SAPF: Statistical Aligner and Phylogenetic Footprinter
Rahul Satija, Lior Pachter, Jotun Hein

USAGE:

After initial setup (see SETUP) The program can be run with a command
of the form:

	perl Sapf.pl alignment tree model [-noparams]

Results are stored in the file results.prob . The results consist of
base-by-base probabilites, for one reference sequence, that the
basewas generated by a "slow" state in the model. The estimated
parameter set is reported in params.output

example usage:
	
	perl Sapf.pl TestThree.aln example.tree models/FullModel3.composed

This command runs SAPF on one of the simulated alignments from the
first simulated dataset. The number of sequences has been reduced to
three to allow the example to run quickly. In the alignment file
TestThree.aln, capital letters represent bases that were simulated as
functional.

ARGUMENTS:

alignment: an initial multiple sequence alignment of the sequences to
be analyzed in clustalw format. The alignment needs to be only a rough
approximation, SAPF will calculate a probability distribution of all
alignment columns within 50bp of this first alignment. A fast,
heuristic aligner like ClustalW is ideal. While tree and model files
can be located anywhere, the alignment file should be located in the
directory with the perl scripts.

tree: a tree file, in newick format, relating the species in the
alignment. The tree must contain all species in the alignment, but
additional species are OK (SAPF automatically prunes the tree to
consider only relevant species). The reference sequence, for which
results are reported, is specified by the left-most node of the tree

model: a model file generated by phyloComposer [1]. phyloComposer
generates a multiple HMM based off of a branch HMM and a guide
tree. phyloComposer is available as part of the DART package at
biowiki.org/DART if you would like to make your own models, but I have
included sample models for 4, 3, and 2 sequences in the models
directory.

-noparams (optional) : skips the parameter estimation process and uses
the inital guess.

SETUP: 

After installing (see INSTALL) there are 2 files to setup with SAPF.

1. sapf.dir : to save time and memory, SAPF pre-computes emission and
transition probabilities in advance. Edit the file sapf.dir to specify
the directory where they will be stored

2. initial.param : holds the initial parameter guess. The order is the
same as Table 3 in the Supplementary Material. The format is identical
for the output file params.output

ADVANCED:

Corner Cutting. SAPF first calculates a set of pairwise alignments and
then restricts a path through the multidimensional DP matrix based on
these. Pairwise homologies below a certain probability cutoff are
discarded. The cutoff is currently set to ~ln(10) = 2.3. This means
that all pairwise alignment columns less than 1/10 as probable as the
Viterbi alignment column will be discarded. Increasing this cutoff,
set by the $ProbCutoff variable in the beginning of Sapf.pl, will
increase the number of alignment columns considered but will decrease
the running time. While we currently use the Viterbi alignment for a
reference while corner cutting, we hope to replace this with the MPP
alignment (see manuscript) in a future version.

CONTACT

Please address all questions and comments to Rahul Satija at
satija@stats.ox.ac.uk

REFERENCES
[1]	Holmes, I. (2007). Phylocomposer and Phylodirector: Analysis and Visualization of 
		Transducer Indel Models. Bioinformatics.