Analysis of binding site conservation
In order to determine the number of binding sites gained, lost, or conserved between two sequences we first performed a pairwise alignment between two sequences using the R package Biostrings. The log-likelihood ratio (LLR) of binding was calculated at every position in each aligned sequence. Sequences were called binding sites if the LLR was greater than zero. In order to accommodate gaps in sequence alignments, sites were considered conserved if they aligned within 3 bp. Sites were considered lost if there was no site with LLR greater than 0 within 3 bp on the corresponding aligned sequence. For the background distribution we use the frequencies of nucleotides in the Drosophila genome (P bg(A)=P bg(T)=0.297, P bg(C)=P bg(G)=0.203).