Cloud based interactive analysis

Prerequisites
Before starting this lesson, you should be familiar with:

Basic computer usage

Learning Objectives
After completing this lesson, learners should be able to:

Understand interactive analysis platforms, including virtual desktops, Galaxy interactive tools, and Jupyter notebooks for bioimage analysis.

Run bioimage tools, such as CellPose on virtual desktops

Run Galaxy Interactive Tools for bioimage analysis

Motivation

Bioimage analysis requires interactive tools that combine computational power with user-friendly interfaces, enabling researchers to process and analyze complex image datasets efficiently.In this course, we introduce BARD Virtual Desktop, Galaxy interactive tools, providing hands-on experience in deploy bioimage tools to these platforms.

Concept map

graph TD T1("Bioimage interactive analysis") T2("Virtual Desktops")--> T1 T3("Galaxy Interactive Tools")--> T1 T4("Containerized GUI app")--> T1

Figure

Key components that enable efficient bioimage interactive analysis

Activities

Using bioimage interactive tools

Show activity for:

BARD Desktops
We will use the BARD desktop EMBL instance.

Access the BARD Desktop

Go to the BARD

Click Connect with EMBL Guest Account.

You will be given login details beforehand.

Username is trainingXX, replace XX with your own number.

Default password is tim2025.

After successful log in, you should see your desktop in your browser.

At the bottom of your desktop is the application dock. We keep multiple versions of software on BARD, but only the latest is shown in this dock.

To access older versions, click the application menu on the top right corner. Application -> Development

Each application is in its own container and independent of each other.

All applications start as the current logged-in user.

Your home directory is shared among all containers, which means if you need data accessible for all applications, you will need to store them in your own home folder.

Using microSAM on BARD

From the dock in your BARD desktop, click microSAM, since microSAM is based on Napari, a Napari GUI will appear.

Start Annotator 2D from Plugin -> Segment Anything for Microscopy -> Annotator 2D.

Go to File -> Open Sample -> Segment Anything for Microscopy -> LiveCell example data

Click Compute Embeddings from the right pane.

Click Automatic Segmentation button on the right.

For details on how to use microSAM, please refer to its Documentations

Segmentation using CellPose on BARD

Start Firefox from the bottom dock on your BARD desktop.

Go to the Bioimage archive AI Gallery and click on Analysed datasets.

Select the first dataset S-BIAD1026, and scroll down to the Model predictions section.

For illustration purpose, we will download the first image for 2D nucleus segmentation task.

Click Download under the Original image to download it. The image will be saved in /home/YOURNAME/Downloads

Click CellPose 1.1.1.1 (2D) to launch CellPose in 2D mode. A konsole terminal will open first,keep it open for logs and debugging.

In the CellPose window, go to File -> Load image, and select the downloaded image 5215c722-0e4d-58a9-a52b-478194c5bab0.png.

Click the run cyto3 button to start the segmentation. You can experiment with different parameters and pre-trained models using the CellPose GUI.

Galaxy Interactive Tools
We will use the Galaxy’s Training Infrastructure as a Service(TIaaS) in the Galaxy EU instance for this course.

Access the Galaxy Interactive Tool

Go to https://usegalaxy.eu/join-training/tim-bioimage-tools/ and join the training group. By joining the group, your Galaxy jobs are submitted to a special queue for fast execution.

You could also go directly go https://usegalaxy.eu without joing the group. This way your Galaxy job will be in the public queue.

Register an account in usegalaxy.eu .Some tools are only available for registered users.

Once you get to the Galaxy main page, on the left pane from the tool menu, scroll down and click the section Interactive tools. Here you will find all interactive tools available in Galaxy EU.

You could also search for a particular tool using the search box on top.

Upload image to Galaxy

Download an example image from here

Ensure you logged in to Galaxy

On the History pane, click the + sign to create a new history.

Click Upload at the top of the tool panel of the left of your screen.

Select Choose local files, and choose the dowloaded image.

Press Start, wait for the upload to finish, then close the window.

On the right hand side of your browser, in your History pane, wait for the image name to turn green.

Segmentation with CellPose Interactive tool

We will be using an image from Bioimage Archive AI Gallery.

Click Upload from the Galaxy tools menu.

Select Paste/Fetch data in the pop up window, paste the url https://www.ebi.ac.uk/bioimage-archive/galleries/assets/example-images/5215c722-0e4d-58a9-a52b-478194c5bab0.png, start the upload and close the window.

Click Tools again from the menu on the left.

Search CellPose in the search tools box, and click CellPose interactive

Select the uploaded image as input to the Images dropdown,and click Run Tool,

Monitor the blue info box on the next page, which will inform you when the Interactive Tool is accessible and provide you with a link to access it. This could take a while.

If you navigate away from this page, you can view your running Interactive Tools from the Interactive Tools menu item in the menu of the left.

Click the Open link to display CellPose in a new browser tab.

Once you are in CellPose, go to File->Load image, navigate to /home/cp_working_dir and select the image.

Click the run cyto3 button to start the segmentation. You can experiment with different parameters and pre-trained models using the CellPose GUI.

View image with the Napari interactive tool

Click Tools from the menu on the left.

Search napari in the search tools box, and click Run Napari interactive tool

In the Galaxy center pane, select the uploaded image from the Image dropdown, and click Run tool .

Monitor the blue info box on the next page, which will inform you when the Interactive Tool is accessible and provide you with a link to access it. This could take a while.

If you navigate away from this page, you can view your running Interactive Tools from the Interactive Tools menu item in the menu of the left.

Click the Open link to display napari in a new browser tab.

In the Napari window, go to File -> Open Files, go to the input folder, select the image and click open.

You should see your image is displayed in your Napari interactive tool.

To stop the Napari interactive tool, go back to the Galaxy browser tab, in your history, click the Delete button next to Run Napari on data 1.

Deploy bioimage tools containers

Show activity for:

BARD Desktops
In this section, we will demonstrate how to make a custom container available for interactive analysis, using the CellPose container as an example from the previous excercise. We have built and tested our CellPose container locally. To deploy it to BARD, we need to modify the Dockerfile by adding the necessary configurations.

Steps
Modify the Dockerfile

Open this Dockerfie

Have look at the added lines:

LABEL: Since BARD is a desktop environment, it needs to know where to display applications and how to launch them when users click on them. To achieve this, we embed the necessary functions using the LABEL instruction in the Dockerfile.

The oc.icon and oc.icondata labels tell BARD which logo image to use for the application.

The oc.icondata label requires a base64-encoded string of the image file(cellpose.svg). To generate this string, use cat cellpose.svg |base64

The oc.keyword label is used by the web application search engine.

The oc.cat label defines the category of the application in BARD. This helps organize applications within the BARD desktop environment.BARD has these categories ‘development’, ‘office’, ‘education’, ‘utilities’.

The oc.template label specifies the name of the template to use in BARD. It corresponds to the base image used in the FROM statement of the Dockerfile. If the value is incorrect, it won’t prevent the container from running, but the application may not function as expected in the BARD environment.

The oc.name label specifies the name of the application.

The oc.displayname label defines the name displayed in the BARD desktop.

The oc.launch label is a unique X11 window class name that BARD uses to identify the application window. To find out the name, run wmctrl -lx

The oc.path label specifies the path to the application’s binary executable inside the container. This tells BARD where to find and launch the application.

The oc.args label specifies any additional arguments that should be passed to the command defined in oc.path when launching the application.

The oc.app label must be always set to app

The oc.showinview label instructs BARD to place an icon of the application in its dock.

Make app icon images available. Line 66,67

RUN if [ -d /usr/share/icons ]; then cd /usr/share/icons; /composer/safelinks.sh; fi

RUN if [ -d /usr/share/pixmaps ]; then cd /usr/share/pixmaps; /composer/safelinks.sh; fi

ENV variables. Line 69-72

ENV APPNAME: The name of the application.

ENV APPBIN: The binary executable of the application. In the CellPose example, we launch CellPose within a Konsole terminal (/usr/bin/konsole) to monitor the output in real-time.

ENV APP: This is the same as APPBIN

ENV ARGS: This specifies the arguments for APPBIN.

To start container as the current logged-in user. Line 89-98

RUN mkdir -p /run/user && chmod 777 /run/user

RUN mkdir -p /etc/localaccount && \ for f in passwd shadow group gshadow ; do if [ -f /etc/$f ] ; then cp /etc/$f /etc/localaccount; rm -f /etc/$f; ln -s /etc/localaccount/$f /etc/$f; fi; done
CMD [“/composer/appli-docker-entrypoint.sh” ]

BARD Desktop allows users to log in via LDAP or OIDC, so the Docker container must run as the logged-in user. The commands above store the LDAP information and ensure that the logged-in user is available inside the container.
Deploy to BARD
As contributors:

Go to BARD containers repository

Fork the repository to your own user.

Create a branch for your container.

Open a pull request(PR).
As administrator (To deploy apps to your own BARD instance)
Build and test your container locally.

Push your container image to a registry, for example, dockerhub, quay.io etc.

Pull the image from the registry.
Extract the metadata from image using the below command, (replace with the actual image ID)
  $ docker inspect  CELLPOSE_IMAGE_ID > cellpose.json
In a terminal,
  $ curl -X PUT -H 'Content-type: text/javascript https://YOUR_BARD_URL/API/manager/image -d@cellpose.json

Galaxy Interactive Tools
In this activity, we will explain how to make your container available as an Interactive Tool in a public Galaxy instance. While building a Galaxy Interactive Tool requires some knowledge of XML and server administration, which are beyond the scope of this exercise, we will walk you through the essential steps to complete the process.

Things to know

Refer to Galaxy tutorial, if you would like to lear how to build and configure an interactive tool.

Galaxy EU instance (https://usegalaxy.eu)

Steps

Galaxy Interactive Tools require certain dependencies. To simplify the setup, we recommend using one of the base images available here. These images come pre-configured with all necessary dependencies, making the process more efficient and reducing setup complexity.

Have a look at this example, the first line uses one of the base image mentioned.

FROM jlesage/baseimage-gui:ubuntu-22.04-v4.5.2 AS build

Add the main installation steps to your own Dockerfile.

Create a startup script for the container startapp.sh. While not always required, it depends on how your Galaxy XML is structured..

Add the following lines to your Dockerfile

EXPOSE 5800

COPY startapp.sh /startapp.sh

RUN chmod +x /startapp.sh

ENV APP_NAME=”Napari”

ENV KEEP_APP_RUNNING=0

ENV TAKE_CONFIG_OWNERSHIP=1

WORKDIR /config

Build and test your Dockerfile locally. You could refer to this module for building containers.

Create a public repository for your docker recipe, e.g., Github(https://github.com), and make the repository public.

Push your image to a public container registry, such as quay.io

If you would like to make your interactive tool available in Galaxy EU, visit the Galaxy EU repository submit a pull request(PR).

Follow-up material

Recommended follow-up modules:

Cloud based batch analysis

Learn more:

BARD Desktop

Galaxy Interactive Tools