Cancer revealed by uncontrolled cell growth due to genetic mutations, phase a significant challenge to global health. Traditional methods for mutation classification often involve labor intensive processes and are limited by the complexity of genetic data. Evaluation of Deep learning has powerful tool for investigation in recent years and interpreting large-scale genomic data. The primary focus of this study is deep learning methods regarding the coordination of cancer mutations based on DNA sequences. Convolutional neural networks, Graph neural networks and transfer learning are few examples of the varied deep learning applied in order to recognize and categorize genetic mutations connected to different cancer types. Here discusses the stability and drawback of these models, the preprocessing steps involved in handling DNA sequences, and the performance metrics used to evaluate their effectiveness. Also suggest that deep learning models, with their ability to capture intricate patterns in DNA sequences, hold significant promise for advancing cancer genomics and improving clinical outcomes. Future research directions are proposed to address current limitations and to further harness the probability of deep learning in cancer mutation classification.