Leveraging deep learning for epigenetic protein prediction: a novel approach for early lung cancer diagnosis and drug discovery.

Abstract

Epigenetic protein (EP) plays a crucial role in influencing disease development, controlling gene expression, and shaping cell identity. They hold potential as targets for future therapies, and studying their mechanisms can lead to improved diagnosis and treatment strategies for various diseases. Anticipating EP is imperative, yet conventional experimental approaches for prediction prove time-intensive and expensive. This work constructed CNN-BiLSTM, computational method for identification of EP prediction. Utilizing primary sequences, two datasets were constructed, and an amphiphilic pseudo amino acid, group dipeptide composition and group amino acid composition were devised to extract numerical features. Model training incorporated a suite of deep learning architectures, including BiLSTM, GRU, and CNN. Notably, an ensemble model combining CNN and BiLSTM, trained using AmpPseAAC features, demonstrated superior performance across both training and testing datasets compared to other predictors. This research contributes to the ongoing efforts to revolutionize therapeutic approaches by facilitating the identification of novel drug targets and improving disease treatment outcomes.