Integrating deep learning derived morphological traits and molecular data for total-evidence phylogenetics: lessons from digitized collections

Deep learning has previously shown success in automatically generating morphological traits which carry a phylogenetic signal. In this paper, we explore combining molecular data with deep learning derived morphological traits from images of pinned insects to generate total-evidence phylogenies and we reveal challenges. Deep learning derived morphological traits, while informative, underperform when used in isolation compared to molecular analyses. However, they can improve molecular results in total evidence settings. We use a dataset of rove beetle images to compare the effect of different dataset splits and deep metric loss functions on morphological and total evidence results. We find a slight preference for the cladistic dataset split and contrastive loss function. Additionally, we explore the effect of varying the number of genes used in inference and find that different gene combinations provide the best results when used on their own vs in total evidence analysis. Despite the pro..., Images of specimens associated with this dataset can be found in the Rove-Tree-11 dataset (https://doi.org/10.17894/ucph.39619bba-4569-4415-9f25-d6a0ff64f0e3). Molecular data was gathered from Genbank and aligned using MAFFT 7. Original Genbank accession numbers are provided. Alignments were concatenated with FASconCAT-G. Partition scheme and model selection were obtained using PartitionFinder 2.1.1. Trees were obtained via Maximum Parsimony using TNT. Example TNT scripts are provided. , , # Integrating Deep Learning Derived Morphological Traits and Molecular Data for Total-Evidence Phylogenetics: Lessons from Digitized Collections [https://doi.org/10.5061/dryad.9cnp5hqqq](https://doi.org/10.5061/dryad.9cnp5hqqq) This dataset includes all supplemental material for the paper 'Integrating Deep Learning Derived Morphological Traits and Molecular Data for Total-Evidence Phylogenetics: Lessons from Digitized Collections' ## Description of the data and file structure The following gives a description of the files included in this dataset: * GenBank accession numbers of all sequences.xlsx: This excel file contains all the genbank accession numbers for the molecular data used in this paper. Column A lists the species in question, the other columns represent the 7 different genes included. Note that not all genes were available for all species. This is discussed in the paper. * rove_stratified_split.csv: This csv file contains the dataset split for the 'Stratified' dataset co...深度学习此前已在自动生成带有系统发育信号的形态性状方面取得良好进展。本文探讨将分子数据与从针插昆虫图像中提取的深度学习衍生形态性状相结合，以构建总证据系统发育树，并揭示了相关研究挑战。深度学习衍生的形态性状虽具备信息价值，但单独使用时的性能劣于分子分析；不过在总证据分析框架下，这类性状可有效优化分子分析的结果。我们使用隐翅虫图像数据集，对比不同数据集划分方式与深度度量损失函数对形态性状分析及总证据分析结果的影响，发现支序数据集划分与对比损失函数的表现略微更优。此外，我们还探究了推断过程中使用基因数量变化带来的影响，发现单独使用时表现最佳的基因组合，与用于总证据分析时的最优组合并不相同。尽管原文此处未完整展示……本数据集关联的标本图像可在Rove-Tree-11数据集（https://doi.org/10.17894/ucph.39619bba-4569-4415-9f25-d6a0ff64f0e3）中获取。 分子数据采集自基因库（GenBank），并使用MAFFT 7进行序列比对。原始GenBank登录号已一并提供。比对结果通过FASconCAT-G进行拼接。分区方案与模型选择通过PartitionFinder 2.1.1完成。系统发育树通过最大简约法（Maximum Parsimony）结合TNT软件构建。附带了示例TNT脚本。 # 整合深度学习衍生形态性状与分子数据开展总证据系统发育研究：来自数字化标本馆藏的启示 [https://doi.org/10.5061/dryad.9cnp5hqqq](https://doi.org/10.5061/dryad.9cnp5hqqq) 本数据集包含论文《整合深度学习衍生形态性状与分子数据开展总证据系统发育研究：来自数字化标本馆藏的启示》的全部补充材料。 ## 数据与文件结构说明 以下为本数据集包含的文件说明： * 所有序列的GenBank登录号.xlsx：该Excel文件包含本文所用分子数据的全部GenBank登录号。A列列出了所涉及的物种，其余列分别对应本次研究使用的7个不同基因。需注意，并非所有物种都能获取全部7个基因的数据，相关细节已在论文中说明。 * rove_stratified_split.csv：该CSV文件包含「分层（Stratified）」数据集的划分方式相关内容……