These conditions are not real ones, and the most efficient methods could perhaps not be the best in real situations, in particular if different methods detect different classes of HTs [40], [47], [48].
Indeed, it has been suggested that if these methods detect different fragments it is because there are different types of horizontal transfers more or less ameliorated [43] as a function of the time elapsed in the genome [40], [44].
Each of these methods detect different targets (bacterial DNA, antibodies), which have different disease dynamics and will thus be detected at different stages of a Map infection.
This discrepancy could be responsible for some of the conflicting results, since it is not known whether the current methods detect different subsets of the protein.
Because different methods detect different signatures left by recombination in the data, previous studies have suggested the use of several different methods to reject the null hypothesis of clonal evolution (Posada and Crandall 2001; Posada 2002).
Each method detected different clusters, and results show the predicted clusters with maximum pairwise Jaccard similarity index (JI) to each of the 'true' modules M1 M7.
The top 100 ranked proteins suggested by DMNC do not appear in other methods' list except MCC (30%), means this method detect different features from the other ten methods.
We recently proposed a strategy to obtain a comprehensive list of drivers minimizing the probability of detecting false-positive findings by combining complementary methods that detected different signals of positive selection (Tamborero et al., 2013b).
This inconsistency was likely due to the difference between the two methods in detecting different members of gene families.
Of interest is to examine the consistency of the predictions, i.e. to see whether the different methods are detecting different subsets of correlated pairs.
