domains / families / science / alphafold-predictions

alphafold-predictions

separate

Retrieves and interprets AlphaFold Protein Structure Database (AFDB) models by UniProt accession, reading pLDDT and PAE confidence correctly. Use when treating pLDDT as PER-RESIDUE confidence (not global accuracy) and recognizing a long low-pLDDT stretch as an intrinsically disordered region rather than a modeling error; reading PAE to segment confident domains and judge inter-domain/relative-position confidence that high mean pLDDT cannot certify; recognizing a static AFDB model carries NO ligands, ions, cofactors, PTMs, quaternary assembly, or alternative conformations (pLDDT sits in the B-factor column with opposite polarity to thermal motion); and deciding an AFDB entry vs re-running prediction. Keywords AlphaFold DB, pLDDT, PAE, B-factor column, intrinsic disorder, UniProt, Foldseek.

When to use

• Fetching precomputed AFDB entries by UniProt accession
• Reading pLDDT per-residue confidence
• Reading PAE per-residue-pair confidence
• Identifying intrinsically disordered regions
• Segmenting confident domains using PAE
• Judging inter-domain relative position confidence
• Deciding between using an AFDB entry or re-running prediction

When not to use

• Using AFDB coordinates as experimental holo complexes
• Reading pLDDT as thermal motion or B-factors
• Trimming low-pLDDT stretches as modeling errors
• Certifying biological correctness based on high mean pLDDT
• Determining quaternary structure or biological assembly
• Analyzing conformational ensembles or allostery
• Ligand docking or virtual screening without processing
• Mechanism or catalytic geometry analysis without holo structures

Skill page

Source