Research on a novel gene sequence prediction and homomorphic encryption method based on Mamba-VMD.

Gene sequence prediction not only effectively identifies homologous genes but also provides crucial insights into gene function and evolutionary relationships, making it one of the significant research directions in the field of bioinformatics. Given the high sensitivity of bioinformatics data, particularly gene sequences, plaintext transmission and processing in cloud environments pose risks of privacy leakage. To address this, this study innovatively proposes an analytical method combining Mamba neural network-based gene sequence prediction with homomorphic encryption, enabling secure data transmission while employing deep learning for sequence prediction. Using the monkeypox virus experimental ID SRX17751190 and sequencing ID SRR21755835 as examples, the study first decomposes the original gene sequence using VMD modal decomposition, then predicts the gene sequence from the decomposed components using the Mamba neural network, and finally performs homomorphic encryption and spatial similarity analysis on the predicted data in a cloud computing environment. Experimental results demonstrate that the average MAE, MSE, RMSE, MAPE, and MSPE for the top ten predicted gene sequences are 0.0140, 0.0003, 0.0190, 0.6527, and 789.52959, respectively, with an average CKKS homomorphic encryption computation error of 0.582492456. This ensures secure similarity calculations for gene sequence data in cloud environments.