Skip to main content
Fig. 4 | BMC Biology

Fig. 4

From: Systematic prediction of degrons and E3 ubiquitin ligase binding via deep learning

Fig. 4

E3 motifs calculation and evaluation. a Workflow of motifs calculation. 1. Summary of the ESI dataset collected from PubMed and related works. 2. Fifty-five E3s possessing at least ten substrates in the collected ESI dataset were selected for motif calculation. 3. Predicting Degpred degrons on substrates in the ESI dataset for each E3. 4. Degron alignment and outlier removal with GibbsCluster for each E3. 5. Aligned regions present on degrons were considered to represent the binding sites for each E3. 6. Motif calculation using aligned peptides for each E3. 7. Cut-off calculation, degrons with motif matching scores larger than the cut-off were predicted as binding degrons of the E3. b Comparison of our generated motifs of βTrCP, SPOP, and FZR1 with their experimentally verified motifs in the ELM database. ELM motifs are represented as regular expressions, and consistent regions are highlighted in red. c Distribution of normalized maximal motif matching scores for our ESIs, the ESIs collected in Ubibrowser2.0 and not in our ESIs, and random pairs. Kernel density estimate plot showing the distribution of scores. The maximal motif matching scores were normalized to Z-scores for each E3 respectively. d Comparison of Ub-site abundance change and protein abundance change of predicted SPOP substrates in three methods after SPOP overexpression. Abundances of replicate 1 in original paper [38] were compared, and the results of replicate 2 were identical (data not shown). ChenESI predicted 1195 SPOP substrates and 867 of them were detected in the SPOP overexpression MS; Ubibrowser2.0 predicted 892 SPOP substrates and 622 of them were detected in the MS; we selected the top 2000 predicted SPOP substrates and 669 of them were detected in the MS. Kernel density estimate plots showing the distribution of the ratios after and before SPOP overexpression

Back to article page