Factor-specific accuracy at increasing degrees of model complexity.

Accuracy of model's predictions at increasing degrees of model complexity. Shown are factor-specific correlations between the predicted binding landscape and measured occupancies for train- (left) and test set loci (right). Variations of the generalized hidden Markov model include (in increasing levels of complexity): (1) independent predictions per factor; (2) joint predictions (allowing for direct binding competition); (3) predictions at single-nucleus resolution; (4) with sequence-specific model of nucleosome binding; (5) with sequence-independent model of nucleosome binding; (6) with non-uniform prior on protein binding, based on DNase I hypersensitivity assay; (7) with cooperative binding interactions.