OME2021 meeting workshop

[will-moore]

Somewhere to record prep notes, scripts etc.

Import and prep steps:

Some of these steps use proposed features in https://github.com/ome/training-scripts/pull/88. (Need to use that PR):

• Import idr0079-ExperimentA images. See https://docs.google.com/document/d/18cLSvUKVn8jEp7KSW7e48NvuxHT_mp3tyh98loTTOYY/edit
• Add 5* rating to 1 image each from membranes_cisgolgi, membranes_nuclei, membranes_recendo.
• Set rendering settings: Channel-1: Green, Channel-2: Red. Maybe adjust levels too.
• Apply Map Annotations from OMERO: $ python idr_get_map_annotations.py username password Project:1102 Project:1301 --server localhost
• Use the Channels KVP to set channel names AND add other Map Annotations we can use for labels: $ python channel_names_from_maps.py username password 1301 --server localhost --use_stain --add_map_anns
• Set Pixels Size (estimated): $ python calibrate_images.py password Project:1301 0.09x0.19 --user username --server localhost
• For chosen images, copy Masks from IDR images to create Polygons gist copy_masks_2_polygons.py: $ python copy_masks_2_polygons.py username password localhost 9837230 5952

idr0079 data

• Extra scripts for using analysis results from idr0079. Need to git clone git@github.com:IDR/idr0079-hartmann-lateralline.git to get the csv files. Then the following scripts are in my branch of that repo:
• In order to use OMERO.parade with these images, need to create OMERO.table on the parent project, using csv_to_table_on_project.py
• To create OMERO.table on each Image, with ROIs and Shapes, use csv_to_roi_table.py

Current figure proposal:

OMERO.parade

• Plot Raw_X_range vv Raw_Y_range to find long thin Zebrafish embryos

Analysis

• If Images have OMERO.tables linked to them (created above) and they have Shapes from OMERO (have OMERO shape IDs) then we can use JavaScript to load OMERO.tables value and set values on the Shapes.
• For example, this script loads the "Sphericity" value for each ROI, and colour: White (high Spericity) -> Red (low).

async function shapeData(panel) {
    const shapeIds = panel.get("shapes").map(s => s.id).filter(id => id > 0);
    console.log('shapeIds', shapeIds);
    let vals_by_shape = {};
    for (let i=0; i<shapeIds.length; i++) {
        // Load one at a time - more reliable
        let url = `/webgateway/table/Image/${panel.get('imageId')}/query/?query=Shape-${shapeIds[i]}`;
        let r = await fetch(url).then(rsp => rsp.json());
        let colIndex = r.data?.columns?.indexOf("Centroids_RAW_X");
        let shapeIndex = r.data?.columns?.indexOf("Shape");
        if (colIndex && shapeIndex && r.data?.rows.length > 0) {
            console.log("Value", r.data.rows[0][colIndex]);
            vals_by_shape[r.data.rows[0][shapeIndex]] = r.data.rows[0][colIndex];
        }
    };
    // Once all loaded, we can calculate range and assign colours to shapes
    const values = Object.values(vals_by_shape);
    let minVal = Math.min(...values);
    let valRange = Math.max(...values) - minVal;
    console.log('min, range', minVal, valRange);
    const new_shapes = panel.get("shapes").map(shape => {
        // hide any shapes we don't have data for
        if (!vals_by_shape[shape.id]) return { ...shape, strokeWidth: 0.01};
        let value = (vals_by_shape[shape.id] - minVal) / valRange;
        let red = parseInt(value * 255).toString(16);
        let blue = (255 - parseInt(value * 255)).toString(16);
        red = red.length == 1 ? `0` + red : red;
        blue = blue.length == 1 ? `0` + blue : blue;
        return { ...shape, strokeColor: `#${red}00${blue}`, strokeWidth: 5 }
    });
    panel.set('shapes', new_shapes);
}
figureModel.getSelected().forEach(shapeData);

[pwalczysko]

@will-moore So, youir list of the idr studies here is

idr0028 idr0089 idr0086 idr0050 idr0079

5 studies to consider. Please correct me if the list is wrong. We can discuss the details about how to minimize the effort by importing only parts of the studies into training servers later, but as discussed with @sbesson , the plan would be to have all these studies mounted on the volume which is visible from the training servers, and implement this logic into the ansible playbook. This will give us more flexibility at such time as new studies are asked for.

[will-moore]

Hi Petr, Sorry, that list is just ones I was thinking of using (just personal notes). But so far I only know that I want to use images from idr0079 (experimentA). I need to try a run-through and think if I have time for any more different images, but going on last year's workshop, 1 main set of images will be enough. Thanks.