DNA sequence analysis is at the core of bioinformatics approaches in the study of genomes, genes and proteins. Recent studies revealed that codon (triplet nucleotides) is a useful means to detect patterns of genome sequences and is the basis of amino acids, which in turn translate to proteins. Sixty four codons are commonly grouped into an alphabet of twenty amino acids. Grouping twenty amino acids into reduced alphabets may aid in detecting patterns, modeling and designing proteins, and finding consensus sequences. It can make patterns simply. In this paper, a novel approach is used to generate a reduced alphabet of twenty amino acids. It is based on base-pairing and incorporates physical features of amino acids. The base-pairing is formulated by a complementary tree algorithm. The tree preserves the relationship among four groups of amino acids (nonpolar, polar, basic, and acidic), and yet generates fifteen amino acid groups. The resulting reduced alphabets are tested on alternative splicing patterns, which is the process related to protein functions and an important process for increasing the diversity arising from a single gene, in a genome sequence. The evaluation shows that this grouping is efficient in detecting patterns of genome sequences.