Skip to content

Latest commit

 

History

History
230 lines (177 loc) · 7.66 KB

README.md

File metadata and controls

230 lines (177 loc) · 7.66 KB

SENDQSAR

SENDQSAR: A R Package for QSAR Modeling with SEND Database

About

  • This package facilitates developing Quantitative Structure-Activity Relationship (QSAR) models using the SEND database. It streamlines data acquisition, pre-processing, organ wise toxicity score calculation, descriptor calculation, and model evaluation, enabling researchers to efficiently explore molecular descriptors and create robust predictive models.
  • Detailed descriptions of each function are available in the “Articles” section of the GitHub-hosted website.

Features

  • Automated Data Processing: Simplifies data acquisition and pre-processing steps.
  • Comprehensive Analysis: Provides z-score calculations for various parameters such as body weight, liver-to-body weight ratio, and laboratory tests.
  • Machine Learning Integration: Supports classification modeling, hyperparameter tuning, and performance evaluation.
  • Visualization Tools: Includes but not limited to histograms, bar plots, and AUC curves for better data interpretation.

Workflow

  • Input Database Path: Provide the path for database or .xpt files containing nonclinical study data in SEND format.
  • Data Pre processing: Use functions f1 to f8 to clean, harmonize, and prepare data for Machine Learning (ML).
  • Model Building: Employ ML functions (f9 to f18) for ML model training and evaluation.
  • Visualization: Generate plots and performance metrics for better interpretation (f12 to f15).
  • Automated Pipelines: Use functions f15 to f18 to perform the above workflows in A single step by providing the database path and a .csv file containing the label (TOXIC/NON-TOXIC) of the STUDYID.

Modular Functions Overview

  • Liver Toxicity Score Calculation for Individual STUDYID :

    • f1: get_compile_data - Fetches structured data from the specified database path.
    • f2: get_bw_score - Calculates body weight z-scores for each animal (depends on f1).
    • f3: get_livertobw_zscore - Computes liver-to-body weight z-scores(depends on f1).
    • f4: get_lb_score - Calculates z-scores for laboratory test results(depends on f1).
    • f5: get_mi_score - Computes z-scores for microscopic findings(depends on f1).

  • Liver Toxicity Score Calculation and Aggregation for Multiple STUDYID:

    • f6: get_liver_om_lb_mi_tox_score_list
      • Combines z-scores for LB, MI, and liver-to-BW ratio into a single data frame.
      • Internally calls f1 to f5.

  • Machine Learning Data Preparation:

    • f7: get_col_harmonized_scores_df - Harmonizes column names across columns for consistency from the data frame (depends on f6).

    • f8: get_ml_data_and_tuned_hyperparameters - Prepares data and tunes hyper parameters for machine learning (depends on f7).

  • Machine Learning Model Building and Performance Evaluation:

    • Model Training
      • f9: get_rf_model_with_cv
        • Builds a random forest model with cross-validation (depends on f8).
    • Improved Classification Accuracy
      • f10: get_zone_exclusioned_rf_model_with_cv
        • Enhances classification accuracy by excluding uncertain predictions (depends on f8).
    • Feature Importance
      • f11: get_imp_features_from_rf_model_with_cv
        • Computes feature importance for model interpretation.
    • Model Performance Visualization
      • f12: get_auc_curve_with_rf_model
        • Generates AUC curves to evaluate model performance.

Notes for MOdular Functions

  • Data Preparation
    • Functions f1 to f8 must be executed sequentially to prepare the Data argument required by these functions.
    • Alternatively, the composite function f18 can be used to directly generate the Data argument, combining the functionality of f1 to f8.
    • For f9, f10, f11, and f12, Functions f1, f2, f3, f4, f5, f6, f7, and f8must be executed sequentially to prepare the Data argument. Alternatively, the composite function f18 can be used to directly generate the Data argument.

Composite Functions Overview

  • Combine multiple modular functions for complex operations.

  • Visualization and Reporting :

    • f13: get_histogram_barplot - Creates bar plots for target variable classes (depends on functions f1 to f8).
    • f14: get_reprtree_from_rf_model - Builds representative decision trees (depends on functions f1 to f8)..
    • f15: get_prediction_plot - Visualizes prediction probabilities with histograms(depends on functions f1 to f8)..

Automated Pipelines

  • f16: get_Data_formatted_for_ml_and_best.m
    • Creates machine learning-ready data by executing f1 to f8 -Formats data for ML pipelines.
    • Provide the same result as f8 by merging functionality of functions from f1 to f7
  • f17: get_rf_input_param_list_output_cv_imp
    • Automates pre-processing, modeling, and evaluation.
    • Provide the same result as f9 by merging functionality of functions from f1 to f8
  • f18: get_zone_exclusioned_rf_model_cv_imp
    • Similar to f17 but excludes uncertain predictions.
    • Provide the same result as f10 by merging functionality of functions from f1 to f8
    • Optional argument for hyperparameter tuning.

Helper Functions

  • h1: get_treatment_group_&_dose - Retrieve treatment groups from the tx domain.
  • h2: -get_repeat_dose_parallel_studyids
    • Retrieves STUDYIDs for repeat dose and parallel study designs.
    • Optional filtering for “rat” species studies.

Functions in Development

  • fid1: get_all_LB_TESTCD_score - Calculates scores for each LBTESTCD based on get_lb_score.
  • fid2: get_indiv_score_om_lb_mi_domain_df - Returns domain-specific scores including liver-to-BW ratio, LB, and MI scores.

Dependencies

  • randomForest
  • ROCR
  • ggplot2
  • reprtree

Installation

# Install from GitHub
devtools::install_github("aminuldu07/SENDQSAR")

Examples

Example 1: Basic Data Compilation

library(SENDQSAR)
data <- get_compile_data("/path/to/database")

Example 2: Z-Score Calculation

bw_scores <- get_bw_score(data)
liver_scores <- get_livertobw_zscore(data)

Example 3: Machine Learning Model

model <- get_rf_model_with_cv(data, n_repeats=10)
print(model$confusion_matrix)

Example 4: Visualization

get_histogram_barplot(data, target_col="target_variable")

Contribution

Contributions are welcome! Feel free to submit issues or pull requests via GitHub.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

For more information, visit the project GitHub Page or contact [email protected].