Automated exploitation of deep learning for cancer patient stratification across multiple types

• DOI: 10.1093/bioinformatics/btad654
• Publication Date: 2023-11-07T19:03:13Z
• Authors: Pingping Sun; Shijie Fan; Shaochuan Li; Yingwei Zhao; Chang Lu; Ka-Chun Wong; Xiangtao Li
• Keywords: Stratification (seeds); Risk Stratification

Recent frameworks based on deep learning have been developed to identify cancer subtypes from high-throughput gene expression profiles. Unfortunately, the performance of is highly dependent its neural network architectures which are often hand-crafted with expertise in networks, meanwhile, optimization and adjustment usually costly time consuming.To address such limitations, we proposed a fully automated architecture search model for diagnosing consensus molecular data (DNAS). The uses ant colony algorithm, one heuristic swarm intelligence algorithms, optimize architecture, it can automatically find optimal diagnosis space. We validated DNAS eight colorectal datasets, achieving average accuracy 95.48%, specificity 98.07%, sensitivity 96.24%, respectively. Without loss generality, investigated general applicability further other types different platforms including lung breast cancer, achieved an area under curve 95% 96%, In addition, conducted ontology enrichment pathological analysis reveal interesting insights into subtype identification characterization across multiple types.The source code be downloaded https://github.com/userd113/DNAS-main. And web server publicly accessible at 119.45.145.120:5001.