Galled Perfect Transfer Networks

Predicting horizontal gene transfers often requires comparative sequence data, but recent work has shown that character-based approaches could also be useful for this task. Notably, perfect transfer networks (PTN) explain the character diversity of a set of taxa for traits that are gained once, rarely lost, but that can be transferred laterally. Characterizing the structure of such characters is an important step towards understanding more complex characters. Although efficient algorithms can infer such networks from character data, they can sometimes predict overly complicated transfer histories. With the goal of recovering the simplest possible scenarios in this model, we introduce galled perfect transfer networks, which are PTNs that are galled trees. Such networks are useful for characters that are incompatible in terms of tree-like evolution, but that do fit in an almost-tree scenario. We provide polynomial-time algorithms for two problems: deciding whether one can add transfer edges to a tree to transform it into a galled PTN, and deciding whether a set of characters are galled-compatible, that is, they can be explained by some galled PTN. We also analyze a real dataset comprising of a bacterial species trees and KEGG functions as characters, and derive several conclusions on the difficulty of explaining characters in a galled tree, which provide several directions for future research.
Comment: extended article based on previously accepted manuscript at RECOMB-CG 2024 conference