‐linear fusion mechanisms such as distillation and meta‐learning can be uninterpretable and time‐consuming. To this end, we propose the hypothesis of linear fusion based on the output representations of deep learning models, and design a interpretable linear fusion method based on this hypothesis. This method applies a transform layer to map the output representations of different models to the same classification center. Experimental results demonstrate that compared to directly averaging the representations, our method achieves better performance. Additionally, our method retains the convenience of direct averaging while offering improved performance in terms of time and computational efficiency compared to non‐linear fusion. Furthermore, we test the applicability of our method in both computer vision and natural language processing representation tasks using supervised and semi‐supervised approaches.