Extra base hits: Widespread empirical support for instantaneous multiple-nucleotide changes

Despite many attempts to introduce evolutionary models that permit substitutions instantly alter more than one nucleotide in a codon, the prevailing wisdom remains such changes are rare and generally negligible or reflective of non-biological artifacts, as alignment errors. Codon continue posit only single change have non-zero rates. Here, we develop test simple hierarchy codon-substitution with rates for one-nucleotide (1H), one- two-nucleotide (2H), any (3H) codon substitutions. Using over 42, 000 empirical alignments, find widespread statistical support multiple hits: 61% alignments prefer 2H allowed, 23%—with 3H allowed. Analyses simulated data suggest these results not likely be due artifacts model misspecification Further modeling reveals synonymous island jumping among codons encoding serine, especially along short branches, contributes significantly this signal. While serine were prominently involved multiple-hit substitutions, there other common exchanges contributing better fit. It appears small subset sites most unusual dynamics well explained by existing formalisms, commonly estimated quantities, dN/dS ratios may biased misspecification. Our findings highlight need continued evaluation assumptions underlying workhorse subsequent inference techniques. We provide software implementation biologists assess potential impact extra base hits their HyPhy package Datamonkey.org server.