ELF-NET Data S1: Domain-environment association and modeling results

ELF-NET Data S1: Domain–environment association and modeling results Pfam domain count matrices, Pfam–GEE and Pfam–AlphaEarth correlation tables, SHAP feature importance values, PCA/UMAP/t-SNE embeddings, CCA joint embeddings, XGBoost bidirectional predictions, per-sample summaries, GPS mappings, and stratified correlation results by data source. The primary algaGPT classifications were produced by a 70-million-parameter GPT-2/nanoGPT-based protein classifier implemented in the LA4SR framework. Separate files labeled Pythia/LA4SR contain results from the Pythia-70M greedy-decoding comparison model; algaGPT itself is not Pythia-based. Related to Figures 2–5, Figure S3, and STAR Methods in Nelson et al., Coupling of oceanographic state to the dark proteome: a foundation for genome-informed marine productivity modeling.