Multiz Threaded-Block-Alignment

http://www.ncbi.nlm.nih.gov/pubmed/15060014
    MULTIZ uses pairwise alignments produced by BLASTZ. These are filtered to select the best match in the genome, so that each base in human aligns to a single base in the other species. We tried two different methods to filter these alignments, axtBest (Schwartz et al. 2003a) and the “net” approach described in Kent et al. 2003. In general axtBest aligns slightly more human bases, and the net approach is somewhat better at identifying orthologs. We prefer the latter.

http://www.bx.psu.edu/miller_lab/
http://www.bx.psu.edu/miller_lab/dist/tba_howto.pdf

The starting material for running TBA includes:
1. A set(s) of sequences from different species
2. An evolutionary tree describing the relationship of these species
3. A parameter file describing how to run blastz for different species. For example,
you might want to treat alignments involving non-placental mammals differently
from placental mammals. This file is optional.

3 Overview of running TBA
There are typically three steps involved with generating a multiple alignment:
1. generating a series of pair-wise alignments to “seed” the multiple alignment process (This process is performed with the program all bz, which is essentially a program that executes a series of blastz commands., outputs a "lav" format that is converted to MAF)
2. generating the multiple alignment (tba, calls multiz)
3. “projecting” the alignment onto a reference sequence

http://genome.csdb.cn/cgi-bin/hgTrackUi?hgsid=446052&c=chrX&g=mostConserved28way