Introduction | GPU Acceleration | Checking GPU Availability | Specifying Device | Custom Model Architectures | Creating Custom Networks | Architecture Selection Guidelines | Training Optimization | Early Stopping | Learning Rate Tuning | Batch Size Considerations | Data Quality Optimization | Handling Unknown Cell Types | Optimal Sample Simulation | Gene Selection Strategies | Working with Multiple Datasets | Merging Simulations | Cross-validation Strategy | Model Persistence and Deployment | Saving Models with Metadata | Loading and Deploying | Troubleshooting | Common Issues and Solutions | Memory Management | Reproducibility | Setting Seeds | Best Practices Summary
Package: TorchDecon 1.0.0
TorchDecon: Deep Learning-Based Cell Type Deconvolution Using torch
A deep neural network ensemble approach for cell type deconvolution of bulk RNA-seq data. TorchDecon uses simulated bulk samples generated from single-cell RNA-seq data to train deep neural networks that predict cell type fractions. This package is an R-native implementation based on the Scaden algorithm, built on the torch framework (LibTorch C++ backend) for GPU acceleration and cross-platform compatibility. Seamlessly integrates with Seurat objects (v4 and v5 compatible).
Authors:
TorchDecon_1.0.0.tar.gz
TorchDecon_1.0.0.zip(r-4.7)TorchDecon_1.0.0.zip(r-4.6)TorchDecon_1.0.0.zip(r-4.5)
TorchDecon_1.0.0.tgz(r-4.6-any)TorchDecon_1.0.0.tgz(r-4.5-any)
TorchDecon_1.0.0.tar.gz(r-4.7-any)TorchDecon_1.0.0.tar.gz(r-4.6-any)
TorchDecon_1.0.0.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION
card.svg |card.png
TorchDecon/json (API)
| # Install 'TorchDecon' in R: |
| install.packages('TorchDecon', repos = c('https://zaoqu-liu.r-universe.dev', 'https://cloud.r-project.org')) |
Bug tracker:https://github.com/zaoqu-liu/torchdecon/issues
Last updated from:139bac3eec (on main). Checks:9 OK. Indexed: yes.
| Target | Result | Time | Files | Syslog |
|---|---|---|---|---|
| linux-devel-x86_64 | OK | 252 | ||
| source / vignettes | OK | 290 | ||
| linux-release-x86_64 | OK | 264 | ||
| macos-release-arm64 | OK | 205 | ||
| macos-oldrel-arm64 | OK | 163 | ||
| windows-devel | OK | 225 | ||
| windows-release | OK | 225 | ||
| windows-oldrel | OK | 208 | ||
| wasm-release | OK | 218 |
Exports:ApplyScalingCalculateAccuracyCalculateCorrelationCalculateMAECalculateMRECalculateRMSECreateTorchDeconCreateTorchDeconEnsembleEvaluateDeconvolutionEvaluatePredictionsExportSimulationExtractCellTypesExtractSeuratDataGenerateExampleDataGetTrainingHistoryLoadModelMergeSimulationsPlotEvaluationPlotTrainingHistoryPredictFractionsProcessPredictionDataProcessTrainingDataQuickPredictRunTorchDeconSaveModelSimulateBulkSimulationToDataFrameTrainModel
Dependencies:abindaskpassbase64encBHbitbit64bitopsbslibcachemcallrcaToolscliclustercodetoolscommonmarkcorocowplotcpp11crosstalkcurldata.tabledeldirdescdigestdotCall64dplyrdqrngevaluatefarverfastDummiesfastmapfitdistrplusFNNfontawesomefsfuturefuture.applygenericsggplot2ggrepelggridgesglobalsgluegoftestgplotsgridExtragtablegtoolsherehighrhtmltoolshtmlwidgetshttpuvhttricaigraphirlbaisobandjquerylibjsonliteKernSmoothknitrlabelinglaterlatticelazyevallifecyclelistenvlmtestmagrittrMASSMatrixmatrixStatsmemoisemimeminiUInlmeopensslotelparallellypatchworkpbapplypillarpkgconfigplotlyplyrpngpolyclipprocessxprogressrpromisespspurrrR6RANNrappdirsRColorBrewerRcppRcppAnnoyRcppArmadilloRcppEigenRcppHNSWRcppProgressRcppTOMLreshape2reticulaterlangrmarkdownROCRrprojrootRSpectraRtsneS7safetensorssassscalesscattermoresctransformSeuratSeuratObjectshinysitmosourcetoolsspspamspatstat.dataspatstat.explorespatstat.geomspatstat.randomspatstat.sparsespatstat.univarspatstat.utilsstringistringrsurvivalsystensortibbletidyrtidyselecttinytextorchutf8uwotvctrsviridisLitewithrxfunxtableyamlzoo
Advanced Usage and Best Practices
Last update: 2026-01-26
Started: 2026-01-26
Algorithm and Mathematical Framework
Introduction | Problem Formulation | The Deconvolution Problem | Traditional Approaches vs. Deep Learning | Algorithmic Pipeline | Overview | Step 1: Bulk RNA-seq Simulation | Mathematical Formulation | Sparse Sample Generation | Step 2: Data Preprocessing | Log Transformation | Sample-wise Min-Max Normalization | Gene Filtering | Step 3: Neural Network Architecture | Fully Connected Network | Output Layer (Softmax) | Dropout Regularization | Step 4: Training | Loss Function | Adam Optimizer | Step 5: Ensemble Prediction | Architecture Specifications | Theoretical Considerations | Universal Approximation | Advantages over Linear Models | Limitations | References
Last update: 2026-01-26
Started: 2026-01-26
Visualization and Result Interpretation
Introduction | Setup | Generate Example Data and Run Deconvolution | Visualization Methods | 1. Stacked Bar Plot | 2. Heatmap Visualization | 3. Box Plot Comparison | 4. Training History Plot | 5. Correlation Plot (with Ground Truth) | 6. Per-Cell Type Accuracy | 7. Pie Chart for Individual Samples | Advanced Visualization | Ensemble Model Comparison | Summary
Last update: 2026-01-26
Started: 2026-01-26
Introduction to TorchDecon
Overview | Key Features | Installation | Quick Start | Step-by-Step Workflow | Step 1: Simulate Bulk Data | Step 2: Process Training Data | Step 3: Create and Train Model | Step 4: Predict Cell Fractions | Step 5: Save and Load Model | Model Architecture | Evaluation | Tips for Best Results | Citation | Session Info
Last update: 2026-01-25
Started: 2026-01-25
