Supplementary Data: hPartition: An Efficient and Robust Alignment Partitioning Framework for Large-Scale DNA and Protein Phylogenomics

Supplementary Data

hPartition: An Efficient and Robust Alignment Partitioning Framework for Large-Scale DNA and Protein Phylogenomics
Submitted to PeerJ

This record contains the empirical and simulated alignments used to benchmark the efficiency, statistical fit and topological accuracy of the hPartition framework, together with the per-alignment measurements behind the figures and tables in the manuscript. The framework is natively integrated into a custom fork of IQ-TREE v2.4.0.

Empirical DNA Datasets

The DNA suite (DNA/empirical) consists of 17 multi-locus alignments drawn from previous studies. Sixteen were used in the evaluations of mPartition, gPartition and PsiPartition; a genome-scale rodent dataset was added to test upper-bound scalability.

• Small to mid-scale alignments: 10 datasets from 4,435 sites (Coenonymphina) to 25,919 (Vertebrata), including Arctiina, Calisto, Morpho, Noctuidae, Choreutidae, Pieridae, Geometridae and Osteichthyes.
• Large-scale alignments: 7 datasets above 100,000 sites — Spermatophyta (1,296,042), Rodent (1,207,638), Xenopsminutus (825,804), Phasianidae (614,159), Neoaves (539,526), Aculeata (183,747) and Actinopterygii (149,366).
• Format: concatenated multi-locus alignments in PHYLIP, representing real sequence divergence and dataset size.

Empirical Protein Datasets

The protein suite (Protein) comprises 13 datasets from large-scale phylogenomic studies.

• Subsampled sets: 12 alignments built by drawing independent subsets of 5, 10 and 20 loci from the Ballesteros, Irisarri, Jarvis and Ran series.
• Genome-scale set: the complete, un-subsampled Ran dataset — 38 taxa, 432,014 amino-acid sites, 1,308 loci — used to test upper-bound scalability.

Alignments Containing Structural RNA

The RNA suite (RNA) holds 4 multi-locus alignments in which ribosomal or transfer RNA forms part of the concatenation, covering RNA fractions from 5.5% to 55.2% of sites. All four are datasets benchmarked by Lanfear et al. (2012), used here as they were deposited with that study.

• barkbeetles (44 taxa, 1,897 sites, 16S), midges (74, 2,701, 16S), hummingbirds (164, 3,821, tRNA) and saxifragales (40, 9,237, 18S and 26S).
• These are nucleotide alignments: no method compared here models the covariation between paired rRNA stem positions, since IQ-TREE 2.4.0 provides no such model.

Simulated DNA Datasets

The simulated alignments (DNA/simulated-1 and DNA/simulated-2) are reused from the PsiPartition study (Xu and Onoda, 2024) and were used to measure topological accuracy as the Robinson-Foulds (RF) distance between inferred and true trees.

• SimDNA1: 100 alignments of 49 sequences, 5,000 sites partitioned into 1 to 10 equal-length loci.
• SimDNA2: 100 alignments of 6 to 15 loci of 1,000 sites each, so 6,000 to 15,000 sites in total.
• Generation: produced with Seq-Gen under the General Time Reversible (GTR) model with a 4-category Gamma distribution.

Raw Results

Raw_results holds the measurements the manuscript's figures and tables are computed from.

• SimDNA1.csv and SimDNA2.csv: one row per alignment per method — 300 rows each, covering hPartition, gPartition and PsiPartition on all 100 alignments — with the RF distance, its maximum, the normalised RF and the BIC. Because every method analyses the same alignments, these files support the paired per-alignment comparison and the Wilcoxon signed-rank tests reported in the Results, not only the distributions shown in the figures.
• Var_Invar.csv: the per-subset composition of the schemes hPartition builds on eight small empirical DNA alignments — partition name, total sites, variant and invariant counts. These are the figures behind the invariant-site analysis, including the subsets that contain no variant column at all.

Data Availability

The hPartition software is open-source and freely available at https://github.com/SliferSkyd/iqtree2/tree/hPartition. The version used in this study is archived at https://doi.org/10.5281/zenodo.23563227.