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

ELF-NET Data S1: Domain-environment association and modeling results PFAM domain count matrices (CLR-transformed and raw), PFAM-GEE Spearman correlation tables (20,318 domains × 37 environmental variables; n = 1,809), PFAM-AlphaEarth correlation tables (20,318 domains × 64 embedding dimensions; n = 995), SHAP feature importance values from XGBoost reverse models (32 environmental targets), PCA/UMAP/t-SNE embeddings of the PFAM manifold, CCA joint embeddings and canonical loadings, XGBoost bidirectional cross-validated predictions (spatial block CV with 2° grid cells), per-sample summaries with GPS coordinates and ocean basin assignments, stratified correlation results by data source (metagenome, transcriptome, reference genome), and temporal hold-out cross-validation results. Related to Figures 2-6 and STAR Methods.