Welcome to QSARtuna Documentation!

Welcome to QSARtuna: QSAR using Optimization for Hyperparameter Tuning.

Build QSAR models with hyperparameters optimized by Optuna.

Documentation:

• Intro and Quick Start
  • Background
  • JSON-based Command-line interface
  • Run from Python/Jupyter Notebook
  • Running via CLI
  • Optional: inspect
  • AutoML (Automatic Machine Learning)
  • Adding descriptors to QSARtuna
  • Adding machine learning algorithms to QSARtuna
• Jupyter Notebook: Preprocessing Data for QSARtuna
  • Introduction
  • Translate from SDF to CSV (if needed)
  • Deal with duplicates (Deduplication)
  • Splitting the data
  • Transform of data inputs
  • Log transform of user input data
  • Probabilistic Threshold Representation (PTR) and experimental error
• QSARtuna CLI Tutorial
  • This tutorial
  • Background
  • Preparation
  • Preprocessing: splitting data into train and test sets, and removing duplicates
  • Choosing scoring function
  • Advanced functoinaility: algorithms & runs
  • Probabilistic Random Forest (PRF)
  • ChemProp
  • Probability calibration (classification)
  • Uncertainty estimation
  • Explainability
  • Log transformation
  • Covariate modelling
  • Advanced options for QSARtuna runs
• AutoML (Automated model retraining)
  • Overview
  • Note on High-Performance Computing (HPC) Setup
  • Data extraction options
  • Walkthough running an AutoML pipeline
  • Schedule AutoML as a daemon for up-to-date models
• List of available ML algorithms
  • AdaBoostClassifier
  • Lasso
  • KNeighborsClassifier
  • KNeighborsRegressor
  • LogisticRegression
  • PLSRegression
  • RandomForestClassifier
  • RandomForestRegressor
  • Ridge
  • SVC
  • SVR
  • XGBRegressor
  • PRFClassifier
  • ChemPropRegressor
  • ChemPropClassifier
  • ChemPropHyperoptClassifier
  • ChemPropHyperoptRegressor
  • ChemPropRegressorPretrained
  • CalibratedClassifierCVWithVA
  • Mapie
• List of available molecular descriptors
  • Avalon
  • ECFP
  • ECFP_counts
  • PathFP
  • MACCS_keys
  • UnscaledPhyschemDescriptors
  • UnscaledJazzyDescriptors
  • UnscaledZScalesDescriptors
  • PhyschemDescriptors
  • JazzyDescriptors
  • PrecomputedDescriptorFromFile
  • ZScales
  • SmilesFromFile
  • SmilesAndSideInfoFromFile
  • ScaledDescriptor
  • CompositeDescriptor
• List of available evaluation splits
  • Random
  • Temporal
  • Stratified
  • Predefined
  • ScaffoldSplit
• List of available data transform
  • PTR
  • ModelDataTransform
  • VectorFromColumn
  • ZScales
• List of available deduplicators
  • KeepFirst
  • KeepLast
  • KeepRandom
  • KeepMin
  • KeepMax
  • KeepAvg
  • KeepMedian
  • KeepKeepAllNoDeduplication

Development

• Test report

• Test coverage

• Public release (3.1.2)