Skip to content

Latest commit

 

History

History
130 lines (84 loc) · 5.96 KB

README.md

File metadata and controls

130 lines (84 loc) · 5.96 KB

Havec

This is a implementation of Efficient de Bruijn graph construction for genome assembly using a hash table and auxiliary vector data structures.

Pre-requisite

To use this software you need g++ (GCC 4.7 or later to support -std=c++11) must be pre-installed. For more information about the compiler here.

Build instruction

Havec build now depends on make. If make is not installed install it using following

  
  sudo apt-get install build-essential
  

To build the Havec

  # you can clean the project by `make clean`
  make havec # only `make` will also suffice

It will generate the executable Havec.out for you in the same directory.

Sample Test run

Sample input files are provided into the input_files. A simple test run is also possible after building the havec. Run followig script to go through a test run:

  sh test_run.sh

A dated html file will be served upon the script ends, which will contain all the test set run results in tabular form. Open the file in your convenient browser.

Some of the example are given following:

./Havec.out dataset_1_7GB.fa 32 8 NC NP 138568143 1.4
./Havec.out dataset_1_9GB.fa 32 8 NC NP 113775137 1.4
./Havec.out Ecoli_MG1655_s_6_1_bfast.fasta 32 8 NC NP 13795561 1.4
./Havec.out Ecoli_MG1655_s_6_2_bfast.fasta 32 8 NC NP 133958572 1.4
./Havec.out NA19240_GAIIx_100_chr21.fasta 32 8 NC NP 113775137 1.4
./Havec.out Human1_95G_CASAVA1.8a2_NCBI37_18Jan11_chr19.sorted.fasta 32 8 NC NP 222896820 1.4
./Havec.out Human1_95G_CASAVA1.8a2_NCBI37_18Jan11_chr21.sorted.fasta 32 8 NC NP 138568143 1.4

Unit testing

Unit testing is been integrated in order to maintain the integrity of this software. CUnit is been used for unit testing. For Ubuntu makefile is already configured to install this unit testing library.

To install the unit testing depedencies

make install_deps

To remove the unit testing dependencies

make remove_deps

To run the unit tests

make test
./test_common.out

Havec Supported parameter

This software accept a number of parameter to run on different data-set. They all are listed following. Those who are impatient, can have a look at the following,

  ./Havec.out [file-path] 
              [k-size] 
              [Number of hash function] 
              [track-cutoff] 
              [print-verbose] 
              [Hashtable min size] 
              [Multiplying-factor]

All the parameter ar listed following

  1. File-path
  2. k-size
  3. Number of Hash function
  4. Track Cutoff
  5. Print-verbose
  6. Hashtable min size
  7. Multiplying factor

1. File-path Input sequence file in fasta or fastq format
2. k-size The size of the khmer. Until now we support upto 32.
3. Number of Hash function To insert the khmer in our Hashtable data-structure we use a number of hash function upon collision. As the number of this hash function is higher, less collision are assumed to occur. Until now we support upto value 8.
4. Track-Cutoff In genome assembly, some algorithm requires the cutoff value of a khmer, some doesn't. To track the cut-off value this parameter is responsible. To enable Cutoff tracking use C, otherwise use NC.
5. Print-verbose To print out the activity enable this. But this will flood you with the messages. Turn it on for the devlopment purpose & for extra log. To enable use P otherhand use NP.
6. Hashtable min size This is the value which will be the used for the hashtable. Exactly this value will not be used, next prime of this value will be used as the hash-table size for more collision proof.
7. Multiplying Factor This value is in aid to take as a multiplying factor in hashtable size. This value will be Multiplied with the 6th parameter to increase the hashtable size. This value is just for ease of usage.

Using the Result Analyzer

Running the Havec tool will output the result in a human readable text file (file with a txt extension). Created output filename will follow the input parameter. Each run of the tool single text formatted output. In each output file there will be verbose data, to sort-out the exact very necessary data a result_analyzer tool is attached with the repository.

To use the result_analyzer tool first make a all_results empty directory inside the result_analyzer directory. Then put all the resulted output text file into this directory & run the parse.sh tool using the following command

sh parse.sh

This parse.sh will create a html file named after the current system date & time. This html file will contain a table with all the data in the all_results directory.

A brief story of the Result Analyzer

The result analyzer actually a java program, which parse out all the results inside a directory. It parses every single line inside the text file which is the output of the Havec executable. So the input to the java program is a directory containing all the resulted text file. The input directory in our case is all_results directory. Running the parse.sh script will generate the output in a dated html (html file with the name of current date) file inside the result_analyzer folder.