SSE-TSR

Overview

SSE-TSR integrates secondary structure information with the Triangular Spatial Relationship (TSR) framework. Traditional TSR creates triangles between Cα atoms in protein structures and assigns each triangle an integer key based on its residue labels and geometric features such as distance and angle; triangles alike in those respects share a key. SSE-TSR builds on this by categorising TSR keys into 18 distinct groups based on secondary structural elements — alpha helices, beta sheets and coils.

This offers a more detailed representation of protein structures and improves classification accuracy. By considering not only the geometric arrangement of Cα atoms but also secondary structure, SSE-TSR gives a richer characterisation of proteins. This dual-level representation supports efficient, alignment-free comparison, capturing both high-level structural features and subtle substructural variation.

The method has been tested on two major protein datasets, showing strong performance in classifying proteins by structure.

Run it with BioTSR

SSE-TSR is an option of BioTSR's protein Cα methods, --sse. To compute it for chain A of ubiquitin:

python -m pip install "biotsr==1.0.0"
biotsr fetch 1UBQ --pdb-dir structures
biotsr run --method protein --pdb structures/1UBQ.pdb --chain A --sse --output results

The option changes no key. Ubiquitin chain A gives the same 70,300 triangles and 45,349 distinct keys as a run without it; what is added is the secondary-structure category of every triangle, so each key's count is split over the 18 categories. The keys file's category column, which in a plain run reads regular, names the category, with one row for each key and category it occurs in. The files are written under the prefix results/1UBQ_A_protein_sse.

• Where the secondary structure comes from: the structure file's own HELIX and SHEET records. BioTSR does not run DSSP or assign secondary structure itself, so a file without those records has every residue in coil.
• Which methods take it: the two protein Cα methods, protein and aa_selection. It combines with amino acid grouping, mirror images and size filtering on the same command.
• From Python: biotsr.analyze(..., sse=True); the result's category_totals() gives the count per category.

Tutorial 03 walks through one run, and the SSE-TSR method record, §4 defines the 18 categories.

How BioTSR differs from the earlier program

Besides the protein Cα conventions listed on the TSR method page, the SSE-TSR method record, §7 describes three differences specific to the earlier SSE program:

• β-strand identity. The earlier program identified a strand by its serial number alone, which restarts in each sheet, so strands of different sheets could be counted as one. BioTSR names a strand by its sheet and serial number. This changes no key, only the category of some triangles, and only on chains with more than one sheet.
• Insertion codes. The earlier program left out residues with an insertion code; BioTSR keeps them.
• The per-key category file. The record describes a defect in the earlier program's per-key category file; its per-triangle category assignments are unaffected.