Examples

[t-kimber]

The aim of this PR is to have scripts that are running for featurization schemes and associated machine learning models on the latest ChEMBL (v.28) data set.

Featurizers

Ligand-based

To run these featurizers, type:
(experiments-binding-affinity) $ bash tests/test_featurization_ligand_only.sh

• Morgan length 1024, radius 2
• Molecular graph
• One-hot SMILES

Kinase-informed

Using the full kinase sequence

To run these featurizers, type:
(experiments-binding-affinity) $ bash tests/test_featurization_kinase_informed.sh

• morgan + hash
• morgan + composition full sequence
• smiles + sequence full sequence

Models

Ligand-based

To run these models, type:
(experiments-binding-affinity) $ bash tests/test_model_ligand_only.sh

• Morgan & Fully connected neural network
• One-hot SMILES & Convolutional neural network
• Graph & Graph neural network

Kinase-informed

To run these models, type:
(experiments-binding-affinity) $ bash tests/test_model_kinase_informed.sh

• morgan + hash & Fully connected neural network
• morgan + composition & Fully connected neural network
• smiles + sequence & Convolutional neural network

⚠️ If you get FileExistsError, use the flag --overwrite.

Status

• Featurization Running

• Models Running

• Ready to go

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[schallerdavid]

Think about switching from ChEMBL subset to Kinase subset, e.g. ABL1+EGFR+AURA.

[schallerdavid]

Merge before fixing everything and create separate PRs afterwards.