Here’s an example figure, but note that the “alignment” is transpose of how we usually think.
Phylogenetic Reconstruction. We reconstructed phylogenies using two in- dependent approaches. First, we calculated a distance matrix for each patient using an “equal or not” distance (31). This method increases the distances between two samples if they have unequal genotypes, regardless of the magnitude of the difference. We then used neighbor-joining (51) in R to infer the phylogenetic relationships between samples. In the very rare case of missing values, we imputed them using the nearest neighbor. We used bootstrapping with 1,000 replicates to test the reliability of the resulting trees (52) and collapsed all interior branches with bootstrap values below 70% into polytomies. Next, we used Bayesian inference of phylogeny— a methodology that relies on a fundamentally different set of principles than neighbor-joining—to construct the phylogenies. The results were al- most identical in all cases, confirming the robustness of our approach. Bayesian phylogenies and posterior probability values for all clades are presented in SI Appendix, Fig. S10. We used the software MrBayes (53) with the same model parameters that were previously used for the analysis of poly-G tract mutation profiles (21).