Automatic identification of group I intron cores in genomic DNA sequences

Automatic identification of the ribozyme core of group I catalytic introns in genomic sequences is shown to be feasible in spite of the scarcity of strictly conserved features in the sequence and secondary structure of group I introns. An algorithm is described that successfully identified 132 out of the 143 currently reported group I cores with a false positive rate of only 10(-6) per nucleotide. The recognition process consists in generating and rating large sets of potential local solutions which are gradually combined into more complex structures until an entire core (six to seven pairings, six connecting segments, three terminal loops) has been assembled. The extent to which successful recognition may be prevented by sequencing errors is assessed. Also discussed are (1) possible relationships between scores allocated by the program and ability to self-splice in vitro and (2) the potential for objectively assessing the degree of relatedness to group I of structures claimed to resemble group I introns.