phylogenetics.dot

digraph G {
  d2toptions ="-ftikz --tikzedgelabels --styleonly --autosize -traw -c --straightedges";
  
  d2tfigpreamble = "\tikzstyle{model}=[draw=yellow!50, very thick, rectangle,fill=yellow!20], \tikzstyle{objective}=[draw=gray!50, very thick, rectangle,fill=gray!20], \tikzstyle{data}=[draw=red!50, very thick, fill=red!20], \tikzstyle{algo}=[draw=blue!50, very thick, fill=blue!20,ellipse], \tikzstyle{treestyle}=[draw=green!50, very thick, fill=green!20];";
   
  d2tdocpreamble = "\tikzset{every text node part/.style={align=center}}"
   
  mindist = 0.5;

  graph [fontname = "arial", ranksep="0.3", nodesep="0.4"];
  node [fontname = "arial"];
  edge [fontname = "arial", arrowhead="vee"];

  sequences [ label="DNA\\\\sequences", style=data];
  
  sequence_alignment [ label="MSA\\\\algorithm", style=algo ];
  
  msa [ label="Multiple\\\\sequence\\\\alignment\\\\($D$: $n \times L$)", style=data];
  distances [ label="Genetic\\\\distance\\\\matrix\\\\($\hat d$: $n \times n$)",style=data];
  
    felsenstein_pruning [ label="Felsenstein's\\\\pruning\\\\algorithm\\\\$\\Pr(D|T,Q)$", style="algo" ];
    substmodel [ label="CTMC\\\\Substitution\\\\model ($Q$)", style=model];
    treeprior [ label="Model-based\\\\tree prior\\\\$f(T|\\theta)$", style=model];
    matrix_exponentiation [label="Matrix\\\\exponentiation\\\\$P(d)=\exp(Qd)$", style="algo"];
  
    maximum_likelihood [label="Maximum\\\\likelihood\\\\tree", style=treestyle ];
  
    bayesian_inference [label="Bayesian\\\\MCMC\\\\algorithm\\\\$f(T,Q,\theta|D)$", style=algo ];

    bayesian_posterior [label="Posterior\\\\distribution\\\\on trees", style=treestyle];
    ml_objective [label="Maximum\\\\likelihood\\\\objective", shape=rect, style=objective];
  
  ls [label="Least-squares\\\\tree", style=treestyle];
 
  distance_correction [label="distance\\\\correction\\\\$\hat d_{ij}=f_{corr}(p_{ij})$", style=algo]

  ls_objective [label="Least-squares\\\\objective", style=objective];

  nj_algorithm [ label="Neighbour-joining\\\\algorithm\\\\$O(n^3)$", style=algo];
  upgma_algorithm [ label="UPGMA\\\\algorithm\\\\$O(n^2)$ or $O(n^3)$", style=algo];
  
  parsimony_algorithm [ label="parsimony\\\\algorithm\\\\$O(nL)$",style=algo];
  parsimony_objective [label="Maximum\\\\parsimony\\\\objective", shape=rect, style=objective];
  maximum_parsimony [label="Maximum\\\\parsimony\\\\tree\\\\(unrooted)", style=treestyle];
    
  nj_tree [ label="Neighbour-joining\\\\tree\\\\(unrooted)", style=treestyle];
  upgma_tree [ label="UPGMA\\\\tree\\\\(rooted, ultrametric)", style=treestyle];

  substmodel -> distance_correction;

  substmodel -> matrix_exponentiation; 
  
  matrix_exponentiation -> felsenstein_pruning[label="Transition\\\\probabilities,\\\\$P(d)$"];

felsenstein_pruning -> bayesian_inference; 
treeprior -> bayesian_inference[weight=2];

felsenstein_pruning -> ml_objective; 
ml_objective -> maximum_likelihood[label="heuristic\\\\tree\\\\search"];

parsimony_algorithm -> parsimony_objective;
parsimony_objective -> maximum_parsimony[label="heuristic\\\\tree\\\\search"];

  sequences -> sequence_alignment

  sequence_alignment -> msa;

  msa -> distance_correction[label="p-distances\\\\($p$: $n \times n$)"];
    
  distance_correction -> distances;

  msa -> parsimony_algorithm
  msa -> felsenstein_pruning
  
  bayesian_inference -> bayesian_posterior;

  distances -> nj_algorithm
  distances -> upgma_algorithm
  distances -> ls_objective;
  
  ls_objective -> ls[label="heuristic\\\\tree\\\\search"];
  
  nj_algorithm -> nj_tree

  upgma_algorithm -> upgma_tree

}