For AlphaFold models

Computational models, especially those from AlphaFold, do not have experimental data fit metrics like R-factors or FSC because they are not derived directly from experimental scattering or resonance data. Instead, their quality is assessed by their predicted accuracy and confidence scores.

Predicted Local Distance Difference Test (pLDDT)

pLDDT is a per-residue score (ranging from 0 to 100) that estimates the local confidence of the model at each amino acid residue. It predicts how well the model’s predicted structure for that residue would agree with the true structure.

Visually, pLDDT is often colour-coded onto the protein backbone (e.g., dark blue for very high confidence, light blue for high, yellow for low, and red for very low). Pay close attention to the pLDDT of your region of interest.

Predicted Aligned Error (PAE)

This metric shows the expected error in the relative positions of any two residues. PAE is defined as the expected positional error at residue X, measured in Ångströms (Å), if the predicted and actual structures were aligned on residue Y.

PAE is particularly useful for understanding the relative orientations of domains or subunits in multi-domain proteins, indicating which parts are confidently positioned relative to each other versus those that might be flexible.

For a deeper dive into AlphaFold usage and model interpretation, refer to the AlphaFold: A Practical Guide online course.