GNN Based Drug Response Prediction GSoC 2022

View the Project on GitHub cannin/graph_neural_network_drug_response

7 August 2022

Week eight | PCA, visualization, and automatic data extraction

by Yoshitaka Inoue

GSoC Project

Mentor: Augustin Luna (@cannin)

Tasks

  1. Re-train DrugCell model.
    Currently working on HP Tuning but it takes around two weeks.

  2. weight visualization

    • Visualization
      • Graph Structure with PCA score
      • PCA for overall hiddens
      • PCA for each hidden layers

  3. Train Benchmark AutoML Model

    I run the model with only SMILES and Cell Line name as feature and it was 0.25 correlation. Then still running the model with morgan finger printing and cell2mutation data which is after using the PCA with 10 components.

  4. Understand RLIPP and implement (or use) it for describing the result with biological background

    Still running. This takes a week or more.

  5. Document Project Dependencies

    Working on this one right now. I organized some files for Pipfile and r-requirement. Also set github actions for data extraction.

Comments

Basically current tasks are really time consuming, so plan to work in parallel, running some script and writing documents.

Next Step

tags: gsoc