Written by: Guillaume Bernard
jD2Stat (previously known as JIWA*) is a Java package that utilises a series of D2 statistics to
extract k-mers (subsequences at defined k length) from a set of biological sequences, and generate
pairwise distances for each possible pair. This distance can be used directly for phylogenetic inference
using neighbour-joining.
*: This program is not to be confused with, and has no relation to, the proprietary trademark and software of Jiwa Financials, North Sydney, Australia.
The D2 methods included are:
- D2 based on exact k-mer counts (
D2) - D2S - similar to D2, normalised based on probability of occurrences of specific k-mers (
D2S) - D2∗ - similar to D2, normalised based on means and variance (
D2St) - D2n - extension of D2 to allow for n number of wildcards (n neighbourhood)
jD2Stat runs on all platforms that support the Java Runtime Environment JRE 1.7.x or higher. jD2Stat comes in a .jar file that can be invoked on command line directly under JRE, so no compilation or installation is necessary. In the rare occasion that you have no Java installed on your computer, Java is freely available from http://www.java.com/. Memory requirement varies depending on the number and length of the sequences, overall sequence similarity and the k-mer length, for obvious reasons. RAM of 1GB would be sufficient for most simple analyses. For instance, in our analysis of 5000 16S rRNA sequences, the memory usage is about 2.5GB.
jD2Stat accepts DNA and protein sequences in FASTA format. A file should contain a set of two or more sequences, and MUST have any one of these extensions: .faa, .fas, .fasta, .ffn, .fna or .frn. Both single- and multi-line FASTA format are accepted.
IMPORTANT: Note that jD2Stat considers only conventional nucleotide and amino acid characters in the analysis: A, C, G and T for DNA, and the standard 20 amino acids (single-letter code) for proteins. The presence of any non-conventional characters in the sequences will result in an error.
For each D2 analysis, jD2Stat generates a pairwise distance matrix in PHYLIP format, which can be
used as input for neighbor in PHYLIP to infer a phylogenetic tree using neighbour-joining.
jD2Stat can be invoked directly on command line:
java –Xmx<vmem> –jar jD2Stat_1.0.jar <options> -i <input> -o <output>
In following examples: input sequences in myseq.fas, output matrix file named myseq.matrix.
java –Xmx4g –jar jD2Stat_1.0.jar –d D2 –k 8 -i myseq.fas -o myseq.matrix
The above command runs D2 statistic on DNA sequences at k-mer length 8 with max memory set at 4GB.
java –Xmx4g –jar jD2Stat_1.0.jar –a aa –d D2St –k 4 -i myseq.fas -o myseq.matrix
The above command runs D2* statistic on protein sequences at k-mer length 4 with max memory set at 4GB.
java –Xmx8g –jar jD2Stat_1.0.jar –n 1 –k 8 -i myseq.fas -o myseq.matrix
The above command runs D2n (with neighbourhood n = 1) on DNA sequences at k-mer length 8 with max memory set at 8GB.
java –Xmx12g –jar jD2Stat_1.0.jar –a aa –d all –k 4 -i myseq.fas
The above command runs all three D2, D2S, and D2* independently on protein sequences at k-mer length 4 with max memory set at 12GB. Three distance matrices in separate files will be generated.
| Parameter | Description |
|---|---|
-a <a> |
<a> is the alphabet/character type of the sequences: dna (default) or aa (amino acid). |
-d <d> |
<d> is the choice of D2 method: D2 (default), D2S, D2St or all to run all three methods of D2, D2S, and D2*; –d all yields all three matrices in separate files even without –o specification. |
-h |
Display all available options on the screen |
-k <k> |
<k> is the length of k-mers to be used in the analysis. Default: 8 |
-n <n> |
<n>is the neighbourhood value of n in D2n. For instance, –n 1 specifies D2n=1. When -n is specified, only D2n is run regardless of specification in -d. |
-t <t> |
<t> is the number of threads to use in the analysis. Default: 4 |
If you use jD2Stat or D2n in your work, please cite:
Chan CX, Bernard G, Poirion O, Hogan JM and Ragan MA (2014). Inferring phylogenies without multiple sequence alignment. Scientific Reports 4: 6504. http://dx.doi.org/10.1038/srep06504.
This replaces the the original URL indicated in the paper, hosted on bioinformatics.org.au.