Skip to content

Latest commit

 

History

History
172 lines (110 loc) · 6.14 KB

Handson.md

File metadata and controls

172 lines (110 loc) · 6.14 KB
Raw

4.1_writing_and_sharing_computational_analyses_in_jupyter_notebooks

Agenda

  1. Test your Conda Installation
  2. Introduction
  • Brief introduction to ML model used for this project
  • Sharing Computational Analyses in Jupyter Notebooks
  1. Run notebooks-sharing Conda Enviroment using the Expanse Portal
  2. Make a prediction with the ML model
  3. Share Notebooks in the public cloud using MyBinder

Test your Conda Installation

Run the galyleo script to create a URL for your Jupyter session.

galyleo.sh launch --account sds184 --reservation 'ciml-day3' --partition 'shared' --cpus-per-task 1 --memory-per-node 4 --time-limit 00:30:00 --jupyter 'lab' --notebook-dir "/home/${USER}" --conda-env 'notebooks-sharing'

After you run this command, a URL is printed. Copy this URL and paste it into a web browser to launch your interactive session.

If, and only if, the galyleo command fails, run the following commands to add a preinstalled notebooks-sharing conda environment from xdtr99 to your .bashrc file.

echo "source /home/xdtr99/miniconda3/etc/profile.d/conda.sh" >> ~/.bashrc 

source ~/.bashrc

Then, run the galyleo script again.

Run the Jupyter Notebooks from the Expanse Portal

1. Login to the Expanse Portal

Use your XSEDE username (trainXX) and password to login to the portal.

https://portal.expanse.sdsc.edu

2. Select Jupyter from the Interactive Apps menu on the Expanse Portal.

3. Specify your Job Parameters

Fill in the following parameters in the jobs parameter page. Be careful use exactly the same options as shown in the image below. Double-check all options before you click the submit button.

4. Monitor the Job Status

After you click submit, the next page shows a progress bar for your interactive session. Once all for stages are green, you will be directed to the next page.

5. Choose your Interactive Session

Once the interactive session is active, select the session you want to launch from the list of your currently running sessions. The latest session appears at the bottom of the list.

6. Run Jupyter Lab

Next, Jupyter Lab starts. Navigate to the notebooks-sharing/notebooks directory in the file browser.

Run the following notebooks:

  • 1-CreateDataset.ipynb
  • 2-CalculateFeatures.ipynb
  • 3-FitModel.ipynb
  • 4-Predict.ipynb

7. Make a prediction Predict the protein fold class for one of the following proteins using the protein sequence.

Structure 1

Source: PDB:7L6U

APADNAADARPVDVSVSIFINKIYGVNTLEQTYKVDGYIVAQWTGKPRKTPGDKPLIVENTQIERWINNGLWVPALEFINVVGSPDTGNKRLMLFPDGRVIYNARFLGSFSNDMDFRLFPFDRQQFVLELEPFSYNNQQLRFSDIQVYTENIDNEEIDEWWIRGKASTHISDIRYDHLSSVQPNQNEFSRITVRIDAVRNPSYYLWSFILPLGLIIAASWSVFWLESFSERLQTSFTLMLTVVAYAFYTSNILPRLPYTTVIDQMIIAGYGSIFAAILLIIFAHHRQANGVEDDLLIQRCRLAFPLGFLAIGCVLVIRGITL

Structure 2

Source: PDB:7B5J

TNKELQAIRKLLMLDVSEAAEHIGRVSARSWQYWESGRSAVPDDVEQEMLDLASVRIEMMSAIDKRLADGERPKLRFYNKLDEYLADNPDHNVIGWRLSQSVAALYYTEGHADLI

Structure 3

Source: PDB:7O3I

GARSSSYSGEYGSGGGKRFSHSGNQLDGPITALRVRVNTYYIVGLQVRYGKVWSDYVGGRNGDLEEIFLHPGESVIQVSGKYKWYLKKLVFVTDKGRYLSFGKDSGTSFNAVPLHPNTVLRFISGRSGSLIDAIGLHWDVYPSSCSRC

8. Use a different ML Model [optional]

Build a Logistic Regression or Neural Network model and rerun the prediction.

Share Notebooks using the MyBinder.org Service

Now, we want to host the notebooks-sharing repository on the MyBinder.org service, so the notebooks can be run anywhere, by anyone, in a web browser.

1. Create a fork of the notebooks-sharing Git repository

A fork is a copy of a repository in your GitHub account. Forking a repository allows you to freely experiment with changes without affecting the original project.

Log into your GitHub account and go to the repo: https://github.com/sdsc-hpc-training-org/notebooks-sharing

In the top-right corner of the notebooks-sharing GitHub page, click Fork.

2. Create a Binder Badge using MyBinder.org

Go to MyBinder.org and paste the GitHub URL of your forked Git repository into the form (and do not click the Launch button).

Expand the Binder Badge section at the bottom and copy the Markdown (first box).

2. Add a Binder Badge to the README file in the forked repository

Click the "pencil" to edit the README.md file

Paste the Markdown for the Binder Badge into the README.md file (find placeholder).

Then append ?urlpath=lab to the Binder Badge path. This instructs MyBinder to launch Jupyter Lab instead of Jupyter Notebook.

3. Add a Binder Badge for the 4-Predict.ipynb Notebook

Go back to the MyBinder form and add the path to notebook notebooks/4-Predict.ipynb.

Expand the Binder Badge section at the bottom and copy the Markdown (first box).

Paste the Markdown for the Binder Badge into the README.md file (find placeholder).

**4. Commit changes to your README.md file To save the README.md file, go to the bottom, enter a comment and click Commit changes.

5. Launch Jupyter from the Binder Badge On your README.md page, click on the first badge. The MyBinder.org service will launch. The first time the service is launched, it will create a Docker image from the environment.yml file. This can take several minutes. Once this step is complete, Jupyter Lab will launch in your web browser.

Now run the any of the notebooks to test them.

The next time you use this service, it will be faster since it uses the cached Docker image.