Figure 2

Metamotif estimation from simulated motif data. Metamotifs estimated with the metamotif nested sampler algorithm with varying relative frequency of metamotif samples. The top row in each metamotif alignment contains the "correct" metamotif that was sampled to the input weight matrix data in six different relative frequencies: 0.0, 0.1, 0.3, 0.5, 0.7, 0.9, 1.0. Frequency 0.0 which is shown in the bottom of the graph refers to a control experiment where all columns of the motif set are samples from the background (a Dirichlet distribution with parameters α = {0.5, 0.5, 0.5, 0.5}). A Cartesian-like distance between the sampled metamotif column mean nucleotide weights of the shown metamotif and the spiked metamotif mean nucleotide weights is presented above the relative frequency. An empirical p-value as described by [3] is also shown for the Cartesian distances (100,000 shuffles made for each motif).