ImmunoDeepR

ImmunoDeepR provides a streamlined, object-oriented framework for building deep learning analysis pipelines tailored to immunology and omics data. Leveraging the power of keras3 via an intuitive R6 class architecture, the package simplifies the deployment of neural networks for complex biological datasets.

Beyond model training, ImmunoDeepR emphasizes interpretability and visual data exploration by integrating dimensionality reduction (PCA, MDS, UMAP, t-SNE), interactive visualizations (ggiraph), and model-agnostic explanations (SHAP).

Key Features

• Unified R6 API: Manage your entire workflow—from data loading to prediction—within a single NeuralNetwork object.
• Versatile Architectures: Automatically detects and constructs architectures for:
  • Multi-class Classification (MC)
  • Multi-label Classification (ML)
  • Regression (R)
  • Autoencoders (AE)
• Built-in Preprocessing: Handle missing values (kNN imputation, omission) and perform effect-size feature selection (Cohen's d, Cliff's delta).
• Advanced Visualizations: Generate publication-ready, interactive plots for multidimensional scaling, training history, evaluation metrics, and prediction scatter plots.
• Explainable AI (XAI): Native integration with fastshap and shapviz to compute and visualize SHAP values (waterfall, beeswarm, bar plots).

Installation

You can install the released version of ImmunoDeepR from this repo with:

install.packages("ImmunoDeepR")