This chapter introduces gene feature prediction methods for the nonprotein-coding part of the genome, including tools to detect RNA genes. Sequence-defined features are exclusively defined by information contained in the sequence itself and are not subject to experimental verification. The compositional properties of DNA sequences include DNA base composition and short oligomer distributions up to the hexanucleotide level. Compositional properties have also been used in gene prediction to distinguish introns from coding exons. The two types of sequence alignment algorithms are commonly distinguished which include global and local. A global algorithm attempts to align two sequences over their entire lengths, regardless of whether there is any similarity at all in a particular region. A local alignment algorithm searches for matching segments between two sequences. Sequence alignment algorithms are often used to compare a query sequence against a large sequence database. It is found that weight matrices are more flexible descriptors of degenerate sequence motifs which can account for varying degrees of severity of base substitutions in a functional element.