#ARKQuikr
ARKQuikr is an AggRegation of Kmers modification of the QUadratic, Iterative, K-mer based Reconstruction technique (Quikr) that utilizes sparsity promoting ideas from the field of compressed sensing to reconstruct the composition of a bacterial community (when the input data is a FASTA file of 16S rRNA reads). This extremely fast method comes with a several databases that can be custom trained. Typically reconstruction is accurate down to the genus level.
#What does this repository contain?
This repository is a Julia implementation of the ARKQuikr algorithm. For a Matlab version of this code, see [this website] (http://www.kth.se/en/ees/omskolan/organisation/avdelningar/commth/research/software).
- Mac OS X 10.6.8 or GNU/Linux
- 4Gb of RAM minimum. Absolutely necessary.
- gcc that supports OpenMP
- dna_utils must be installed
- Mac OS X 10.6.8 (what we have tested)
- GCC 4.7 or newer. (gcc 4.2 did not work, and is the default installation)
- OpenMP libraries (libgomp, usually comes with gcc)
- GCC 4.7 or newer
- OpenMP libraries (libgomp, usually comes with gcc)
After cloning and installing the dna_utils repository, just clone this repository. As the code contained herein are Julia scripts, no compilation is necessary.
The code only works on FASTA files (not FASTQ or any other format). Here's an example using 10 clusters:
julia ARK.jl -i /path/to/FASTA.fa -o /path/to/Output.tsv -n 10
Another example using deterministic clustering, 5 clusters, and the SEK training database:
julia ARK.jl -i /path/to/FASTA.fa -o /path/to/Output.tsv -n 10 -c Deterministic -t SEK
Other options are available, see julia ARK.jl -h.
The output format is consistent with the CAMI challenge and is similar to the output produced by MetaPhlAn.
If your installation of dna_utils results in the executable being located in a non-standard location, specify this location using the option -k /path/to/./kmer_counts_per_sequence
It is very important that your installation of BLAS matches the architecture of your hardware (if not, significant increases in computation time might be observed). We recommend using OpenBLAS.