Ge-SAND: an explainable deep learning-driven framework for disease risk prediction by uncovering complex genetic interactions in parallel

Table 6 Classification metric deltas (Δ%) for Ge-SAND and its ablated variants across three diseases

Disease	Metric	Ge-SAND	TBEM	Δ(%)	WGNLN	Δ(%)
Schizophrenia	AUC-ROC	0.646	0.597	+ 8.208	0.602	+ 7.380
	AUC-PR	0.637	0.568	+ 12.187	0.624	+ 2.132
	KS	0.295	0.180	+ 64.345	0.180	+ 64.345
	MCC	0.282	0.180	+ 56.319	0.181	+ 55.801
	ACC	0.641	0.590	+ 8.699	0.590	+ 8.699
	F1 score	0.641	0.568	+ 12.932	0.615	+ 4.312
Alzheimer’s disease	AUC-ROC	0.716	0.653	+ 9.614	0.670	+ 6.914
	AUC-PR	0.717	0.663	+ 8.194	0.663	+ 8.210
	KS	0.380	0.305	+ 24.509	0.324	+ 17.320
	MCC	0.381	0.301	+ 26.578	0.325	+ 17.411
	ACC	0.690	0.650	+ 6.121	0.662	+ 4.234
	F1 score	0.695	0.654	+ 6.220	0.652	+ 6.567
Crohn's disease	AUC-ROC	0.700	0.551	+ 27.042	0.660	+ 6.125
	AUC-PR	0.637	0.546	+ 16.752	0.693	− 8.081
	KS	0.386	0.158	+ 144.459	0.281	+ 37.367
	MCC	0.393	0.158	+ 148.734	0.298	+ 32.012
	ACC	0.684	0.589	+ 16.129	0.640	+ 6.808
	F1 score	0.731	0.586	+ 24.702	0.568	+ 28.607

ISSN: 1471-2164