| The alignment of multiple sequences of divergent proteins with an average identity less than 20% remains an unsolved challenging problem. Currently available multiple alignment algorithms in the public domain often do not accurately align divergent proteins even though structural comparisons have shown the existence of a correct alignment. In this study, known physical and chemical properties of amino acids were used to convert protein sequences from an alphabetic symbolic representation into a numerical representation for the easy application of fast numerical algorithms such as the Fast Fourier Transform (FFT) correlation to significantly reduce the overall search space for multiple sequence alignment while retaining a high level of sensitivity in detecting true motifs critical for the alignment. Additional numerical and pattern recognition procedures were also used to further refine the results for a final set of candidate alignments.; As a result of this study, a multiple alignment algorithm, DIVergent Alignment (DIVA), was developed. Using an independent test data set, DIVA identified 56% of the correct alignments, which is an improvement over the alignment procedures currently available. Unlike the publicly available methods, the current implementation of DIVA does not return a typical final multiple alignment. A list of alignments is returned instead. |
