Skip to main content

AlphaFold3

AlphaFold3 AlphaFold 3 (AF3) is an advanced AI model that uses diffusion to predict the 3D structures of proteins, nucleic acids, small molecules, and post-translational modifications. As a newer version of AlphaFold 2, it supports multi-chain structures for co-folding and docking. AF3 enhances molecular docking studies by providing detailed insights into protein conformations, improving ligand binding predictions, and aiding in the identification of optimal docking poses, making it a valuable tool in drug discovery and design.

Inputs for Molecular Docking​

  • PDB File: The 3D structure of the target protein in PDB format.
  • Ligand Structure: The structure of the small molecule or ligand, often provided in PDB or SMILES format.
  • Binding Site Information: Details about the specific region on the protein where docking will occur.
  • Configuration Parameters: Settings that define the docking process, including options for multi-chain interactions.

Outputs From AlphaFold 3​

  • PDB File: The predicted 3D structure of the protein-ligand complex, showing the docked arrangement.
  • Confidence Scores: Metrics indicating the reliability of the predicted binding poses.
  • Interaction Details: Information on binding interactions, such as hydrogen bonds and hydrophobic contacts.

Analyzing Alphafold3 Predictions​

pLDDT Scores:​

A per-residue confidence score that indicates structural reliability, with scores ranging from 0-50 (very low confidence) to 90-100 (very high confidence). Understanding pLDDT scores is essential for assessing the accuracy of protein structure predictions.

PAE Scores:​

Evaluates confidence between residue pairs, where scores of 0-5 Å indicate low confidence (known relative positions) and scores of 20+ Å indicate high confidence (unknown relative positions). Analyzing PAE scores helps assess structural alignment and flexibility.