UI to run models

[vedina]

Moved from https://github.com/ideaconsult/Toxtree.js/issues/68 and https://github.com/ideaconsult/Toxtree.js/issues/110

[vedina]

See T171

[thejonan]

There are two more skills added to jToxKit: TaskPollng and ModelRunning - they work together to make the prediction running an easy task. (Check this commit: 504c6afc372dcc44677064e5a9ee7d9f0b029425)

If they are combined with the Running widget which automates the process of attaching a method invocations to UI element, the entire wiring to enable running predictions on datasets is very easy. There is an example in model_demo.html and it goes like this:

  <input type="text" placeholder="Dataset URI_" id="dataset-uri" size="100"/><br/>
  <input type="text" placeholder="Model URI_" id="model-uri" size="100"/><br/>
  <input type="button" value="Run" id="run-prediction"/>&nbsp;<span id="algorithms-count">0</span><br/>
  <textarea id="result-area" placeholder="Results_" cols="100" rows="20"></textarea>

<!-- referred scripts, including jtox-kit.js -->
  <script type="text/javascript">
    var Manager = new (a$(Solr.Management, Solr.QueryingURL, Solr.Configuring))({ 
          serverUrl: "https://apps.ideaconsult.net/data/",
          connector: $
        }),
        Runner = new (a$(jT.ModelRunning, jT.TaskPolling, jT.Running))({ 
          /* Running skill settings */
          runSelector: '#run-prediction',
          runMethod: "onRun",
          target: document,

          /* ModelRunning settings */
          algorithms: true, 
          listFilter: "ToxTree", 
          loadOnInit: true, id: 'toxtree',
//           forceCreate: true,
          onRun: function (e) {
            this.runPrediction(datasetUri, $('#model-uri').val(), function (results) {
              $('#result-area').html(JSON.stringify(results, null, 2));
            });
          },
          onLoaded: function (result) {
            var self = this;

            $('#algorithms-count').html(result.algorithm.length);

            self.getModel(result.algorithm[0].uri, function (modelUri) {
              $('#model-uri').val(modelUri);
            });
          }
        });

    Manager.addListeners(Runner);
    Manager.init();

Pay attention to the fact that this is an exemplary Manager object, and that it also refers to AMBIT server.

[thejonan]

The challenging part could be matching the proper column for each predicted feature.

This is actually easy, the features should have sameAs label which is the same as the endpoint category (which defines the column of the read across matrix)

[thejonan]

I've added some version of this functionality. It can be found on this commit: 33e6f9fe247c399e407dd407236058766937406a of the old repo (branch: feature/ui-model-running). It does the same:

• Adds an additional tab Models next to Export. In there are listed all models found on the server. (Yes, the jToxComound extraTabs feature is added).
• Fills this tab as the normal jToxModel widget does - i.e. with quite a good information and leaving a checkbox in front.
• Adds a button after the whole list, which runs selected models (actually - selection is on algorithms).
• After all models ran finish, it issues a simple matrix reload (i.e. jToxCompund re-query), hoping that the newly available (due to initiated calculation) features will arrive and show up in the table.