Opened 4 years ago

Closed 4 years ago

#1186 closed task (fixed)

Implement plug-in for running Single Sample Predictor models

Reported by: Nicklas Nordborg Owned by: Nicklas Nordborg
Priority: major Milestone: Reggie v4.27
Component: net.sf.basedb.reggie Keywords:
Cc:

Description

To start testing the Single Sample Predictor integration with BASE a simple plug-in should be implemented.

It should be possible to select a number of raw bioassays (StringTie) and SSP models and have the plug-in run the R scripts to evaluate the selected models for all raw bioassays. Results can be imported back into BASE and saved as annotations associated with the raw bioassays.

Some kind of configuration is needed in order to map the outcome from a given model to an annotation type.

Change History (10)

comment:1 by Nicklas Nordborg, 4 years ago

In 5626:

References #1186: Implement plug-in for running Single Sample Predictor models

Added a configuration section to reggie-config.xml for mapping SSP models to annotation types. The SspModel class is used to read that information and provides methods for enumerating the models and check if the mapped annotation type exists.

Note that it is expected that all model .RData files can be found in given directory.

comment:2 by Nicklas Nordborg, 4 years ago

In 5627:

References #1186: Implement plug-in for running Single Sample Predictor models

Added wrapper class for starting the SSP script. A new run() method has been implemented which takes a single raw bioassay and a list of SspModel:s as parameters. The result is returned as a string by Result.getStdout() with one value per line in the same order as the original model list.

comment:3 by Nicklas Nordborg, 4 years ago

In 5628:

References #1186: Implement plug-in for running Single Sample Predictor models

Implemented the plug-in. It can be started from the list or single-item page of a raw bioassay. It will only accept raw bioassays of the StringTie raw data type.

The plug-in can be configured with a pre-selected set of SSP models (which can be changed when running the plug-in). It can be run in dry-mode (annotations are not saved) and can then produce a log-file which is a tab-separated file with the values produced by the selected models.

comment:4 by Nicklas Nordborg, 4 years ago

Status: newaccepted

comment:5 by Nicklas Nordborg, 4 years ago

In 5667:

References #1186: Implement plug-in for running Single Sample Predictor models

Add <> around all output values from the predictor models since the output may contain other R-typical messages (eg. [1] ...) which will cause the output data to become misaligned with the models when parsing the result.

comment:6 by Nicklas Nordborg, 4 years ago

In 5668:

References #1186: Implement plug-in for running Single Sample Predictor models

Re-factored the parsing of SSP results into the SspAnalysis.Result.getModelResults() method so that we don't have to repeat the same code again in the ScanBReportWorker implementation.

comment:7 by Nicklas Nordborg, 4 years ago

In 5737:

References #1186 and #1190. Disabling code related to the SSP plug-in and SCAN-B report. They will be postponed to a later release.

comment:8 by Nicklas Nordborg, 4 years ago

Milestone: Reggie v4.24Reggie v4.x

comment:9 by Nicklas Nordborg, 4 years ago

Milestone: Reggie v4.xReggie v4.27

The plug-in was included in Reggie 4.27.

comment:10 by Nicklas Nordborg, 4 years ago

Resolution: fixed
Status: acceptedclosed
Note: See TracTickets for help on using tickets.