starcalibre / microscopium-ui

Microscopium web interface.

Microscopium User Interface

The web interface for Microscopium.

Prerequisites

• Python 3.4
• NPM package manager
• Bower package manager
• MongoDB version > 2.6
• A modern browser. The interface is tested as working on latest Chrome, Firefox, Edge and IE11. Due to the heavy use of the HTML5Canvas API, performance is best on Chrome.

Install Instructions (Development)

Install Python package dependencies.

$ pip install -r path/to/requirements.txt

Install JavaScript dependencies using NPM and Bower.

$ npm install
$ bower install

Run grunt. The grunt tasks browserifies, lints and tests the JavaScript code upon any changes.

$ grunt

If you don't have the grunt command in your path, you can install it with:

$ sudo npm install -g grunt-cli

Run the Flask server.

$ python manage.py runserver -r

The interface will now be available in localhost:5000. The port can be changed using the -p flag when using runserver, e.ge

$ python manage.py runserver -r -p 8080

MongoDB Schema

Documents should by default be stored in the the database microscopium, however this can be changed by updating the config.py file.

Document Schemas

While MongoDB does not enforce a schema for data, certain fields must be present in each document to ensure proper operation of the UI.

Screen descriptions are saved in the screens collection with the following format:

var Screen = {
    _id: String, // A short ID for the screen.
    screen_desc: String, // A description of the screen.
    number_samples: Number, // The number of
    screen_features: Array, // An array of strings labelling each feature used in
                           // the screen.
    available_overlays: Array // An optional field. An array of additional overlay
                              // information available for samples in this screen.
                              // e.g. if samples have ``screen_run`` and ``z_score``
                              // overlay information, this array should be 
                              // ["screen_run", "z_score"]
}

Each sample saved in document with format:

var Sample = {
        _id: String,  // A unique identifier for each sample. Currently takes
                      // screen_id-plate-well
        screen: String  // The screen_id for the sample. An index should be
                        // setup on this field.
        gene_name: String,  // The name of the gene silenced or treatment
                            // applied to the sample.
        feature_vector: Array,  // An array of Numbers, the vector of
                                // unstandardised features
        feature_vector_std: Array  // An array of Numbers, the vector of standardised
                                   // features.
        neighbours: Array  // An array of sample_id strings, the 25 nearest
                           // neighbours of the sample.
        column: String,  // The column the sample is on in the plate
                         // e.g. '01', '02', '03', etc.
        row: String, // The row the sample is on in the plate
                     // e.g. 'A', 'B', 'C', etc.
        dimension_reduce: {
            pca: Array // the co-ordinates for the sample's point after PCA dimension reduction
            tsne: Array // the co-ordinates for the sample's point after TSNE dimension reduction
        },
        overlays: Object // An object of the additional screen overlay information
                         // for this sample. This information is optional. If no 
                         // overlay information is available, the corresponding
                         // sample document should no, or an empty, ``available_overlays``
                         // field. Currently supports numeric and Date data.
    }

Each image saved in document with format:

var Image = {
        _id: ObjectID, // Mongo generated unique ID for this image
        sample_id: String, // The _id of the sample corresponding to this image.
        image_full: String, // The fullsize image.
        image_large: String, // A large preview image, 512px wide.
        image_thumb: String // A smalller, thumbnail sized preview image, 150px wide.
    }

For each feature belonging to each screen, a document is saved detailing the feature across the whole dataset. Think of this is the column axis if you were to look at the dataset as a dataframe or table.

var Feature = {
        _id: ObjectID, // Mongo generated unique ID for this feature.
        screen: String,
        feature_name: String, // the actual name of the feature
        feature_dist: Array, // A vector of unstandardised values for this feature.
        feature_dist_std: Array, // A vector of standardised values for this feature.
}

Indices

By default, MongoDB only creates indices on the primary key for each document. The UI makes queries to a few non-primary keys, so indices need to be created on these. From your Mongo terminal, load up the microscopium database and enter these commands to ensure you've got the right indices.

db.samples.createIndex({'screen':1});
db.images.createIndex({'sample_id':1});
db.features.createIndex({'screen':1, 'feature_name':1});