Codon Usage in SARS-CoV-2 Spike: Global Dominance of the Suboptimal TAG Stop Codon and Amino Acid Specific Biases from 9.3 Million Genomes

We analyzed approximately 9.3 million global SARS-CoV-2 genomes to characterize codon usage in the spike gene. Sequences were processed in 188 chunks of ~50,000 genomes each, with results aggregated into tab-separated value (TSV) files. Standard indices including relative synonymous codon usage (RSCU), effective number of codons (ENc), stop codon distribution, and GC3s were computed. The TAG stop codon predominated (43.1%), despite being suboptimal in human cells. Strong biases were observed for AGA (Arg), CCA (Pro), and TCA (Ser) codons which is underrepresented in highly expressed human genes. A consistent preference for U-ending codons (e.g., TTT, ATT) was evident across the dataset. The corrected ENc value (~42) indicates pronounced codon bias, not random usage. We, therefore conclude that the spike codon usage is shaped more by mutational pressure and evolutionary constraints than by selection for translational efficiency. The persistence of non-optimal codons challenges the narrative of progressive host adaptation and may instead reflect regulatory or structural roles.