Detecting frame shifts by amino acid sequence comparison.

Various amino acid substitution scoring matrices are used in conjunction with local alignments programs to detect regions of similarity and infer potential common ancestry between proteins. The usual scoring schemes derive from the implicit hypothesis that related proteins evolve from a common ancestor by the accumulation of point mutations and that amino acids tend to be progressively substituted by others with similar properties. However, other frequent single mutation events, like nucleotide insertion or deletion and gene inversion, change the translation reading frame and cause previously encoded amino acid sequences to become unrecognizable at once. Here, I derive five new types of scoring matrix, each capable of detecting a specific frame shift (deletion, insertion and inversion in 3 frames) and use them with a regular local alignments program to detect amino acid sequences that may have derived from alternative reading frames of the same nucleotide sequence. Frame shifts are inferred from the sole comparison of the protein sequences. The five scoring matrices were used with the BLASTP program to compare all the protein sequences in the Swissprot database. Surprisingly, the searches revealed hundreds of highly significant frame shift matches, of which many are likely to represent sequencing errors. Others provide some evidence that frame shift mutations might be used in protein evolution as a way to create new amino acid sequences from pre-existing coding regions.