Spearman correlations of protein property prediction methods do not correlate perfectly with absolute error

Summary

Spearman values of protein property prediction methods do not correlate with their mean squared errors (1).

Figures

Method	Pearson r	MAE	RMSE
MSAesm_ddG	0.54	0.96	1.39
PROSTATA	0.49	1.00	1.45
ACDC-NN	0.46	1.05	1.49
ACDC-NN-Seq	0.42	1.08	1.53
ThermoMPNN	0.43	-	1.52
DDGun3D	0.43	1.11	1.60
DDGun	0.41	1.25	1.72
PremPS	0.41	1.08	1.50
RaSP	0.39	1.14	1.63
ThermoNet	0.39	1.17	1.62
Rosetta	0.39	2.08	2.70
Dynamut	0.41	1.19	1.60
INPS3D	0.43	1.07	1.50
INPS-Seq	0.43	1.09	1.52
SDM	0.41	1.26	1.67
PoPMuSiC	0.41	1.09	1.51
MAESTRO	0.50	1.06	1.44
FoldX	0.22	1.56	2.30
DUET	0.41	1.10	1.52
I-Mutant3.0	0.36	1.12	1.52
Dynamut2	0.34	1.15	1.58

Table from (1)

1.

Cuturello F, Celoria M, Ansuini A, Cazzaniga A. Enhancing predictions of protein stability changes induced by single mutations using MSA-based language models. Bioinformatics. 2024;40(7). Available from: https://doi.org/10.1093/bioinformatics/btae447