rasar

The components of this module can be used for read-across related calculations.
It is a crucial module for cheminformatics applications.

🚀 Installation

pip install rasar

📌 Usage

This module supports five different read-across tasks:

• Pairwise similarity calculation
• Read-across prediction
• Read-across optimization
• Read-across feature importance calculation
• RASAR descriptor calculation

🧪 Getting Started

import pandas as pd
from rasar import (
    ra_similarity,
    ra_pred,
    ra_optimization,
    ra_importance,
    calculate_descriptor
)

# Load data
tr = pd.read_excel('train.xlsx', index_col=0)
te = pd.read_excel('test.xlsx', index_col=0)

# Split features and target
xtr = tr.iloc[:, :-1]
ytr = tr.iloc[:, -1]
xte = te.iloc[:, :-1]
yte = te.iloc[:, -1]

# Similarity calculation
sim = ra_similarity(des_tr=xtr, des_te=xte)
sim1 = sim.similarity_calculation(method='Euclidean Distance')

# Prediction
pred = ra_pred(df1=tr, df2=te).weighted_prediction(
    method='Laplacian Kernel',
    ctc=6,
    gamma=0.5
)

# Optimization
opt = ra_optimization(
    method='Laplacian Kernel',
    data=tr,
    parameters={
        'CTC': [1, 3, 6, 10],
        'Gamma': [0.1, 0.5, 1],
        'Threshold': [0.0]
    },
    objective_function="MAE",
    cv_fold=5
)

# Feature importance
imp = ra_importance(df1=tr).imp_calculation(
    method='Laplacian Kernel',
    ctc=6,
    gamma=0.5,
    ths=2
)

ra_importance(df1=tr).plot_importance(
    imp_df=imp,
    plot_type='coefficient',
    color="winter_r",
    index=1
)

# Descriptor calculation
des_tr, des_te = calculate_descriptor(
    df1=tr,
    df2=te,
    method='Laplacian Kernel',
    ctc=6,
    gamma=0.5,
    merge=True
)

📖 Citation

If you use this module, please cite:

Pore, S. and Roy, K., 2025.
“intelligent Read Across (iRA)” — A tool for read-across-based toxicity prediction of nanoparticles.
Computational and Structural Biotechnology Journal.
https://doi.org/10.1016/j.csbj.2025.07.032

📜 License

Apache License 2.0