Correction of weighted orthology and paralogy relations. Complexity and algorithmic results

A relation graph for a gene family is a graph with vertices representing the genes, edges connecting pairs of orthologous genes and “missing” edges representing paralogs. While a gene tree directly leads to a set of orthology and paralogy relations, the converse is not always true. Indeed a relation graph cannot necessarily be inferred from any tree, and even if it is “satisfiable” by a tree, this tree is not necessarily “consistent”, i.e. does not necessarily reflect a valid history for the genes, in agreement with a species tree. Here, we consider the problems of minimally correcting a relation graph for satisfiability and consistency, when a degree of confidence is assigned to each orthology or paralogy relation, leading to a weighted relation graph. We provide complexity and algorithmic results for minimizing corrections on a weighted graph, and also for the maximization variant of the problems for unweighted graphs.