Prediction of Proto-Oncogene Using Bidirectional GRU and Attention

One of the key responsibilities of bioinformatics is now protein sequence prediction, thanks to the advancements in genome sequencing technology. The primary means of uncontrolled cancer growth is the absence of tumour suppression gene (TSG) regulatory ability and proto-oncogene (OG) mutations. Even though a cancer is a complicated mixture of several disorders, computational research may be able to identify genes linked to OG or TSG activity, which may help with the creation of drugs that target the condition directly.

Recently, the attention mechanism in deep learning has emerged as a cutting-edge method for protein sequence classification. The attention-based strategy can provide a reliable and comprehensible way to help overcome current challenges in characterising deep neural networks for protein sequence classification. This study proposes two approaches like Attention with Convolutional Neural Network (ACNN) and Attention with Bi directional Gated Recurrent Units (ABiGRU) to predict Proto-oncogene protein sequence. The proposed deep learning with Attention model is validated using Independent test and K-fold cross-validation test. Moreover this study has performed Ablation study and Statistical significant Testing to access the superiority of the proposed model.

The results are analyzed by the benchmark Uniprot dataset. Independent testing of ACNN model gives 96.85% of accurate results and ABiGRU model gives 97.53%of accurate results.

According to these findings, the suggested model may be crucial in determining a cancer patient's early prognosis and in helping researchers identify cancer-fighting systems.