Help for using HNADOCK server


1. How to provide input for the molecules to be docked

The HNADOCK server is to predict the binding complex structure between two nucleic acid molecules through a hierarchical docking algorihtm of an FFT-based global search strategy and an intrinsic scoring function for nucleic acid interactions. Users are required to provide the three-dimensional (3D) structures of the two molecules to be docked. The HNADOCK server accepts three types of input for the molecules:

Only ONE type of input is needed for each molecule.

If more than one types of input are provided, the first one will be used. If the uploaded pdb file contains multiple models like the following,

MODEL  1
...
ENDMDL
MODEL  2
...
ENDMDL
...
only the first one will be used for docking. For the "PDB ID:ChainID" input, the user can provide one single chain ID or multiple chain IDs. For example, "1KD5:A" stands for the chain A of the pdb file of 1KD5; "1KD5:AB" stands for the chains A and B of the pdb file of 1KD5.

RNA structure prediction: If an RNA sequence is provided, the server will automatically constuct a model by using our RNA homology modeling protocol HMODELRNA if there is a good template in the PDB or our ab initio RNA structure prediction algorithm 3dRNA if no template is found. Users are recommended to submit their own pdb file if the RNA contains multiple chains, as our pipeline is currently designed to model single-chain RNAs.

Secondary Structure Prediction: By default, the server uses the "RNAfold" to predict the secondary structure of an RNA sequence for ab initio three-dimensional structural modeling if no template is found for the sequence. Users may also choose other secondary structure prediction method like Fold, MaxExpect, ProbKnot, and IPKnot, where Fold, MaxExpect, and ProbKnot are taken from the RNAstructure package.

RNA-RNA Interaction Prediction: By default, HNADOCK server builds the complex structures between two RNAs without a priori binding information through ab initio docking. However, the server also offers users an option to choose a method (RNAup or RactIP) for RNA-RNA interaction predictions. The predicted interaction information will be then used as inter-RNA distance constrains during docking.

Model refinement: The server offers users an option to refine the top 10 predicted models by a short MD simulation. If selecting "Yes", the job will take longer to finish, and the result package will also include the refined models named like "model_[1-10]_ref.pdb" in addition to the initially generated models.

2. How to specify the binding site [optional]

The HNADOCK performs global docking to predict the binding complexes between two molecules. Therefore, no information about the binding site is required for the docking job. However, the server also gives users options to specify the binding site residues if such information is available, so that the predicted models may get a higher accuracy. Two types of binding site information can be provided.

3. About the docking results

When a job is finished, the status web page will automatically show the docking results for download and visualization. The docking results include two types of files for download, the individual pdb files and the docked complex models. Users may also download all the results in a single package that includes the Receptor PDB file, Ligand PDB file, and the top 100 predictions.

As the top 10 binding models are normally deemed as the most important models in realistic macromolecular docking, like CAPRI (https://www.ebi.ac.uk/msd-srv/capri/), the result page also provides an interactive visualization of the top 10 models using the NGL viewer. Users can choose to view any of the top 10 models or all together by different colors, representations and styles.

The result page also gives a summary of the rankings and docking scores for the top 10 complex models, where the score is based our scoring function for RNA-RNA interactions DITScoreRR. However, it should be noted that the docking scores here do not reflect the true binding affinities, but a relative ranking of the complex models, as DITScoreRR was not calibrated with experimental binding data.

It is recommended that users download their docking results as soon as their job is done, as the job results will only be stored on our server for two weeks.