Table 1.

Performance of evaluation metrics. In the test set

	Default		Linear regression		MLP		OTU latent space		Combined latent space
Input Variables	Pearson	Bray–Curtis	Pearson	Bray–Curtis	Pearson	Bray–Curtis	Pearson	Bray–Curtis	Pearson	Bray–Curtis
Age, T, rain, line, variety	0.5852	0.5140	0.6629	0.4659	0.7219	0.4169	0.7340	0.4222	0.7229	0.4065
Age, T, rain	0.5852	0.5140	0.6641	0.4638	0.6927	0.4527	0.7348	0.4181	0.7220	0.4072
T and rain	0.5852	0.5140	0.5881	0.5089	0.6323	0.5047	0.6087	0.4686	0.6676	0.4557
Age and T	0.5852	0.5140	0.6622	0.4664	0.6814	0.4591	0.7323	0.4204	0.7189	0.4094
Age and rain	0.5852	0.5140	0.6628	0.4728	0.7155	0.4211	0.7361	0.4200	0.7048	0.4139

Note: In Pearson, higher scores are better, because it is a correlation metric. In Bray–Curtis, lower scores are better, as it is a dissimilarity metric. Bold means the best model per metric and row. Underline means the best model per metric in the table.