Table 3.

NucleoNets model comparison with other models. ✓: This symbol means the related part is available in the model. ✖: This symbol means the related part is unavailable in the model. *Not mentioned in the original paper⁶. **The Scikit-learn library does not support the p-value calculation. On the contrary, the Stasmodels library does not have an ENET function. ***NucleoNets results from ABST-6.

Polygenic model	GGDPR	OLS	OLS + ENET	NucleoNetV1	NucleoNetV2	NucleoNetV3	Wide and deep model
Total Indonesian rice SNPs	1232	1232	1232	1232	1232	1232	1232
SNP data	✓	✓	✓	✓	✓	✓	✓
SNP position data	✖	✖	✖	✖	✖	✖	✓
Covariate: sample location	✓	✓	✓	✓	✓	✓	✓
Covariate: sample variety	✓	✓	✓	✓	✓	✓	✓
Shrinkage prior/regularization	Generalized double pareto	✖	ENET	✖	Modified ENET	Modified ENET	Modified ENET
Shannon’s entropy	✖	✖	✖	✖	✖	✓	✓
Evaluation: MSE	N/A*	4.104	2.517	2.779***	2.799***	2.863***	8.535
Evaluation: RMSE	N/A*	2.026	1.587	1.667	1.673	1.692	2.921
Evaluation: MBE	N/A*	− 0.236	− 0.404	0.099	0.015	− 0.074	− 2.148
Evaluation: MAE	N/A*	1.673	1.321	1.407	1.412	1.433	2.497
Evaluation: MSLE	N/A*	0.286	0.185	0.184	0.191	0.197	0.468
Evaluation: SMAPE	N/A*	64.843%	45.432%	47.156%	47.960%	47.481%	63.668%
Significance/importance level	N/A*	$p<0.05$	N/A**	$a{}^{\prime }\ge 0.025$	$a{}^{\prime }\ge 0.025$	$a{}^{\prime }\ge 0.025$	N/A
Number of significant/important SNP	9	16	N/A**	29	35	23	N/A
Execution time	N/A*	< 2 s	< 2 s	1630 s	5120 s	4910 s	6070 s