MiRNAClassifier: Improving Cancer Biomarker Identification with A Spatio-temporal Adjustable Deep Learning Algorithm

Precision cancer detection and therapy need cancer circulating microRNA (miRNA) biomarkers, crucial indications of cancer genesis and progression. These biomarkers may be detected via liquid biopsy and tissue-based procedures, which need advanced sequencing and analytics to interpret. Traditional data mining and machine learning methods have poor classification performance, notably for several miRNA classes. The innovative framework ATSTCNN-SNet-AGO-CBC improves miRNA biomarker classification and identification. A newly developed technique called Dynamic Multi-Stage Noise Suppression Filtering pre-filters the pictures before using gene expression input information from the Cancer Genome Atlas. The twin spatio-temporal convolutional neural network with a SqueezeNet layer optimized for feature discrimination is used in this new model. Batch normalization is not used. The proposed Adaptive Gorilla Optimization (AGO) technique tunes hyperparameters for optimal weight updates. Using this method, miRNA biomarkers may be classified as diagnostic, therapeutic, or prognostic with unprecedented accuracy. Experimental optimization improves accuracy by 28.76% and precision by 45.62% over state-of-the-art techniques. These findings demonstrate the ATSTCNN-SNet-AGO-CBC framework's potential to improve miRNA biomarker categorization for cancer diagnosis, treatment, and prognosis. The model sets the standard for cancer biomarker discovery using state-of-the-art deep learning algorithms for adaptive optimization.