Inconsistent performance of multi-type genomic data in phylogenomics of neuropteridan insects, with solutions toward conflicting results

Reconstructing the tree of life is facing challenges to infer accurate and
robust phylogeny based on large data in the genomic era. Currently,
universal single-copy orthologs (USCOs), ultraconserved elements (UCEs),
and mitochondrial genomes (mitogenomes) are widely used to reconstruct
phylogeny. In this study, the higher-level phylogeny of Neuropterida is
reconstructed based on USCOs, UCEs, and mitogenomes assembled from 42
newly sequenced low-coverage genomes (above 32.80X), representing all
orders and almost all families except Rhachiberothidae, under various
settings in data filtering, model selection, and strategy of tree
reconstruction. According to cautious evaluation, the topology based on
amino acids matrices of the USCOs filtered by multifactorial strategies
under the site heterogeneity model (LG + PMSF (C20)) is demonstrated to be
the most robust. The average bootstrap support (ABS) values, an important
criterion in gene filtering, show great variations between different
repetitions. Applying fossils calibration at deeper nodes close to the
root of the phylogeny is demonstrated to facilitate accurate estimation of
evolutionary timescales by comparing three different calibration schemes
(deeper nodes, shallower nodes and a combination of both). These results
highlight the complexity of genomic data and offer an integrative solution
to overcome the systematic error in phylogenomic inference.