pyRforest: a comprehensive R package for genomic data analysis featuring scikit-learn Random Forests in R.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomic Data Analysis

Background:

• Random Forest models are crucial for analyzing complex genomic data, revealing non-linear and interactive feature effects.
• Python offers computationally efficient Random Forest implementations, while R is preferred by biologists for integrated statistical analysis and visualization.
• A gap exists in seamlessly combining the strengths of both R and Python for advanced genomic data analysis.

Purpose of the Study:

• To introduce pyRforest, an R package that bridges Python's scikit-learn Random Forest algorithms with the R environment.
• To provide biologists with an efficient tool for classification tasks on large genomic datasets, such as RNA-seq data.
• To enhance the interpretability and utility of Random Forest models in genomic research.

Main Methods:

• Integration of Python's RandomForestClassifier into R via the pyRforest package.
• Leveraging Python's efficient memory management and parallelization capabilities within the R environment.
• Implementation of novel rank-based permutation for biomarker identification and SHapley Additive exPlanations (SHAP) for feature interpretability.

Main Results:

• pyRforest enables efficient classification on large genomic datasets by combining R and Python.
• A novel rank-based permutation method allows for robust P-value estimation and visualization for feature significance.
• The package supports comprehensive downstream analysis, including gene ontology and pathway enrichment.

Conclusions:

• pyRforest effectively merges the computational strengths of Python with the analytical ecosystem of R for genomic data analysis.
• The package enhances the identification and interpretation of biomarkers using Random Forest models.
• pyRforest offers a unified platform for advanced genomic analysis, improving biological mechanism discovery.