AlphaFind v2

AlphaFind v2 enables fast, structure-based similarity search for protein structures against the AlphaFold DB and TED DB using protein-level embeddings for rapid pre-filtering followed by structural alignment refinement.

Key Features:

• Whole-chain search: Searches over full protein chains for structural similarity.
• pLDDT-aware search: Restricts similarity computation to high-confidence regions defined by the AlphaFold pLDDT metric.
• Domain search: Searches over protein domains from the TED database.
• Multidomain search: Combines multiple chain-level domain matches into a single score and alignment.
• Filtering options: Allows restriction of queries to a given organism, CATH superfamily, or proteins with experimental structures.
• Query input: Accepts queries by UniProt/AlphaFold identifier.
• Similarity metrics: Computes TM-Score, RMSD, and residue-level alignments for top candidates.
• Results output: Returns a ranked list with similarity metrics, rich metadata, and an interactive 3-D superposition view.

Scientific Applications:

• Large-scale structure exploration: Enables practical exploration of large predicted and experimental protein structure collections that remain computationally infeasible with traditional three-dimensional structure comparison methods.
• Functional and evolutionary analysis: Supports structure-based analysis of protein function and evolution beyond what sequence-based methods alone can achieve.
• Domain-level investigation: Facilitates targeted search for structurally similar protein domains through integration with the TED database.
• Confidence-restricted comparison: Allows biologically relevant similarity assessment limited to the most stable and structurally reliable regions of predicted models.

Methodology:

AlphaFind v2 uses fast pre-filtering via state-of-the-art protein embeddings that preserve structural information, followed by refinement with US-align. Top candidates from the embedding-based pre-filter undergo TM-Score, RMSD, and residue-level alignment computations.