Overlapping genes encode two different proteins from the same DNA sequence, read in different frames. This is not rare — viruses rely on it heavily, and overlapping genes appear across all domains of life. The standard view treats overlap as a compression trick: small genomes need to reuse sequence space. Kirsch, Wood, Redford, and Husain ask a different question. Not whether overlap is possible, but whether the genetic code makes it easy.
They use Potts models trained on protein family alignments to evaluate the fitness of computationally designed overlapping sequences. The results identify a fundamental trade-off — dual encoding imposes costs on both proteins — but also find that these costs are navigable. Diverse overlapping sequences are connected by mutation networks with minimal fitness penalties.
The surprising finding: the natural genetic code is uniquely well-suited for overlapping genes. Compared to alternative codes with the same amino acid assignments but different codon tables, the real code permits more viable overlapping solutions and lower fitness penalties for dual encoding. One class of reading frames is markedly more permissive than others, and the natural code exploits this asymmetry.
The through-claim: the genetic code is not just a mapping from codons to amino acids — it is a structure optimized for multiplexed information. The capacity to encode two functions in one sequence is not an incidental byproduct of degeneracy. It is a feature the code appears selected for. The redundancy that looks wasteful when you analyze one gene at a time becomes essential when you consider that the same sequence may need to serve two masters simultaneously.
This reframes what “efficient” means for an encoding scheme. A code optimized purely for single-gene fidelity would look different from one optimized for overlap tolerance. The real code leans toward overlap tolerance, suggesting that multiplexed encoding has been a persistent selective pressure throughout the history of life.