I recently submitted a sequence to run on I-TASSER (job ID S63581) and
have a couple of questions since I am a little confused about what I
see as I am not a bioinformaticist or a crystallographer. So please
pardon my stupid questions.
•Based on the sequence identity score, the program predicted a single
model instead of the Top 5 Models. The single model displayed (C-score
= 1.98) is the Streptococcus pneumonia TIGR4 LplA molecule or something else?
•Top 10 templates used by I-T are all based on Streptococcus pneumonia
sequences. Residues that are used for alignment with high confidence
are all in colored font.
•Top 10 proteins in PDB that are structurally closest to the I-T
model, shows that Spn TIGR4 LplA (1vqzA) has a TM score of 0.9833 with
an identity score of 0.459% and 0.99% coverage of the sequence. So the
submitted sequence is more closely related to Spn LplA.
•Other proteins have a TM score of 0.5801 (2po1A), 0.6158 (2artA
Thermoplasma acidophilum), 0.6456 (3a7rA E. coli). The structures of
2artA and 3a7rA in complexes with AMP + Mg + lipoic acid are different.
What is the explanation for that?
•The total # of ligands shown in the Jmol view for 1vqzA are too many
and none of these are either AMP or octanoic acid or lipoic acid. So
what is the identity of the many ligands shown in the structure?
•Predicted EC # shows that the submitted protein has an EC score of
1.8814 indicating a high degree of confidence that the query protein
is likely to have LplA-like activity.
•In the predicted binding site model (spinning structure with binding
residues identified), the top ranking protein is 3a7rA with a BS score
of 1.3622 indicating high degree of confidence with the prediction. Is
this the same model displayed in the Top 5 models (without the binding
residues) predicted by I-T?
My feeling is that it would certainly help to know which favored model
is being displayed in the TOP 5 models predicted by I-TASSER in the
results. I have also submitted some other LplA-like proteins and for a
couple of them, 5 models are displayed, albeit some of them have a
C-score of -1.xxx indicating that structures predicted with a negative
C-score are unlikely to be correct. But for job ID S63581, only one
model is predicted with a very high C-score.