As a main method for predicting the functionality of genes, the sequence similarity querying technique is becoming one of the research hotspots in bioinformatics. The similarity of gene sequence and structure usually determines the similarity of gene functionality, and the function of an unknown gene can be predicted by sequence similarity querying. After analyzing the advantages and shortcomings of related work such as frequency transformation and wavelet transformation used in MRS, a new sequence similarity query processing technique based on the two-Partitioning Frequency Transformation 2-PFT is proposed. Firstly, the Two-partitioning frequency transformation and the corresponding distance function are designed. They have a higher filtering ability than frequency transformation and wavelet transformation, and the system performance is thus improved significantly. Secondly, the problem of processing the queries with any length is solved. Theoretical proof and experimental results show that the 2-PFT system outperforms the MRS system greatly.