Direct whole-plastome alignment systematically inflates long-branch artifacts in chloroplast phylogenomics: a 468-paper audit and three-method reanalysis

Abstract Plastid genomes are the workhorse of plant phylogenomics, but their quadripartite structure, inverted-repeat (IR) and small single-copy (SSC) orientation changes, gene rearrangements and length-variable noncoding regions make direct whole-plastome multiple sequence alignment (MSA) problematic. We audited 468 chloroplast/plastid genome papers published in Mitochondrial DNA Part B between October 2023 and October 2026, extracted their materials and methods, and flagged 53 papers with suspicious long branches or topology. Each candidate was reanalyzed with three independent strategies: (C) whole-genome MAFFT + trimAl + IQ-TREE, mimicking the common published workflow; (A) SibeliaZ locally collinear blocks (LCBs) concatenated and analyzed with a patched RAxML-NG implementing T92+F+FreeRate (2-10 categories, BIC-selected); and (B) single-copy orthologs + IQ-TREE TESTMERGE. We compared Robinson-Foulds distances and branch-length diagnostics across methods and manually reviewed every case against the published tree. Thirty-six papers were strict positives: the published long branch was reproduced by C and disappeared or shortened under A/B, giving a confirmed positivity rate of 7.69% (95% CI 5.61-10.47%; 10.75% among 335 branch-length-auditable papers); 34 of the 36 (94.4%) had used direct whole-plastome MSA in the original materials and methods, including one case with an ambiguous figure legend that we mark with an asterisk and test separately. The remaining candidates comprised eight CDS/PCG-based studies, six whole-genome cases attributable to local-clock or genuine rate heterogeneity, and three negatives, one of which was reclassified after re-review. A BLAST screen of the 24 positives that lacked SSC/IR assays, plus a LASTAL re-check of the nine whose original assays were negative, attributed 16 additional cases to SSC orientation discordance and 16 to complex rearrangement; all 36 positives now have a structural attribution. Positivity was 20.0% in 2023 and approximately 6.7% in 2024-2026. All 36 positives deposited supplementary material, but none deposited a machine-readable alignment, partition scheme or tree topology. Direct whole-plastome MSA is therefore not a neutral shortcut: in this corpus it systematically inflates long branches, and distance-matrix residuals localize the effect to the curated long-branch taxa (26/27 C-A and 25/27 C-B comparisons; A-B shows no directional difference). Shared alignment errors can also produce supported clades when several close relatives carry the same structural variant, so topology alone is not a safeguard. It is not the only cause of long branches; routine locus extraction, LCB- or ortholog-aware alignment, alignment inspection, and explicit rooting/IR-SSC checks are required to separate alignment artifacts from biological rate variation.