Analysis of regulatory sequences. A configuration of s nonoverlapping binding sites is given by the sequence of left initial positions r
= (r1,..., r
) (with rν+1- r
≥ ℓ for ν = 1, 2,..., s - 1). It can be associated with a path m(r) which takes the values m = 1 at the nucleotide positions of binding sites and m = 0 elsewhere. Dynamic programming algorithms based on a Bayesian model (27) of genomic sequences assign to each site configuration a probability of occurence ρ (r|a1,..., a
) for given sequence data a1,..., a
; see eq. (29).