Fig 1.
Flowchart of the SmotifTF prediction algorithm.
Table 1.
GDT_TS values of top scoring models obtained with SmotifTF method using dynamic Smotif library generated at different e-value cutoffs.
Fig 2.
Performance evaluation in the training set.
Prediction quality (assessed as the mean GDT_TS of the top-scoring model against the native structure) is plotted on the X-axis for 20 cases at different e-value cutoffs used in generating the dynamic Smotif library. The data points for different e-value cutoffs are shown in different symbols (no cutoff (square), 10-10 (circle), 10-5 (triangle), 10-1 (star) and 100 (diamond)). The dual Y-axes correspond to the mean number of hits in the dynamic Smotif database (right axis, inversed scale, black data points) and to the mean e-value of the best hit in the dynamic database (left axis, log scale, red data points), respectively.
Table 2.
Performance of SmotifTF on the benchmarking test set in comparison to other methods
Fig 3.
Examples of SmotifTF predictions in the benchmark test set.
The structural superposition of the top-scoring model (pink cartoon) with the native structure (green cartoon) is shown in the middle. The proteins that provide the Smotif fragments to the top-scoring model are shown in grey cartoon, with the Smotif themselves colored according to the secondary structure elements present in them (helix = red, strand = yellow, loop = green). The PDB id, chain id and residue numbers of the Smotif fragments are shown along with the root mean square deviation (RMSD) of the respective Smotif fragments compared to the corresponding native Smotif. The SCOP ids of the proteins are provided, where available. (a) N-terminal domain of a protein with unknown function from Vibrio Cholerae (PDB: 4ro3A) (b) RNA binding protein Tho1 from Saccharomyces Cerevisiae (PDB: 4uzxA) (c) Mammalian Endoribonuclease Dicer (PDB: 4wyqA).