StepBrothers: inferring partially shared ancestries among recombinant viral sequences

Phylogeneticists have developed several statistical methodsto infer recombination among molecular sequences that are evolutionarilyrelated. Of these methods, Markov change-point models currentlyprovide the most coherent framework. Yet, the Markov assumptionis faulty in that the inferred relatedness of homologous sequencesacross regions divided by recombinant events is not independent,particularly for nonrecombinant sequences as they share thesame history. To correct this limitation, we introduce a novelrandom tips (RT) model. The model springs from the idea thata recombinant sequence inherits its characters from an unknownnumber of ancestral full-length sequences, of which one onlyobserves the incomplete portions. The RT model decomposes recombinantsequences into their ancestral portions and then augments eachportion onto the data set as unique partially observed sequences.This data augmentation generates a random number of sequencesrelated to each other through a single inferable tree with thesame random number of tips. While intuitively pleasing, thissingle tree corrects the independence assumptions plaguing previousmethods while permitting the detection of recombination. Thesingle tree also allows for inference of the relative timesof recombination events and generalizes to incorporate multiplerecombinant sequences. This generalization answers importantquestions with which previous models struggle. For example,we demonstrate that a group of human immunodeficiency type 1recombinant viruses from Argentina, previously thought to havethe same recombinant history, actually consist of 2 groups:one, a clonal expansion of a reference sequence and anotherthat predates the formation of the reference sequence. In anotherexample, we demonstrate that 2 hepatitis B virus recombinantstrains share similar splicing locations, suggesting a commondescent of the 2 viruses. We implement and run both examplesin a software package called StepBrothers, freely availableto interested parties.