DGAT: Spatial Transcriptomics and Proteomics Datasets for Model Training, Prediction and Reproduction

DGAT: Spatial Transcriptomics and Proteomics Datasets for Model Training, Prediction, and Reproduction This Zenodo repository provides datasets, pretrained models, and analysis results supporting DGAT (Dual-Graph Attention Network), a deep learning framework for predicting spatial protein expression from spatial transcriptomic data. These resources support DGAT model training, protein expression prediction, evaluation, and reproduction of the results presented in the associated publication. Data Contents The repository contains five downloadable ZIP archives: DGAT_training_datasets.zip (470.2 MB)Spatial transcriptomic and matched protein expression datasets used for DGAT model training and evaluation. DGAT_prediction_ST_data.zip (113.3 MB)Spatial transcriptomic datasets prepared for protein expression prediction using DGAT. DGAT_pretrained_models.zip (578.6 MB)Pretrained DGAT model files for protein expression prediction. DGAT_results_reproduction.zip (192.9 MB)Prediction outputs and processed results used to reproduce the figures and analyses presented in the DGAT publication. revision_datasets.zip (44.4 MB)Additional datasets associated with the DGAT manuscript revision. Code and Reproducibility The DGAT source code, preprocessing procedures, model training and prediction notebooks, and figure reproduction workflows are available in the GitHub repository: https://github.com/osmanbeyoglulab/DGAT Users can download the corresponding datasets and pretrained models from this Zenodo record and follow the instructions in the GitHub repository to reproduce the published analyses or perform protein expression prediction on spatial transcriptomic data. Associated Publication DGAT: a dual-graph attention network for inferring spatial protein landscapes from transcriptomics. Nature Communications (2026). https://doi.org/10.1038/s41467-026-73114-z Citation If you use these datasets, pretrained models, or reproduction resources in your research, please cite the associated DGAT publication and this Zenodo dataset record.