Multi-omics Phylogenetic Analysis — Integrative Phylogenomics

Phylogenomics originally referred to phylogenetic inference using genome-scale data, typically from a single molecular layer such as whole-genome sequences. Multi-omics phylogenetic analysis extends this by integrating data from two or more molecular layers — for example, combining genomic, transcriptomic, and proteomic sequences. The multi-omics approach can reduce biases specific to any single data type and is increasingly standard in studies where transcriptomes or proteomes offer broader taxon coverage than complete genomes.

The supermatrix (concatenation) approach pools all orthologous loci into one large alignment and infers a single tree, maximising power but assuming all loci share the same history. The coalescent approach (e.g., ASTRAL) infers individual gene trees first and then summarises them into a species tree, explicitly modelling incomplete lineage sorting. Use the coalescent approach when your focal clade experienced a rapid radiation or when gene-tree discordance is high; use supermatrix when taxa are deeply diverged and lineage sorting is less of a concern. Running both and comparing topologies is best practice.

There is no universal threshold, but analyses with hundreds of single-copy orthologs consistently outperform single-gene approaches. Studies routinely use 200–2000 loci. More important than raw number is quality: a well-curated set of 300 cleanly aligned, paralog-free orthologs is more informative than 2000 noisy or misaligned loci. Node support typically plateaus after several hundred loci; beyond that, systematic error rather than stochastic error becomes the limiting factor.

Yes. Transcriptome-based phylogenomics using de novo assembled RNA-seq data has become standard for non-model organisms lacking a reference genome. Tools such as Trinity for assembly and BUSCO for completeness assessment enable multi-locus phylogenetic analyses from RNA-seq alone. Proteomics-based approaches using mass spectrometry data are also applicable, though coverage is typically lower than transcriptomics.

Compositional heterogeneity — where nucleotide or amino-acid frequencies differ systematically across taxa — is a known source of artefactual groupings (e.g., GC-rich taxa clustering together irrespective of true relationship). Mitigation strategies include working at the amino-acid level rather than nucleotide level for deep divergences, applying recoding schemes (e.g., Dayhoff-6 amino-acid recoding), using site-heterogeneous mixture models (e.g., C60 or LG+C60 in IQ-TREE), or employing models that explicitly account for compositional non-stationarity such as those implemented in PhyloBayes.