Fig. 1

Illustration of the three models used. A DNA sequence is shown with the complement sequence below. The blue histograms illustrate nucleotide probabilities. The central model with k=7 (upper left) predicts the base in the middle from the adjacent nucleotides in the boxes to the left and right. For this illustration, C has highest probability, which happens to coincide with the correct nucleotide at the position. The Markov model (top right) of order k=14 predicts a nucleotide from the previous 14. In this example A has highest probability although G is the actual reference probability. The bidirectional model (bottom right) use the same model on the reverse complement strand. In this example C has the highest probability, which coincides with the complement base at the position. The probabilities are translated to the direct strand and averaged with the forward model