PDBe-SIFTS for protein sequence-structure mapping

PDBe-SIFTS is now available as a fully open-source, locally deployable software package that enables precise residue-level mapping between protein sequences and their three-dimensional structures. Built on the established SIFTS framework, it identifies the exact corresponding residues between protein sequences from UniProtKB and protein structures in the Protein Data Bank (PDB). 

The Structure Integration with Function, Taxonomy and Sequences (SIFTS) resource was first developed in 2002 to address the need for consistent mapping between protein sequence and structure data. Since then, it has become a central component of data integration across a wide range of biological and bioinformatics data resources. SIFTS was established as a collaboration between UniProtKB and PDBe, which are co-located at EMBL-EBI. This enabled close integration of infrastructure and rapid data exchange, but also meant that the SIFTS mapping workflow remained tightly coupled to internal systems and was not available as reusable software outside EMBL-EBI.

In response to growing demand from the community, these internal dependencies have now been removed. PDBe-SIFTS provides the first fully open-source implementation of the core SIFTS mapping framework, allowing researchers to run, inspect, and extend protein sequence-to-structure mappings locally, including for custom sequences or structures not present in UniProtKB or the PDB.

Faster search, accurate mappings 

PDBe-SIFTS introduces significant improvements in both speed and accuracy. Using the MMseqs2 search tool, it identifies the best-matching UniProtKB/SwissProt entries for protein structures at PDB archival scale, reducing search times from around 6 hours with BLASTP to just 10 minutes.

An updated and interpretable scoring scheme improves the ranking of candidate mappings. Benchmarking against curated datasets shows that the curated UniProtKB–PDB mapping is recovered as the top result in over 93% of cases, achieving accuracy comparable to manual curation while remaining fully automated.

Improved mapping using structural information

In addition to faster sequence matching, PDBe-SIFTS improves the quality of residue-level mappings by incorporating structural context. A refinement step based on protein backbone connectivity helps resolve alignment artefacts and ensures better consistency between sequence and structure.

At the scale of the full PDB archive, this refinement step improves approximately 2% of protein chain alignments. While this may seem small, these corrections are important for ensuring accurate residue-level annotation and downstream analyses.

Visit the PDBe-SIFTS Github repository (https://github.com/PDBeurope/SIFTS) and get started using the included quick-start notebook.