data:
  resource: data/ChEMBL_IC50_CYP3A4_CHEMBL340_aggregated.parquet
  type: intake
  input_col: OPENADMET_CANONICAL_SMILES
  target_cols:
  - pchembl_value_mean
metadata:
  authors: Devany West
  email: devany.west@omsf.io
  biotargets:
  - CYP3A4
  build_number: 0
  description: predicting CYP3A4 IC50 with lgbm model
  driver: sklearn
  name: lgbm
  tag: openadmet-IC50
  tags:
  - openadmet
  - release
  - IC50
  version: v1
procedure:
  feat:
    type: FingerprintFeaturizer
    params:
      fp_type: "ecfp:4"
  model:
    type: LGBMRegressorModel
    
  split:
    type: ShuffleSplitter
    params:
      random_state: 42
      train_size: 0.8
      val_size: 0.0
      test_size: 0.2
  train:
    type: SKLearnBasicTrainer 
report:
  eval:
  - type: RegressionMetrics
    params:
      axes_labels:
      - True IC50
      - Predicted IC50
      max_val: 10
      min_val: 3
      pXC50: true
      title: CYP3A4 True vs Predicted IC50
  - type: SKLearnRepeatedKFoldCrossValidation
    params:
      axes_labels:
      - True IC50
      - Predicted IC50
      max_val: 10
      min_val: 3
      n_repeats: 5
      n_splits: 5
      pXC50: true
      random_state: 42
      title: CYP3A4 True vs Predicted IC50