• Written by Jared Simpson.
• Currently only has a GitHub repository.
• Paper: Efficient construction of an assembly string graph using the FM-index [1]

• Uses the Burrows-Wheeler Transform(BWT)/Ferragina—Manzini(FM)-index to build a string graph.

• Nodes are reads (reads that are substrings are condensed into superstrings). Edges are overlaps between reads, and the non-overlapping prefix is stored in the forward edge and suffix is stored in the backwards edge.
• Condenses repeats like a de Bruijn graph.
• More expensive to construct than a de Bruijn graph.

• Like BWA, it uses the FM-index, which is a compressed method of inferring the suffix array.
• The Burrows Wheeler transform B_X is an array of the last characters in the alphabetically sorted suffix array.
• The FM-index (two data structures: 1. C_X(a) be the number of symbols in X that are lexographically lower than the symbol a, 2. Occ_X(a, i) be the number of occurrences of the symbol a in B_X[1, i], the ) allows substring searching and can be extended to construct the string graph.

Installation of SGA from the GitHub is a major pain, because it has so many dependencies. It needs

• google-sparsehash (also needed for Abyss)
• hoard
• bamtools, which is a particular pain to install because it uses
  • cmake (newer than the version installed on campusrocks), and cmake does not support installing stuff anywhere other than in the initial directory or root-access-only directories, so we can't do the usual “configure –prefix=/campusdata/BME235”

Once cmake, bamtools, hoard, and google/sparsehash are all installed as best they can be, then the UCSC-install.bash script can be run to install SGA. Actually running SGA is still to be tested!