The CARRARA pipeline: Using machine-learning techniques for automated species delimitation in intensively hybridising plant genera based on herbarium specimens

多倍体化和杂交是区分开花植物和高等动物的进化和物种形成的两个重要过程。在SPP的前3年资助期内，我们已经研究了多倍体复合体（在菊属中）物种划分的解决方案，目前的提案将集中在更复杂的问题上，即在密集杂交植物属中使用客观和快速的机器学习方法进行自动物种划分的可能性。通过实施基于植物标本馆材料的高通量分子技术（hyRADseq），结合形态学特征的自动提取和基于博物馆材料的生态位建模，我们的目标是在不需要额外的野外工作的情况下快速发现和划分进化重要单位（物种）。我们的方法将以向日葵科（菊科，菊科）的三个植物类群为例：Senecio nemorensis syngameon（8种；欧洲），Rhodanthemum B.H.Wilcox等人（15种；西北非洲）和智利的Baccharis L.（14种，26种杂交组合）。

Polyploidisation and hybridisation are two paramount processes that distinguish evolution of and species formation in flowering plants from higher animals. While in the first 3-years funding period of SPP we have worked on solutions for species delimitation in polyploid complexes (in the genus Leucanthemum Mill.), the present proposal will focus on the even more complex questions concerning possibilities for objective and fast machine-learning approaches to automated species delimitation in intensively hybridising plant genera. By implementing high-throughput molecular techniques based on herbarium material (hyRADseq) in combination with the automated extraction of morphological characters and ecological niche-modelling based on museum material, we aim at the fast discovery and delimitation of evolutionary significant units (species) without the necessity of additional field work. Our approach will be exemplified in three plant groups of the sunflower family (Compositae, Asteraceae) known for their critical taxonomy caused by extensive hybridisation: the Senecio nemorensis syngameon (8 species; Europe), the genus Rhodanthemum B.H.Wilcox et al. (15 species; NW Africa), and the genus Baccharis L. in Chile (14 species, 26 hybrid combinations).