The COVID-19 pandemic and drug-resistant bacterial infections have caused significant economic losses and harm to individuals. There is an urgent need for safe and efficient anti-coronavirus drugs, including anti-coronavirus peptides. These small-molecule peptides have demonstrated high efficacy, broad-spectrum coronavirus activity, low toxicity, and minimal drug resistance. Traditional identification methods for anti-coronavirus peptides are inefficient and resource-intensive, relying on experimental means. However, the latest prediction methods combine machine learning and bioinformatics, offering a more efficient and cost-effective approach that does not rely on biological experiments.

In this research, we integrated a new machine learning technology with bioinformatics to analyze anti-coronavirus peptides using deep neural networks that incorporated 9 classification models. We conducted a comparative analysis of anti-coronavirus peptide predictions across three different datasets and validated our findings using an independent dataset, which demonstrated better performance.

Through this method, its highest accuracy rate reaches 98%, and the MCC value exceeds 0.9, it tested independently on three different datasets, the average accuracy is 96.33%, Compared with the state-of-the-art predictors, ACP-Dnnel improved the MCC value by 10.1%, Sn value by 16.4% and ACC value by 7.3% respectively, after validation by the latest independent validation dataset, ACP-Dnnel has a maxi-mum Sn value of 98.9% and an ACC value of 96.8%, so the algorithm model can perfectly replace the method of laboratory identification of anti-coronavirus peptides, and also improve the recognition efficiency of anti-coronavirus peptides.