EMnEM is a DNA motif finder that takes multiple alignments and a phylogenetic tree and identifies conserved patterns. 

Like modern phylogenetics methods and many motif finding methods it uses probabilistic models of DNA sequences and their evolution.

The key idea introduced in EMnEM was to use a different model of evolution, m,  for the motifs and for the background sequence, and to equate the probabilities in the motif matrix, f, to the prior probabilities of observing the ancestral bases in the motif evolution model, p(A|m).

It was described in this paper:
Moses AM, Chiang DY, Eisen MB: Phylogenetic Motif Detection by Expectation Maximization on Evolutionary Mixtures.
Pacific Symposium on Biocomputing 2004,9:324-335

download the software, source code and instructions.