-
Notifications
You must be signed in to change notification settings - Fork 32
NightShift_Tutorial
Note to Editors: Note, this page belongs to the ballaxy server, do not remove!
Image(ballaxy_login_short.png)
The workflow NightShift can be accessed via the menu Shared Data -> Published Workflows in the top level menu.
Image(ballaxy_import_workflows_short.png)
A list will show all provided workflows.
Image(ballaxy_import_workflows2.png)
Click on the workflow of your choice (NightShift). The workflow is presented in more detail, i.e. all steps are shown.
Image(ballaxy_import_NightShift.png)
Click on Import Workflow to import.
To return to the main page by click on the ballaxy logo (upper left corner).
Now lets have a look on the workflow to find out what input files we have to prepare.
All workflows of your account can be accessed via the menu Workflow in the top level menu.
Image(ballaxy_access_worklow_menu.png)
Click on the combo box of the NightShift workflow and select Edit.
Image(ballaxy_access_NightShift_worklow_m.png)
The NightShift workflow is a combination of several tools that are represented by boxes:
- DownLoadNMRPredictDB,
- RestrictPDB2BMRBMapping,
- ComputeNMRDescriptors, and
- TrainNMRShiftModel.
Please note that Input dataset is not a tool, just a required input file.
The sequence of action is represented by the connections between the boxes.
Image(ballaxy_NightShift_worklow_view.png)
Each tool, i.e. box, has a defined input(s) and output(s), and the connections between the boxes show the data flow.
E.g. Input dataset is the only required input and refers to a feature specification file, that is needed by the tool ComputeNMRDescriptors.
This input can be created by a tool called Feature Generator, which we will do later.
The tool DownloadNMRPredictDB has no input, but produces a current BMRB-to-PDB entry matching file, that can be further restricted by the tool RestricPDB2BMRBMapping.
By clicking on individual tool-boxes, more information and details of that tool and predefined parameters are shown and can be edited. E.g. we can change the training size percentage or choose another Amber energy cutoff for filtering reasonable atom instances.
Image(ballaxy_NightShift_worklow_edit.png)
For now, lets keep the standard values.
The NightShift workflow needs a set of pdb and bmrb files to train a shift prediction model and a list of features used as predictors for the model. While the pdb and bmrb files are downloaded automatically from <www.pdb.org> and <www.bmrb.wisc.edu>, the user can manually define the features used for training. To this end we use the tool Feature Generator.
Click on Analyse Data in the top level menu.
Image(ballaxy_NMR_tools_closed.png)
Then click on BALL/NMR in the Tools bar on the left hand side.
Image(ballaxy_NMR_tools_open.png)
Select the FeatureGenerator tool
Image(ballaxy_NMR_tools_FeatureGenerator.png)
This tool offers all pre-implemented features by checkboxes, e.g.
- semi classical contributions like ring current (Haigh1972), electric field effect (Buckingham1960), hydrogen bonding (Neal2003), random coil (Wishart1995)
- sequence based features like aa type, align score, seq length, …
- structur based descriptors like psi, phi, chi angles, surface contributions, rotamers, atomic density, S-S-bonds, H-bonds,..
- and force field based features like Amber energies and Amber atomtypes
Deselecting a box leads to exclusion of that feature as predictor.
Image(ballaxy_NMR_tool_FeatureGenerator_footer.png)
Confirm your choice by click on the Execute button.
A feature specification file is created and added to your history (right hand side bar).
Image(ballaxy_NMR_history_FeatureFile.png)
We can now proceed to run the NightShift workflow.
Access the NightShift workflow via the menu Workflow in the top level menu.
Image(ballaxy_access_worklow_menu.png)
Click on the combo box of the NightShift workflow and select Run.
Image(ballaxy_NightShift_worklow_run.png)
All steps of the workflow are shown and the required input data can be specified.
Image(ballaxy_NightShift_enter_input.png)
In our case, we only have to specify the feature description file. Select your feature description file from the combo box.
If you are interested in more details or want to change the default parameters, click on each step to show additional information, e.g. PISCES cut off value.
Image(ballaxy_NightShift_training_details.png)
Press Run workflow to start NightShift.
Now, we have to wait quite some time (2-5 days) depending on the size of your training set.
The results of all intermediate steps and the resulting model (Rdata) as as well as an evaluation summary (in pdf format) will be added to your history.
Image(ballaxy_NightShift_pipeline_result.png)
For chemical shift prediction we offer two tools:
- the PredictNMRShifts tool for application of user created prediction models and
- the Spinster tool for applying our pre-computed model
Image(ballaxy_NMR_prediction_tools.png)
In the following we focus on PredictNMRShifts since it allows to apply the user generated training model. (Spinster uses our pre-trained model)
First the user has to upload the pdb file for which a shift prediction has to be made.
Select the Get data - _ Upload File_ tool from the Tools bar on the left hand side.
Image(ballaxy_NightShift_protein_upload.png)
Specify the path to your pdb file on your file system and press Execute.
Image(ballaxy_NightShift_protein_upload2.png)
The pdb file will be uploaded and add to your history.
Image(ballaxy_NightShift_new_pdb.png)
The new pdb file is added to the history (right hand side).
Select the PredictNMRShifts tool from the Tools bar on the left hand side. (If you have no model created by yourself, use our Spinster tool).
Image(ballaxy_NightShift_run_prediction.png)
The tool expects three inputs:
- select the pdb file of your choice from the combo box input pdb-file, specify the chain, for which a prediction has to be made
- select the feature specification file with which you trained your model from the combo box feature specification file
- the model from the combo box Prediction Model
Image(ballaxy_NightShift_run_prediction2.png)
Press Execute to run the prediction.
The resulting shift file (in csv format) will be added to your history.