Statistics in molecular biology: An example from detection of chimeric 16S rRNA artifacts

Statistical methods have had wide application in molecular biology. Genetic mapping, physical mapping, DNA sequence determination, evolutionary history reconstructions , sequence alignments, and sequence database searches all have substantial statistical components. In this talk we consider 16S rRNA sequences that have been generated during PCR ampliication from mixed populations. 16S ribosomal RNA (16S rRNA) have been selected to determine the phylogeny of orgainsms. It has the advantages of universality, reasonable size (about 1400 bases), and slow evolution, the latter imposed by its extensive secondary structure that maintains contacts between widely separated elements in the primary sequence. Some of data we consider comes from samples drawn from hot springs. Chimeric sequences have been found from these PCR ampliications; that is, part of the sequence is from one species, while another part of the sequence is from some entirely separate species. Clearly, such sequences will cause signiicant problems for researchers , since they will introduce errors in phylogenies. The statistical problem is to detect a change-point between the two sequences forming the chimeria, if there is a change-point. The method uses dynamic programming to generate an optimal chimeric alignment, deened as the highest scoring alignment between a query and a concatenation of a 5 0 and a 3 0 segment from two separate entries from a database of related sequences. Chimeras are detected by studying the score distributions and the form of the chimeric and global sequence alignments.