The comparative studies of k-mer distribution in human and mouse TFBS sequences listed in TRANSFAC database are given. A non-alignment based approach for fast genome-wide discovery of transcription regulatory k-mer motifs (TRKMs) is proposed. The method is called distance-based conservative k-mer searching algorithm (DCKS) which is based on the conservation of k-mer pair distance. By use of DCKS the prediction accuracy of human transcription regulatory 7-mer motifs is: sensitivity 90%, specificity 78%, and correlation coefficient 0.65.