DNA linear motif classes, how many? how different?

ARIEL APTEKMANN

Conferencia; 9CA2B2C; 2018

Sequence motifs are relatively short, recurring patterns.When found in DNA they are presumed to have a biological function. Some of themindicate sequence-specific binding sites for proteins, such as nucleases or transcriptionfactors (TF).Frequently conservation of a sequence implies a selective pressure, which in turnsuggests a function, although there are some motifs with no apparent functionality.In this work we study sequence motif databases by modelling sequence motifs as regularexpressions, which specify the length of the motif and which bases are allowed at eachmotif position.We develop a method for building a regular expression from position specific scoringmatrices.Using this representation we tackle:How many linear motif classes remain to be discovered in nature?How many classes coexist on a genome?How different the motifs on a genome are?As a measure of motif specificity for a pair of linear motif classes, we quantify how manymotif-discriminating positions prevent a subsequence from being an instance of the twoclasses at once.Naturally occurring pairs of DNA linear motif classes present most often one motif-discriminating position, which maximizes the potential number of coexisting linear motifclasses.Increasing the size of the alphabet by means modifications increases the potential numberof coexisting linear motif classes.We calculate the fraction of all possible protein subsequences that would belong to a linearmotif class if the potential number of coexisting linear motif classes came into actualexistence.This number is highest if the specificity requirement is no motif-discriminating positions.We propose that naturally occurring DNA linear motif classes operate under mild specificityrequirements that maximize the potential number of coexisting linear motif classes.