DEAN: Improving the Efficiency of Disease Network Using Selective Ensemble Network Aggregation Model

Disease networks facilitate the investigation of common linkages between the genetic sequences of distinct diseases. Researchers can analyze the implications of inter-disease interactions by discovering common gene links. The construction of these networks requires extensive data collection, preprocessing, variance analysis, grouping of features, and connection. Variance analysis and feature clustering needs huge processing time and are responsible for successfully linking genetic sequences. Existing high AUC (area under the curve) performance designs allow both the detection and analysis of dual disease links, limiting their applicability to networks including several illnesses. This article proposes the development of DEAN, a framework for choosing ensemble networks for aggregation, to improve the scalability of existing disease networks. This approach analyzes and chooses aggregate Kolmogorov–Smirnov (KS), differential expression (DE), empirical mode decomposition (EMD), and DE variance (DEVar) coefficients to assess the relationship between various disease types. These coefficients are processed using adaptable hierarchical clustering (AHC). The AHC approach optimizes cluster assessment by maximizing the sum of squared distances. The optimum clusters model is first enlarged from individual coefficient analysis to integrated coefficient analysis to improve AUC performance. On datasets pertaining to breast cancer, brain cancer, and post-traumatic stress disorder (PTSD), the proposed DEAN model evaluates, and the produced results were compared to those obtained using the EMD, DE, DEVar, and KS coefficient analysis methods. On these datasets, the proposed DEAN model performed 20% better than DE, 18% better than DEVar, 14% better than KS, and 5% better than EMD models. In contrast to individual models, which are only scalable to a single dataset, the proposed model is scalable to a variety of illness types due to its consistent performance across diverse datasets. In addition, this paper provides several recommendations for enhancing DEAN’s AUC performance and incorporating more datasets for better scalability levels.