The original FASTA sequences, alignment matrices, and all resulting phylogenetic tree files used in mito-APG analysis

Resolving the phylogenetic relationships among the five major angiosperm lineages remains challenging, with large-scale nuclear and plastid datasets often yielding conflicting topologies. We successfully assembled the first complete mitochondrial genomes for two of the five major lineages within core angiosperm—Ceratophyllales (Ceratophyllum) and Chloranthales (Chloranthus)—and comprehensively assessed the utility of mitochondrial genome in resolving deep angiosperm phylogeny. Through analyses of multiple mitochondrial and plastid datasets using diverse phylogenetic methods, we found consistent support for two important relationships despite overall topological discordance between genomic compartments: (1) all mitochondrial datasets recovered Ceratophyllum + Chloranthus as a monophyletic clade with strong support, and (2) protein-based analyses from both mitochondrial and plastid genomes, along with approximately half of all mitochondrial trees, supported monocots and eudicots as sister groups. We demonstrate that topological instability often observed in mitochondrial phylogenies can be effectively mitigated through RNA editing adjustment strategies (C-to-U edit or delete), which significantly increased tree consistency across inference methods. However, substantial gene tree conflict persisted, particularly at deep nodes. Our investigation revealed that this conflict frequency is significantly negatively correlated with gene entropy (R = -0.55, p < 0.01) and sequence length (R = -0.63, p < 0.001), but showed no significant direct correlation with overall evolution rate or informative sites ratio. Based on these findings, we developed a marker optimization framework integrating both the rate of evolution and entropy, identifying five optimal low-conflict markers (atp1, atp4, ccmFC, matR, rps3). Furthermore, differences in mitogenomic architectural features exhibited strong phylogenetic signals (Blomberg’s K), with GC content (K = 0.948, p<0.001) and RNA editing site numbers (K = 1.092, p<0.001) showing significant phylogenetic conservation and independently supporting the monocot-eudicot sister relationship through possession of shared characteristics. We also discovered significant negative correlations between branch length and both GC content and RNA editing site abundance. These findings demonstrate that mitochondrial genomic data, when appropriately analyzed considering RNA editing, gene characteristics like entropy, and complemented with genomic feature-based approaches, offers unique and valuable insights for resolving deep phylogenetic relationships in angiosperms.