Table 4. Performance of off-rate classification models for the detection of stabilizing mutations.

A
Descriptor type	Off-rate prediction model	TPR	FPR	MCC	F1	ACC	Spec	Prec
Hotspot+Molecular	RFSpot+MOL	0.839	0.023	0.815	0.839	0.959	0.977	0.839
Hotspot	RFSpot_KFC2	0.806	0.028	0.778	0.806	0.951	0.972	0.806
Hotspot	RFSpot	0.742	0.028	0.735	0.767	0.943	0.972	0.793
Hotspot+Molecular	RFSpot_KFC2+MOL	0.677	0.019	0.723	0.750	0.943	0.981	0.840
Hotspot	KFC2a	0.613	0.023	0.659	0.691	0.931	0.977	0.792
Molecular	MOL	0.581	0.019	0.653	0.679	0.931	0.981	0.818
Hotspot+Molecular	Hotpoint2+MOL	0.548	0.019	0.629	0.654	0.927	0.981	0.810
Hotspot+Molecular	KFC2b+MOL	0.548	0.019	0.629	0.654	0.927	0.981	0.810
Hotspot	Hotpoint2	0.484	0.009	0.621	0.625	0.927	0.991	0.882
Hotspot+Molecular	KFC2a+MOL	0.516	0.019	0.604	0.627	0.922	0.981	0.800
Hotspot+Molecular	Hotpoint1+MOL	0.452	0.019	0.552	0.571	0.914	0.981	0.778
Hotspot	KFC2b	0.516	0.033	0.551	0.593	0.910	0.967	0.696
Molecular	CP	0.355	0.005	0.539	0.512	0.914	0.995	0.917
Molecular	PB	0.419	0.019	0.524	0.542	0.910	0.981	0.765
Molecular	CP_AP	0.323	0.005	0.510	0.476	0.910	0.995	0.909
Hotspot	Hotpoint1	0.387	0.019	0.496	0.511	0.906	0.981	0.750
Molecular	AP	0.355	0.028	0.428	0.458	0.894	0.972	0.647

(A) Shows classification accuracy measures for the for the detection of stabilizing mutations (Δlog₁₀(k_off) <−1) from destabilizing mutations (Δlog₁₀(k_off) >1). This dataset contains 31 stabilizing mutants and 213 destabilizing mutants (referred to as CDS2). (B) Shows results for the detection of stabilizing mutations (Δlog₁₀(k_off) <−1) from neutral to destabilizing mutations (Δlog₁₀(k_off) >0). This dataset contains 31 stabilizing mutants and 503 neutral to destabilizing mutants (referred to as CDS1).