(C) Johannes Soeding, Markus Meier, Christian Roth
PEnG-motif is an open-source software package for searching motifs (position specific weight matrices, PWMs) in a set of DNA sequences.
As the core algorithm operates on kmers, the runtime is practically independent of the number and size of the input sequences. This makes PEnG-motif suitable for de-novo motif discovery on large sequence sets.
The easiest way to get PEnG-motif on your mac is by downloading our precompiled binaries. If you choose that way, please download peng_motif_macOS.zip from our latest release and unpack it.
After unzipping you will find a new binary peng_motif which you can call from command line by the full path to the binary ./peng_motif, or directly by typing peng_motif, if you moved it to a location in your shell path.
The easiest way to get PEnG-motif on your 64bit linux computer is by downloading our precompiled binaries. If you choose that way, please download peng_motif_linux_amd64.zip from our latest release and unpack it.
- CMake 2.8.12 or later
- gcc >=5.2 | gcc =4.9 (when manually setting CXXFLAG flag -std=c++1y)
Download the source code archive from our latest release.
Unzip the source code, and navigate into the freshly unzipped folder in your terminal.
This code will compile and install PEnG-motif
INSTALL_DIR=/path/to/an/installation/directory/of/your/choice
mkdir build
cd build
cmake .. -DCMAKE_INSTALL_PREFIX=$INSTALL_DIR
make && make installYou may want to add the path of $INSTALL_DIR/bin to your shell PATH, to simplify usage.
The repository also contains helper scripts - if you want to use them, you also need:
- python>=3.5
- numpy
- BaMMmotif2
PEnG-motif finds enriched sequences in a fasta file and writes the motifs to an output file in meme format.
If the parent directory of of peng_motif is in your shell path, the simplest way to use PEng-motif is:
peng_motif <path/to/input.fasta> -o output.memeFor a list of all available options, please see the output of peng_motif -h.
If you didn't put PEnG-motif in your shell path, you have to enter the full path to the peng binary, e.g. /path/to/peng/peng_motif -h.
The PEnG-motif algorithm runs several phases; understanding the output printed to the console will help interpreting the final results.
In the first phase, the occurrences of all 4^(pattern_length) kmers are counted. The expected number of pattern counts are calculated under the assumption of a homogenous Markov model. From the observed and expected counts a z-score is computed for each kmer.
pattern observed enrichment zscore
CACTAG 890 2.40 27.00
GCCACC 2456 1.68 26.08
For each pattern that passes a predefined z-score threshold, the number of occurences, the enrichment of observed occurences over the expected and the calculated z-score are reported.
In the second phase the selected base patterns from phase 1 are iteratively optimized by degenerating single nucleotides if the degenerated pattern achieves a higher score. By default we optimize a function based on mutual information of observation and expectation.
CACTAG 890 2.40 -0.036173
CWCTAG 1687 1.89 -0.036744
optimization: CACTAG -> CWCTAG
For each iteration the current IUPAC pattern, the number of observed counts, the enrichment over the expected counts and the optimization score are printed. Once the optimization runs into a local optimum, the original base pattern and its optimal IUPAC pattern are reported.
If the optimization runs into a pattern that has already been seen in a previous optimization, the optimization stops and the base pattern is removed.
At the end of the optimization phase, all IUPAC patterns, their occurrences, enrichment over the expected counts, and the calculated z-scores are reported.
pattern observed enrichment zscore
CWCTAG 1687 1.89 26.51
GYCAYC 4376 1.55 29.29
In phase 3, only the best scoring IUPAC PWMs are retained and are converted to PWMs.
In the final phase, an expectation-maximization algorithm sharpens the PWMs. PWMs that have strong detectable overlaps are merged to form longer PWMs. The so optimized PWMs are written to the output file in meme format.
- PEnG-motif has many parameters that can be set to tune the results. The default parameters are suitable for most DNA binding factors. For RNA binding factors with short motifs a pattern length of 6 or 8 often yields better results.
- If your sequences are strand specific, pass the option
--strand PLUSto avoid mixing the counts of reverse complemented base patterns. - The memory requirement and runtime scales exponentially with the pattern length. It is not recommended to run with base pattern lengths larger than 12.
The shoot_peng.py python script can be used to annotate motifs according how well they can distinguish the given sequences from randomly generated ones.
Benchmarking requires BaMMmotif2.
PEnG-motif is included in our bamm-suite docker image soedinglab/bamm-suite.
After navigating into the directory with your fasta sequences use:
docker pull soedinglab/bamm-suite
docker run -v $UID:$GID -v $PWD/data soedinglab/bamm-suite peng_motif <fastafile.fa> <other options>
The PEnG-motif can be modified and distributed under the GPL-3.0 License.
PEnG-motif uses shared code from BaMMmotif. Many thanks to the developers of BaMMmotif!