Proposed in this paper is a compression algorithm of DNA sequences based on the mixed statistical mo-del,which estimates the probability of each symbol of a DNA sequence in line with the principle of expert model al-gorithm (XM algorithm) and the mixed finite context statistical model.Then,the estimated probability is applied tothe arithmetic coding to encode each symbol of standard DNA sequences.Experimental results show that (1) ascompared with the single finite context model,the mixed statistical model helps to obtain better compression effect;(2) the proposed algorithm based on mixed statistical model helps to achieve higher compression ratio than those ofsome other classical compression algorithms; (3) it effectively overcomes the deficiencies of XM algorithm for thestandard dataset compression of DNA sequences,although the XM algorithm based on statistical information is ratheradvanced; and (4) the proposed algorithm needs to be improved for the compression of high- throughput DNA se-quences.