Figure 4From: Identifying protein complexes directly from high-throughput TAP data with Markov random fieldsCluster quality. The distribution of the quality of clusters as predicted by MRF. Note that negative values for the quality score indicate that a cluster is observed better supported by the data than expected. The zero-line indicates that the observed error corresponds to the expectation. (a) shows that most predicted clusters fit the model for the spoke model. Outliers with high error are points on the top of the figure; the clusters contain largely artifacts. (b): Also for the matrix model, MFA is robust against the high false negative rate. For the list of clusters, refer to the supplementary material.Back to article page