Large Scale Insect Phylogenomics Analyses reveal new insights into insect Evolution
Description
Understanding insect evolution requires resolving deep phylogenetic relationships and divergence times, yet conflicting hypotheses persist due to biological and methodological complexities. Here, we address this challenge by reconstructing a robust, time-calibrated phylogeny of Insecta using 481 single-copy orthologous genes from 794 high-quality genomes. Our analyses resolve most inter-ordinal relationships with high confidence and employ a multi-faceted approach to diagnose the nature of two contentious nodes. We show that the conflict surrounding the relationships among Paraneoptera is the result of a hard polytomy, consistent with an ancient radiation event where phylogenetic signal is genuinely absent. In contrast, the conflict at the base of Holometabola is characterized by a dominant signal competing with a significant secondary one, a pattern consistent with deep coalescence that allows for a statistically confident resolution. Our divergence time estimates place the origin of insects in the Early Ordovician (~469 Ma) and the evolution of flight in the Silurian, coinciding with the rise of terrestrial vascular plants. Ultimately, our study provides a comprehensive framework for the insect tree of life and demonstrates how moving beyond a single topological estimate to diagnose the distinct evolutionary processes underlying phylogenomic conflict can resolve key relationships and explain the sources of persistent incongruence.
Files
Steps to reproduce
Data Collection We initially obtained genome assemblies from InsectBase 2.0 and the NCBI Genome Database (NCBI Resource Coordinators et al. 2018; Mei et al. 2022). To ensure the quality and comprehensiveness of gene sets for ortholog identification, we employed BUSCO v5.2.0 to assess genome completeness (Manni et al. 2021). For this purpose, we used the Insecta official ortholog sets from OrthoDB (Insecta_odb10.2020-09-10) (Zdobnov et al. 2021). We retained genome assemblies with BUSCO completeness scores greater than or equal to 80% for subsequent analysis. This resulted in 694 genome assemblies representing 414 genera from 146 families across 19 orders and two Diplura species to serve as outgroups (Table S1). Data Processing and Alignment We identified orthologous groups (OGs) using BUSCO v5.2.0 (Manni et al. 2021). To ensure a comprehensive representation of species, we specifically chose single-copy OG genes that exit in over 95% of the species (662 species), resulting in 698 single-copy OGs. Amino acid sequences for each OG were aligned with MAFFT v7.505 (Rozewicki et al. 2019) using the parameters "-localpair -maxiterate 1000". Subsequently, we trimmed each OG alignment using TrimAl v1.4 (Capella-Gutiérrez et al. 2009) with the settings “-resoverlap 0.5 -seqoverlap 50”. We employed IQ-TREE 2 (Minh et al. 2020b)with ModelFinder (Kalyaanamoorthy et al. 2017) to determine the optimal substitution model, and conducted 1000 ultrafast bootstrap replicates with UFBoot2 (Hoang et al. 2018), including 1000 corresponding branch support metrics via SH-aLRT (Anisimova et al. 2011) (Shimodaira-Hasegawa approximate likelihood ratio test), to obtain the maximum likelihood (ML) trees. As part of refining the dataset, we used the best-scoring ML trees for each gene. We then identified branches where the length exceeded 20% of the total tree length (i.e., the sum of all branch lengths). This relative branch length test is invaluable for detecting misaligned or misidentified orthologs within alignments, as it can reveal abnormally long branch lengths (Dos Reis et al. 2012; Springer and Gatesy 2018). We detected 15 ortholog alignments with at least one relative branch length exceeding 20% of the total tree length, and therefore excluded those specific ortholog alignments from further analysis.