Modelling trait heterogeneity and inferring causal links in the macroevolution of growth habit in eudicot angiosperms

Phylogenetic comparative methods (PCMs) help researchers understand and predict trait evolutionary relationships. While improvements to PCMs have focused on increasing model complexity, understanding processes remains difficult due to persistent challenges in grounding complex models in biological reality and synthesizing findings across multiple analyses. We examined the evolution of growth habit in eudicots (75% of all angiosperms) and tested how variables such as vessel diameter, leaf phenology, and minimum temperature influence macroevolutionary inference. We used a series of PCMs to synthesize our understanding of trait interrelationships, explored plausible causal relationships using phylogenetic path analysis, and employed phylogenetic cross-validation to assess predictive performance among taxa. We found that discrete coding of growth form was linked to other measured and unmeasured traits, and that these interrelationships can help overcome limitations arising from incomplete data and simplistic coding of complex traits. Analysis of growth form using phylogenetic path analysis helps reconcile competing views of trait interrelationships from previous studies. Furthermore, including identified covariates improves prediction of growth habit and other traits. Our study shows that incorporating causal structure improves macroevolutionary inference, identifies when analyses that omit key causal traits become unreliable, and underscores the importance of integrating phylogenetic models with natural-history knowledge.

系统发育比较方法（Phylogenetic Comparative Methods，PCMs）可帮助研究者理解并预测性状演化关联。尽管针对PCMs的改进多聚焦于提升模型复杂度，但由于始终存在将复杂模型锚定至生物学现实、整合多项分析结果等难题，研究者仍难以明晰演化过程。本研究以占所有被子植物75%的真双子叶植物为对象，考察其生长习性的演化历程，并检验导管直径、叶片物候、最低气温等变量如何影响宏观演化推断。本研究采用一系列PCMs整合对性状间关联的认知，通过系统发育路径分析探索合理的因果关联，并借助系统发育交叉验证评估不同类群的预测性能。研究发现，生长型的离散编码与其他已测定及未测定性状存在关联，这类关联可帮助克服因数据不全、对复杂性状进行简化编码所带来的局限。通过系统发育路径分析开展的生长型研究，有助于调和此前研究中关于性状关联的不同观点。此外，纳入已确定的协变量可提升对生长习性及其他性状的预测精度。本研究表明，纳入因果结构可优化宏观演化推断，可识别出遗漏关键因果性状的分析何时失效，并强调了将系统发育模型与自然史知识相结合的重要性。