The invention relates to a characteristic extraction and classification method and device for DNA (Deoxyribo-Nucleic Acid) binding protein sequence information. The method comprises the following steps that: firstly, carrying out theory argumentation, and analyzing and arranging collected data to obtain a reliable dataset with a biological meaning and a statistical meaning; then, extracting effective protein sequence data characteristic parameters from a complex protein three-dimensional structure as a key link which is a way of converting sequence character information into digital characteristic information, designing a reasonable classification algorithm for extracted characteristic data, and screening the characteristics favorable for classification for realizing target classification;and finally, adopting a reasonable and fair evaluation system, including testing methods, checking means, evaluation index selection and the like, for classification performance. By use of the method, requirements on high-flux protein sequencing function annotation can be met, automated DNA binding protein sequence function annotation can be realized, and meanwhile, the characteristics which areput forward can assist biologists in carrying out experimental analysis and research on the DNA binding protein sequence.