Fig. 1. Maximum Entropy (MaxEnt)-based ecological niche modeling of four antagonistic microbial families across Punjab rice agroecosystems. A, Receiver Operating Characteristic curves plotting sensitivity against 1-specificity. Red lines represent training data, blue lines show test data, and black diagonal indicates random prediction [AUC (area under the curve) = 0.5]; training AUC values were 0.893, 0.971, 0.899, and 0.972 for Nocardioidaceae, Sclerotiniaceae, Solirubrobacteraceae, and Lysobacteraceae, respectively, while corresponding test AUC values were 0.721, 0.969, 0.704, and 0.985, respectively. Lysobacteraceae demonstrated the strongest generalization with test AUC exceeding training AUC, while Sclerotiniaceae showed near-perfect discrimination on both datasets. B, Jackknife analysis of regularized training gain. Green bars show gain when each variable is excluded, blue bars show gain using only that variable, and red bars indicate gain with all variables. Bio_15 dominates for Nocardioidaceae; Bio_15 and Bio_06 co-dominate for Sclerotiniaceae; Bio_03 dominates for Solirubrobacteraceae; Bio_07 shows exceptional dominance for Lysobacteraceae with the highest independent gain across all analyses. C, Jackknife analysis of AUC confirming variable importance rankings. Bio_15 achieves highest independent AUC for Nocardioidaceae; Bio_15/Bio_06 shows balanced contributions for Sclerotiniaceae; Bio_03 dominates for Solirubrobacteraceae; Bio_07 achieves the highest independent AUC (0.88) for Lysobacteraceae, confirming annual temperature range as the dominant spatial driver. D, Predicted habitat suitability maps for Nocardioidaceae, Sclerotiniaceae, Solirubrobacteraceae, and Lysobacteraceae across Punjab Province, Pakistan. Color gradient represents suitability from unsuitable (blue, 0.00-0.10), low (light blue, 0.10-0.33), moderate (green-yellow, 0.33-0.66), to high (orange-red, 0.66-1.00); white and violet dots indicate training and test presence locations; rectangular dots depict biodiversity hotspot zones where multiple families co-occur; color scale bar (right) applies to all maps; Solirubrobacteraceae displays the broadest suitable range, Sclerotiniaceae shows the most restricted (eastern districts), Nocardioidaceae shows concentrates in northeastern-central zones, and Lysobacteraceae exhibits an intermediate fragmented pattern; southern and western Punjab consistently show low suitability across all families.