BLAT, a BLAST-like alignment tool, is designed for rapid mRNA/DNA and protein alignments between vertebrate genomes. It is significantly faster and more accurate than existing tools, achieving 500 times the speed of mRNA/DNA alignments and 50 times the speed of protein alignments at typical sensitivity settings. BLAT's efficiency stems from indexing all non-overlapping K-mers in the genome, which fits into the RAM of inexpensive computers. The tool performs several stages: finding homologous regions, aligning these regions, stitching aligned regions into larger alignments, and revisiting small internal exons. BLAT is available in client/server and stand-alone versions, with a web-based server hosted at http://genome.ucsc.edu. It is used for various applications, including aligning ESTs and whole-genome shotgun data, and has been shown to be highly effective in detecting homologies between mouse and human genomes. The paper discusses the optimization of BLAT, its performance on different K-mer sizes, mismatch schemes, and the number of required index matches, and compares it with other alignment programs.BLAT, a BLAST-like alignment tool, is designed for rapid mRNA/DNA and protein alignments between vertebrate genomes. It is significantly faster and more accurate than existing tools, achieving 500 times the speed of mRNA/DNA alignments and 50 times the speed of protein alignments at typical sensitivity settings. BLAT's efficiency stems from indexing all non-overlapping K-mers in the genome, which fits into the RAM of inexpensive computers. The tool performs several stages: finding homologous regions, aligning these regions, stitching aligned regions into larger alignments, and revisiting small internal exons. BLAT is available in client/server and stand-alone versions, with a web-based server hosted at http://genome.ucsc.edu. It is used for various applications, including aligning ESTs and whole-genome shotgun data, and has been shown to be highly effective in detecting homologies between mouse and human genomes. The paper discusses the optimization of BLAT, its performance on different K-mer sizes, mismatch schemes, and the number of required index matches, and compares it with other alignment programs.