Dear Conrad,

First, a disclaimer. HyPhy was not intended to be a fast topology inference package (the exception is NJ, which is fairly quick). You can do some topology inference with codon models; I would recommend getting a starting NJ tree and then branch swapping (using TopologyInference:BranchSwap.bf), preferrably on a cluster.

To answer your question about including NJ code into your analysis: look at the NeighborJoining.bf in TemplateBatchFiles and, specifically at the InferTreeTopology function for the general case (which prompts for models and such). This function depends on the data being in the [bf]ds[/bf] dataset object and functions in distanceMethodNPBootstrap.bf.

The minimalist function for using NJ (based on SingleBreakPointRecomb.bf) will need to include the following functions:

InitializeDistances
ComputeDistanceFormula
TreeMatrix2TreeString
InferTreeTopology

The sequence of calls to build a tree would be

[code]
DataSetFilter filteredData = CreateFilter ... /* make your filter */
InferTreeTopology (0); /* do NJ and return internal tree representation */
treeString=TreeMatrix2TreeString(0); /* convert to string */
Tree givenTree = treeString; /* build a tree object */
[/code]

HTH,
Sergei