Consensus methods for DNA and protein sequence alignment.

Introduction The increasing body of nucleic acid sequence data has created interest among many scientists in computational approaches to macromolecular sequence analysis. Several international databases have been created in C141 order to store the data in a useful format, both for archival and analysis purposes.' Both DNA and protein sequences databases are maintained. The value of simply having easy access to all membrane protein sequences, for example, is not to be underestimated. The quantity of data has naturally led to the development of computer approaches to sequence analysis .* The purpose of this chapter is to present some of the tools that we have created in order to analyze multiple sequences in a rigorous, efficient, and systematic way. Much computer analysis of molecular sequences is directed toward discovery of biologically significant patterns. These patterns include ho-mologous genes, RNA secondary structure, tRNA or structural RNAs, palindromes in DNA sequences, regulatory patterns in promoter regions, and protein structural patterns. Once the patterns have been located they can often be tested by experiment, as in the case of promoter elements. Evolutionary relationships, however, cannot be directly tested, and increasing emphasis is being attached to the discovery and interpretation of sequence evolution. , Sequence alignment is a popular approach to pattern analysis.2 Computer alignments are often based on an explicit optimization function, rewarding matches and penalizing mismatches, insertions, and deletions. Sequence alignment often gives useful information about evolutionary or functional relationships between sequences. Our approach is based on what we refer to as consensus a n a l y ~ i s. ~-~ Consensus sequence analysis is usually performed by visual inspection of the sequences and by experiment. Of course, a protein binding site can only be verified by experiment, and analysis by " eye " can be biased. Thus, it is useful to have computer methods that can find consensus patterns best fitting explicitly stated criteria. Some algorithms have been developed along these line^,^-^ and they are described here, along with some biological examples. Our earlier methods applied only to DNA; here we also describe recent extensions to protein sequences. In 1970 Needleman and Wunsch5 published an approach sequence comparison (alignment) using a dynamic programming algorithm. Their algorithm find' maximum similarity between two sequences, where matches score positive weight and mismatches, insertions, and deletions 223 score nonpositive weight. Mathematicians began to attempt to define a distance between sequences and so to construct a …