Table 4.

Performance metrics of DeepBGC models for A. niger test sets generated with 10 000-amino acid sliding window

Training	DeepBGC	Sliding
dataset	model	window	Overlap	P	R	F-m
DeepBGC	Original	10 000	50%	0.114	1	0.205
DeepBGC	Original	10 000	30%	0.159	1	0.274
50%–50%	Original	10 000	50%	0.075	1	0.140
50%–50%	Original	10 000	30%	0.105	1	0.191
50%–50%	Fungal	10 000	50%	0.464	0.765	0.578
50%–50%	Fungal	10 000	30%	0.580	0.682	0.627