RESUMO
Mitochondria and plastids power complex life. Why some genes and not others are retained in their organelle DNA (oDNA) genomes remains a debated question. Here, we attempt to identify the properties of genes and associated underlying mechanisms that determine oDNA retention. We harness over 15k oDNA sequences and over 300 whole genome sequences across eukaryotes with tools from structural biology, bioinformatics, machine learning, and Bayesian model selection. Previously hypothesized features, including the hydrophobicity of a protein product, and less well-known features, including binding energy centrality within a protein complex, predict oDNA retention across eukaryotes, with additional influences of nucleic acid and amino acid biochemistry. Notably, the same features predict retention in both organelles, and retention models learned from one organelle type quantitatively predict retention in the other, supporting the universality of these features-which also distinguish gene profiles in more recent, independent endosymbiotic relationships. A record of this paper's transparent peer review process is included in the supplemental information.
