Write CONECT records for bonds between residues differing in insertion code

[haeganm]

Fixes #945

_remove_non_conect_bonds() found inter-residue bonds by comparing res_id and chain_id only. Two residues that differ just in ins_code (e.g. 52 and 52A) looked like one residue, so a bond between them got no CONECT record and was lost when writing a PDB file.

This now compares residue starts via get_residue_starts_for(), the same way _filter_bonds() does for the PDBx writer.

Added one small test with a disulfide between 52 and 52A. It fails on main and passes here. tests/structure/io/test_pdb.py, test_pdbx.py and tests/structure/test_bonds.py pass, ruff 0.9.7 and pyright are clean on the changed file.

[padix-key]

Good catch! Thank you for the fix 👍

[padix-key]

Would it be possible to generalize this test? For example we could check that PDBFile.set_structure() adds the same CONECT records as there are in the PDB file downloaded from RCSB. The test assets do contain structures with insertion codes, so this should catch it. If for some reason complete CONECT equivalence is not possible, we could at least compare the number of CONECT records.

[codspeed]

Merging this PR will degrade performance by 11.38%

❌ 3 (👁 1) regressed benchmarks
✅ 106 untouched benchmarks
⏩ 14 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	benchmark_match_kmer_selection[KmerTable-None]	246.2 µs	278.2 µs	-11.5%
❌	benchmark_match[KmerTable-None]	294 µs	327 µs	-10.11%
👁	benchmark_match_kmer_selection[KmerTable-11*11*1*1***111]	242.1 µs	276.8 µs	-12.51%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing haeganm:fix-conect-ins-code (9777950) with main (1aaf784)}

Footnotes

1. 14 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[haeganm]

Good idea. I replaced the synthetic test with one over every PDB test asset. It checks that the written CONECT records include all pairs from the original file's CONECT records (matched by chain, residue, insertion code and atom name, since serials get renumbered) and all inter-residue bonds of the parsed structure.

One thing I found while doing this: the RCSB comparison alone doesn't catch the bug. RCSB only writes CONECT for hetero groups and disulfides, and hetero bonds are kept by the separate hetero check. What the fix changes in 1igy are 8 peptide bonds between Kabat-numbered residues like 82/82A, which RCSB doesn't record. So the inter-residue check is the half that fails on main (for 1igy only) and passes with the fix. The RCSB check holds on both.

The failing "databases" job is the biotite.database.rcsb doctest, which is also failing on main.

[padix-key]

Shouldn't these two sets be equal?

Suggested change

	assert _conect_pairs(ref_file) <= test_pairs
	assert _conect_pairs(ref_file) = test_pairs