Leveraging Hybrid Deep Learning Models for Predictive Modelling and Early Diagnosis of Adrenocortical Carcinoma using Multi-Omics Data

This study addresses the challenges in diagnosing adrenocortical carcinoma (ACC), a rare and aggressive cancer with poor prognosis and limited treatment options. Conventional diagnostic methods often fall short due to ACC's complex clinical and molecular heterogeneity. Recent advances in artificial intelligence (AI) combined with multi-omics data integration offer promising avenues for improving ACC detection and classification. The study surveys three key computational models-Graph Convolutional Networks (GCN), Weighted Majority Rule-based Clustering Algorithm (WMRCA), and Residual Networks (ResNet)-highlighting their strengths and limitations in processing diverse biological datasets. Recognising that no single model fully captures ACC's complexity, a novel hybrid framework is proposed, integrating these models through an attention-based fusion strategy to enhance diagnostic accuracy, subtype classification, and interpretability. This approach aims to deliver clinically actionable insights, including molecular subtypes and prognostic markers. Future directions emphasise the development of interpretable, robust AI systems and collaborative data sharing to address challenges posed by ACC's rarity and heterogeneity, ultimately supporting early diagnosis and personalised treatment.