Eastman & Pande 2019 reproduction — processed RMA data, predictions, and trained model for rat → human gene expression on Open TG-GATEs

Processed data and a trained neural network from a faithful reproduction of Eastman & Pande, "Predicting Gene Expression Between Species with Neural Networks" (arXiv:1907.03041, 2019). The reproduction pipeline is open-source at Ajay1989Kumar/eastman-pande-2019-repro. This Zenodo record contains the processed artifacts the code produces, so downstream users can run the evaluation, train alternative models, or verify the reproduction without re-running the full pipeline. Contents rma_rat.tsv, rma_human.tsv — RMA + BrainArray ENTREZG v22 normalized expression matrices for Open TG-GATEs rat and human in-vitro liver microarrays. Log₂ scale, Entrez gene IDs as rows, sample barcodes as columns. Shapes: 14,132×3,226 (rat); 20,414×2,397 (human). pairs_rat_X.tsv, pairs_human_Y.tsv — replicate-averaged matrices paired on (compound, dose_level, time). 1,122 paired conditions across 139 compounds. ML-ready: rows aligned across both files via pairs_metadata.tsv and split via pairs_splits.tsv into 1,024 train pairs (125 compounds) and 98 test pairs (14 compounds — the exact set used in Eastman & Pande 2019). results_predictions.tsv, results_sigma.tsv — neural network predictions for the 98 test samples (mean and combined aleatoric + epistemic uncertainty from 50 MC-dropout passes). results_overall.tsv, results_per_sample.tsv, results_top100.tsv, results_fig5.tsv — evaluation outputs: overall Pearson r and MAE, per-sample correlations, top-100 differentially expressed gene overlap per test compound (paper's Table 1), and MAE×compound×dose at 24h (paper's Figure 5). training_log.tsv — per-epoch loss and learning rate, 1,000 epochs. train_means.npz — per-gene training-set means used for centering predictions (required to compare against the paper's centered Pearson r). model_checkpoint.pt (if present) — PyTorch state dict for the trained 1.1 B-parameter network. Single hidden layer of width 20,000, ReLU, dropout 0.5, dual heads for mean and log-variance, trained 1,000 epochs with Adam. Reproduction summary Metrics evaluated in the centered space the paper uses (per-gene training-set mean subtracted from both predictions and ground truth before correlating): MetricEastman & Pande 2019This reproductionOverall Pearson r0.6970.696Overall MAE0.1580.162Per-sample r (median)0.7910.779Top-100 DE overlap (mean / cmpd)41.140.4 End-to-end correctness audit (9 checks; see audit_pipeline.py in the GitHub repo) passes with no data leakage detected: train/test compound disjointness, train-only centering means, NaN-free predictions and weights, monotone learning-rate decay over 1,000 epochs. Source data Raw CEL files (not redistributed here) are from the Open TG-GATEs toxicogenomics database at NIBIOHN. RMA normalization used the BrainArray ENTREZG v22 custom CDF packages from the University of Michigan.