G4 Paradox: Cross-kingdom Comparative Genomics, Time-Calibrated Evolutionary Inference, and Experimental Validation

Reproducibility package for the manuscript: Tanigawa M. & Iwaki T. (2026) "A conserved bidirectional selection regime shapes G-quadruplex genome architecture across eukaryotes" Nature Ecology & Evolution (submitted) This deposit contains four isolated, independently runnable analysis directories (extracted into a single wrapper g4_paradox_reproducibility/): 1. g4_paradox_gmmcmc/ — Generator Matrix Markov Chain Monte Carlo (GM-MCMC) inference of G4-paradox evolutionary dynamics on a Grafen-scaled phylogeny of 163 eukaryotic species. 2. g4_paradox_time_calibrated/ — Time-calibrated phylogenetic Bayesian inference using TimeTree-derived divergence constraints with chronos penalised likelihood, GM-MCMC inference on the calibrated tree, and epoch-binned endpoint-conditioned stochastic mapping (Hobolth-Stone uniformization). Demonstrates 75.8% posterior probability that the Last Eukaryotic Common Ancestor (LECA, ~1.6 Gya) was already in the full G4-paradox state. 3. g4_chip_atlas_validation/ — In vivo experimental validation of in silico stable G4 predictions using 20 BG4 ChIP-seq datasets from human cell lines (K-562, HEK293/HEK293T) downloaded from ChIP-Atlas. Demonstrates promoter G4 odds ratio of 16.3 (P < 10^-300). 4. g4_chip_mouse_validation/ — Cross-mammalian and cross-modality experimental validation: (i) mouse 3T3 G4P-ChIP overlay (Zheng et al. 2020, GSE133379, mm9->mm39 lifted), demonstrating 7-21 fold enrichment of predicted promoter and 5'UTR G4; (ii) human rG4-seq overlay (Kwok et al. 2016, GSE77282, hg19->hg38 lifted), demonstrating 1.51-fold intron G4 enrichment (P < 1e-8) and identifying a complementary RNA-level mechanism distinct from DNA-level chromatin G4. Each directory contains: - All analysis scripts (Python 3.10, R 4.4) - Frozen input data snapshots (TSV, BED, Newick) - Raw MCMC posterior chains (16 chains, 250K iterations each) - Intermediate TSV outputs and contingency tables - Per-directory README with reproduction instructions - LICENSE (Apache 2.0 for code, CC BY 4.0 for data) Random seeds are fixed throughout (MCMC chains 0-3; stochastic mapping seed 20260514) for bit-identical reproduction. Excluded from this deposit (by design): - The publication manuscript itself, supplementary methods PDF, cover letter, and publication-formatted figures (excluded for journal copyright; available from Nature Ecology & Evolution under the journal's terms after publication). - All figure outputs (*/figures/*.{pdf,png,svg}); these are bit-identically regeneratable by re-running the analysis scripts. - Genome FASTA / GFF3 files (publicly available from primary genome consortia: NCBI Genome, Ensembl Plants, Phytozome 13, FungiDB, VEuPathDB, WormBase ParaSite WBPS19; per-species accession identifiers provided in the manuscript Supplementary Table S1). The methodology extends the Generator Matrix MCMC framework of Iwaki & Tanigawa (2025), AIP Advances 15:035102, from continuous-state DNA conformational sampling to discrete-state phylogenetic ancestral inference. Archive structure: - 1 file: g4_paradox_zenodo_upload.tar.gz (~1.03 GB, 1,077,489,059 bytes; 280 entries; md5 6bae72f996f7758e7aa4e6c509cc9ee5) - Symlink-free (Windows compatible) - Extracts to single wrapper g4_paradox_reproducibility/