Table 2.

MCC and ACC of adaptive LSTM and other three methods

Dataset	Metrics	ProbKnot	Cylofold	CentroidFold	Adaptive	Filter
TMR	MCC	0.105	−0.043	0.106	0.434	0.581
	ACC	0.531	0.485	0.561	0.630	0.786
SPR	MCC	0.591	*	0.668	0.786	0.751
	ACC	0.796	*	0.834	0.891	0.870
SRP	MCC	0.262	−0.184	0.177	0.421	0.475
	ACC	0.613	0.396	0.584	0.708	0.690
RFA	MCC	0.398	0.256	0.299	0.451	0.699
	ACC	0.677	0.624	0.650	0.661	0.834
ASE	MCC	0.238	0.043	0.286	0.323	0.484
	ACC	0.611	0.523	0.642	0.556	0.720
Average	MCC	0.319	0.014	0.307	0.483	0.592
	ACC	0.646	0.406	0.654	0.689	0.780

Boldface represents the highest MCC or ACC in comparison with the other three methods

*indicates Cylofold does not generate results on SPR dataset, since Cylofold can not accept the sequence with missing bases in SPR dataset