We propose an alternate measure of information, called superinformation, which has been found to be very effective for analyzing the coding and noncoding regions of the DNA. This superinformation is actually a measure of the "randomness of randomness." It has been found to be highly accurate in classifying coding and noncoding regions of human DNA. In the proposed method, no prior training is required. This technique exhibits higher accuracy than previously reported techniques in distinguishing between the coding and the noncoding portions of the DNA. Superinformation can also be used to analyze the untranslated regions in various genes.