Protein structure-similarity scores

Protein structure-similarity scores —interactive playground for TM-score, GDT_TS, GDT_HA, RMSD

Drag the sliders to set the Cα–Cα distance for each residue (after optimal superposition). All four scores recompute in real time. Try the presets to see how each metric responds to the same prediction.

TM-score

–

[0…1] same fold ≥ 0.5

GDT_TS

–

[0…100] mean of P₁ P₂ P₄ P₈

GDT_HA

–

[0…100] mean of P₀.₅ P₁ P₂ P₄

RMSD

–

Å sqrt(mean(d²))

controls

protein length L:

how the numbers are computed (live)

residues — drag to set Cα distance from experiment (Å)

bar colour: green <1 Å · yellow <2 Å · orange <4 Å · red <8 Å · grey ≥ 8 Å. Vertical lines mark the GDT thresholds.

show definitions / cutoffs

TM-score Zhang & Skolnick, 2004.

TM = (1/L) · Σᵢ  1 / (1 + (dᵢ/d₀)²)
d₀ = max( 1.24 · ∛(L - 15) - 1.8 ,  0.5 )

0 — unrelated · 0.17 random · > 0.5 = same fold · 1 = identical

GDT_TS Zemla, 1999.

GDT_TS = (P₁ + P₂ + P₄ + P₈) / 4
Pₓ = 100 · (# residues within x Å) / L
The four Pₓ are cumulative (nested), not exclusive bins.

GDT_HA = same recipe, tighter thresholds 0.5, 1, 2, 4 Å. Used for high-quality predictions.

RMSD = √(mean of dᵢ²) — sensitive to outliers, no length normalization.