ClusterMerge

Bottom up merging to cluster similar protein sequences.

Install dependencies:

$ sudo apt-get install bazel

Bazel will use the default C++ compiler on your system.

Building

Linux (dbg): bazel build -c dbg src:clustermerge

Note: I'm not sure how to increase the default stack size limit on Mac, so currently clustermerge will only work on Linux. Mac (dbg): bazel build -c dbg --spawn_strategy=standalone src:main_dsym

Mac/Linux (opt) bazel build -c opt src:clustermerge

Running

First, set the thread stack limit to 64MB:

$ ulimit -s 65532000

This is to accomodate some alignment routines that stack allocate a lot of data.

Clustermerge usage:

$ ./bazel-bin/src/clustermerge file1.fa file2.fa

$ ./bazel-bin/src/clustermerge -h  # view help

By default clustermerge will use as many threads as are available on your system. You can optionally provide a list of your datasets to cluster to avoid typing so much crap on the command line:

# data.json
[
  "file1.fa",
  "file2.fa",
  "fileN.fa"
]

and run with

$ ./bazel-bin/src/clustermerge -i data.json

PAM matrices for optimal alignments using SWPS3 are provided in the repo under data/matrices/json, if you run the tool elsewhere you will need to use the -d / --data_dir option to provide the path.

If you have complete all vs. all results you want to compare against, tools has a compare script:

python2 tools/compare_results.py path/to/AllAll/ path/to/clustermerge/output_matches/

Distributed Version

bazel build -c opt //src/dist:dist_cluster