yexianingyue / FFP

Feature Frequency Profile(FFP) two core programs

FFP

Feature Frequency Profile (FFP); two core programs' codes deposit for,
"A genome Tree of Life for the Fungi kingdom", JaeJin Choi and Sung-Hou Kim (2017), PNAS.
"Whole-proteome tree of life suggests a deep burst of organism diversity", JaeJin Choi and Sung-Hou Kim (2019), PNAS.

Requirements

• GCC(g++) version 4.7.1+
• Google sparse hash library. Look here: https://github.com/sparsehash/sparsehash
• zlib version 1.2.8+. Look here: http://www.zlib.net/

Welcome to have questions regarding the program codes, bug report, and usage.

• Please contact JaeJin Choi ([e-mail])

Tutorial / Supplement

• A tutorial you can walkthrough here:
• Additional fungi study supplement files (e.g., tree newick and divergence matrix) are here:

Two versions available: Binary, and Text(string)

• Latest update: 2v.3.0 (2019-12-3); see the , list all versions

FFP_binary version;
FFP_text version;

Old text based FFP 2v.1.0 (before 2018-8);

1. FF Profiler; FFP_x.cpp

Compile: g++ -std=c++11 -o (execute name) (this script) -lz
Run example: [Program path][options][input file path][output file path]
When preparing input files save different taxons in separated files

[Arguments]

• -h
  Show options
• -v
  Show version
• -s [INT]
  Feature size (l-mer)
• -e [INT]
  Feature size (l-mer) range end. Functional only when measuring vocabulary size using with -V
• -a
  Takes 20 amino acids sequences as inputs
• -c
  Convert and accept nucleotide bases (A, G, C, T) into RY bases (Purine, Pyrimidine)
• -k [STR]
  Manual input of alphabet base string. For example input 'HJKL' is ['H', 'J', 'K', 'L'] bases
• -r
  Disable reverse complement accounting. Any bases set other than AGCT code will disable reverse complement accounting
• -n
  Output frequency, i.e., feature count / total feature count
• -u
  Accept masked (lower confidence; softmasked) letters which are in lower cases in FASTA format
• -V
  Count vocabulary size at given range of feature length (l-mer)
• -b [LONG LONG]
  Bottom count limit. Remove features, the count below [-b]
  Default = 1
• -t [LONG LONG]
  Top count limit. Remove features, the count above [-t]
  Default = 0 = maximum
• -B [Double]
  Bottom entropy limit. Remove features, the string entropy below [-B]
  Default = 0.0
• -T [Double]
  Top entropy limit. Remove features, the string entropy above [-T]
  Default = 1.0 = maximum

[Note]

When using option [-a], amino acids input, [-r] turns on automatically that disable reverse complement accounting, because peptide sequences are single strand that have direction (start code -> stop codon). However, nucleotide sequences input is considered as a double helix, of 'forward' and 'backward' strands, and account reverse compliment as default option.

Use [-r] option that turn off reverse compliment accounting if input is single strand nucleotide sequences such as ribosomal DNAs.

Use [-V] option along with [-s], [-e] and feature filtering arguments to estimate a range of optimal l-mer. In general, use [-b 2], remove any feature count less than 2, to determine a l-mer where vocabulary complexity started to maximize. FF Profiler will determine and stop to a point where vocabular size drops in range of [-s] and [-e], or without giving [-e] (default value is 0) will continue search the point by incrementing the l-mer.

• Heuristically, an optimal l-mer for amino acids was 13 and for nucleotides was 23 or 24 (but indefinitive). A population's optimal l-mer likely follow the majority optimal l-mer.

[Input]

FASTA format peptide or nucleotide sequence files.

[Output]

zlib compressed Feature Frequency Profile.

2. JSD Calculator; JSD_matrix_x.cpp

Compile: g++ -std=c++11 -pthread -o (execute name) (this script) -lz
Run example: [Program path][options][input files path] > [output file path (standard output)]

[Arguments]

• -h
  Show options
• -t [INT]
  Number of threads for multiprocessing
  Heuristically, an adequate thread number is the number of cpu cores - 1. Default is 5
• -r [PATH]
  Input previous matrix, and add more items to the matrix without calculating a whole. File name is used as an item name but no longer than 10 characters.
• -d
  Output Jensen-Shannon distance matrix instead of Jensen-Shannon divergence matrix which is default option.
  Square root(JS divergence) = JS distance
• -s
  Output a symmetric matrix instead of a low triangular matrix which is default output.

[Note]

[-r] input low triangular divergence or distance matrix. This requires all pair-wise output of 'FF Profiler'

[Input]

The output files of 'FFP_compress' which are zlib compressed Feature Frequency Profiles (FFPs)

[Output]

Standard output of a low triangular Jensen-Shannon divergence or distance matrix in PHYLIP format (not .tsv).
Support a symmetric matrix output using [-s].

Limitation

Generally, longer feature lengths (l-mer) consume more memory and time.
In fungi proteome study the largest proteome has 35,274 proteins containing 10,866,611 amino acids, this program worked for feature length up to 24 amino acids.