Table 2.

Comparative benchmark of different sequence- and structure-based methods on the S669 independent test set of single-point variations.

MethodInputTotal
Direct
Reverse
Symmetry
PCCRMSEMAEPCCRMSEMAEPCCRMSEMAErd-r⟨δ⟩
DDGembSEQ0.681.400.990.531.400.990.521.400.99−0.970.01
PROSTATASEQ0.651.451.000.491.451.000.491.450.99−0.99−0.01
ACDC-NN3D0.611.51.050.461.491.050.451.51.06−0.980.02
INPS-SeqSEQ0.611.521.10.431.521.090.431.531.11.000.00
PremPS3D0.621.491.070.411.51.080.421.491.05−0.850.09
ACDC-NN-SeqSEQ0.591.531.080.421.531.080.421.531.081.000.00
DDGun3D3D0.571.611.130.431.61.110.411.621.14−0.970.05
INPS3D3D0.551.641.190.431.51.070.331.771.31−0.50.38
THPLMSEQ0.531.630.391.600.351.66−0.96−0.01
ThermoNet3D0.511.641.20.391.621.170.381.661.23−0.850.05
DDGunSEQ0.571.741.250.411.721.250.381.751.25−0.960.05
MAESTRO3D0.441.81.30.51.441.060.22.11.66−0.220.57
ThermoMPNNSEQ0.431.52
Dynamut3D0.51.651.210.411.61.190.341.691.24−0.580.06
PoPMuSiC3D0.461.821.370.411.511.090.242.091.64−0.320.69
DUET3D0.411.861.390.411.521.10.232.141.68−0.120.67
I-Mutant3.0-SeqSEQ0.371.911.470.341.541.150.222.221.79−0.480.76
SDM3D0.321.931.450.411.671.260.132.161.64−0.40.4
mCSM3D0.371.961.490.361.541.130.222.31.86−0.050.85
Dynamut23D0.361.91.420.341.581.150.172.161.690.030.64
I-Mutant3.03D0.321.961.490.361.521.120.152.321.87−0.060.81
Rosetta3D0.472.692.050.392.72.080.42.682.02−0.720.61
FoldX3D0.312.391.530.222.31.560.222.481.5−0.20.34
SAAFEC-SEQSEQ0.262.021.540.361.541.13−0.012.41.94−0.030.83
MUproSEQ0.322.031.580.251.611.210.202.381.96−0.320.95
MethodInputTotal
Direct
Reverse
Symmetry
PCCRMSEMAEPCCRMSEMAEPCCRMSEMAErd-r⟨δ⟩
DDGembSEQ0.681.400.990.531.400.990.521.400.99−0.970.01
PROSTATASEQ0.651.451.000.491.451.000.491.450.99−0.99−0.01
ACDC-NN3D0.611.51.050.461.491.050.451.51.06−0.980.02
INPS-SeqSEQ0.611.521.10.431.521.090.431.531.11.000.00
PremPS3D0.621.491.070.411.51.080.421.491.05−0.850.09
ACDC-NN-SeqSEQ0.591.531.080.421.531.080.421.531.081.000.00
DDGun3D3D0.571.611.130.431.61.110.411.621.14−0.970.05
INPS3D3D0.551.641.190.431.51.070.331.771.31−0.50.38
THPLMSEQ0.531.630.391.600.351.66−0.96−0.01
ThermoNet3D0.511.641.20.391.621.170.381.661.23−0.850.05
DDGunSEQ0.571.741.250.411.721.250.381.751.25−0.960.05
MAESTRO3D0.441.81.30.51.441.060.22.11.66−0.220.57
ThermoMPNNSEQ0.431.52
Dynamut3D0.51.651.210.411.61.190.341.691.24−0.580.06
PoPMuSiC3D0.461.821.370.411.511.090.242.091.64−0.320.69
DUET3D0.411.861.390.411.521.10.232.141.68−0.120.67
I-Mutant3.0-SeqSEQ0.371.911.470.341.541.150.222.221.79−0.480.76
SDM3D0.321.931.450.411.671.260.132.161.64−0.40.4
mCSM3D0.371.961.490.361.541.130.222.31.86−0.050.85
Dynamut23D0.361.91.420.341.581.150.172.161.690.030.64
I-Mutant3.03D0.321.961.490.361.521.120.152.321.87−0.060.81
Rosetta3D0.472.692.050.392.72.080.42.682.02−0.720.61
FoldX3D0.312.391.530.222.31.560.222.481.5−0.20.34
SAAFEC-SEQSEQ0.262.021.540.361.541.13−0.012.41.94−0.030.83
MUproSEQ0.322.031.580.251.611.210.202.381.96−0.320.95

Results for all methods except DDGemb, THPLM, ThermoMPNN, and PROSTATA were taken from (Pancotti et al. 2022). For PROSTATA and THPLM direct and reverse PCC, RMSE, and MAE were taken from the reference papers (Gong et al. 2023, Umerenkov et al. 2023). ThermoMPNN results were taken from Dieckhaus et al. (2024) PROSTATA total PCC, Total RMSE, Total MAE, PCCd-r, and δ were computed using the predictions available at the method GitHub repository. Bold values highlight the top-performing method(s) on the respective metric.

Table 2.

Comparative benchmark of different sequence- and structure-based methods on the S669 independent test set of single-point variations.

MethodInputTotal
Direct
Reverse
Symmetry
PCCRMSEMAEPCCRMSEMAEPCCRMSEMAErd-r⟨δ⟩
DDGembSEQ0.681.400.990.531.400.990.521.400.99−0.970.01
PROSTATASEQ0.651.451.000.491.451.000.491.450.99−0.99−0.01
ACDC-NN3D0.611.51.050.461.491.050.451.51.06−0.980.02
INPS-SeqSEQ0.611.521.10.431.521.090.431.531.11.000.00
PremPS3D0.621.491.070.411.51.080.421.491.05−0.850.09
ACDC-NN-SeqSEQ0.591.531.080.421.531.080.421.531.081.000.00
DDGun3D3D0.571.611.130.431.61.110.411.621.14−0.970.05
INPS3D3D0.551.641.190.431.51.070.331.771.31−0.50.38
THPLMSEQ0.531.630.391.600.351.66−0.96−0.01
ThermoNet3D0.511.641.20.391.621.170.381.661.23−0.850.05
DDGunSEQ0.571.741.250.411.721.250.381.751.25−0.960.05
MAESTRO3D0.441.81.30.51.441.060.22.11.66−0.220.57
ThermoMPNNSEQ0.431.52
Dynamut3D0.51.651.210.411.61.190.341.691.24−0.580.06
PoPMuSiC3D0.461.821.370.411.511.090.242.091.64−0.320.69
DUET3D0.411.861.390.411.521.10.232.141.68−0.120.67
I-Mutant3.0-SeqSEQ0.371.911.470.341.541.150.222.221.79−0.480.76
SDM3D0.321.931.450.411.671.260.132.161.64−0.40.4
mCSM3D0.371.961.490.361.541.130.222.31.86−0.050.85
Dynamut23D0.361.91.420.341.581.150.172.161.690.030.64
I-Mutant3.03D0.321.961.490.361.521.120.152.321.87−0.060.81
Rosetta3D0.472.692.050.392.72.080.42.682.02−0.720.61
FoldX3D0.312.391.530.222.31.560.222.481.5−0.20.34
SAAFEC-SEQSEQ0.262.021.540.361.541.13−0.012.41.94−0.030.83
MUproSEQ0.322.031.580.251.611.210.202.381.96−0.320.95
MethodInputTotal
Direct
Reverse
Symmetry
PCCRMSEMAEPCCRMSEMAEPCCRMSEMAErd-r⟨δ⟩
DDGembSEQ0.681.400.990.531.400.990.521.400.99−0.970.01
PROSTATASEQ0.651.451.000.491.451.000.491.450.99−0.99−0.01
ACDC-NN3D0.611.51.050.461.491.050.451.51.06−0.980.02
INPS-SeqSEQ0.611.521.10.431.521.090.431.531.11.000.00
PremPS3D0.621.491.070.411.51.080.421.491.05−0.850.09
ACDC-NN-SeqSEQ0.591.531.080.421.531.080.421.531.081.000.00
DDGun3D3D0.571.611.130.431.61.110.411.621.14−0.970.05
INPS3D3D0.551.641.190.431.51.070.331.771.31−0.50.38
THPLMSEQ0.531.630.391.600.351.66−0.96−0.01
ThermoNet3D0.511.641.20.391.621.170.381.661.23−0.850.05
DDGunSEQ0.571.741.250.411.721.250.381.751.25−0.960.05
MAESTRO3D0.441.81.30.51.441.060.22.11.66−0.220.57
ThermoMPNNSEQ0.431.52
Dynamut3D0.51.651.210.411.61.190.341.691.24−0.580.06
PoPMuSiC3D0.461.821.370.411.511.090.242.091.64−0.320.69
DUET3D0.411.861.390.411.521.10.232.141.68−0.120.67
I-Mutant3.0-SeqSEQ0.371.911.470.341.541.150.222.221.79−0.480.76
SDM3D0.321.931.450.411.671.260.132.161.64−0.40.4
mCSM3D0.371.961.490.361.541.130.222.31.86−0.050.85
Dynamut23D0.361.91.420.341.581.150.172.161.690.030.64
I-Mutant3.03D0.321.961.490.361.521.120.152.321.87−0.060.81
Rosetta3D0.472.692.050.392.72.080.42.682.02−0.720.61
FoldX3D0.312.391.530.222.31.560.222.481.5−0.20.34
SAAFEC-SEQSEQ0.262.021.540.361.541.13−0.012.41.94−0.030.83
MUproSEQ0.322.031.580.251.611.210.202.381.96−0.320.95

Results for all methods except DDGemb, THPLM, ThermoMPNN, and PROSTATA were taken from (Pancotti et al. 2022). For PROSTATA and THPLM direct and reverse PCC, RMSE, and MAE were taken from the reference papers (Gong et al. 2023, Umerenkov et al. 2023). ThermoMPNN results were taken from Dieckhaus et al. (2024) PROSTATA total PCC, Total RMSE, Total MAE, PCCd-r, and δ were computed using the predictions available at the method GitHub repository. Bold values highlight the top-performing method(s) on the respective metric.

Close
This Feature Is Available To Subscribers Only

Sign In or Create an Account

Close

This PDF is available to Subscribers Only

View Article Abstract & Purchase Options

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

Close