Motif Centrality Analysis of ChIP-Seq (MOCCS) is a method for for clarifying DNA-binding motif ambiguity.
Given ChIP-Seq data of any DNA-binding proteins including transcription factors, MOCCS comprehensively analyzes and describes every k-mer that is bound by the DNA-binding proteins.
MOCCS (version 2.0) is written in Perl and R.
MOCCS was tested on Perl version 5.26.2 and R version 3.6.0.
- 2019/06/14: Version 2.0.
- Added MOCCS2 score in the output (See below for MOCCS2 score).
- MOCCS now generates a plot of cumulative relative frequency curves of top k-mer sequences based on MOCCS2 score, in addition to that based on AUC.
- 2016/04/15: Version 1.7.
- Added 'stranded' option. With this optin, MOCCS will count k-mers on the forward strand and will not reverse-complement k-mers. This is useful when analyzing RNA-binding proteins (RBPs) data (e.g. CLIP-Seq data).
- Added '--low-count-threshold' option.
perl MOCCS.pl -i <string> -k <int>|--regex <string> \
[--label <string>] [--mask] [--threshold <float>]
-i <string>: FASTA file containing fixed-length sequences around transcription factor binding sites (TFBSs) identified by ChIP-Seq. If the file name ends with.gz, the file is treated as gzipped FASTA. If-i stdinis specified, MOCCS receive FASTA from the standard input.- We recommend to use genomic sequences ±350 bp around the TFBSs.
-k <int>: Length of k-mers to calculate AUCs. Mutually exclusive with--regex.- We recommend to set k to 5, 6, 7, or 8.
--regex <string>: (Optional) Regular expression (e.g. CANNTG). When this option is specified, MOCCS calculates AUCs for only k-mers corresponding to the regular expression. Currently, 'N' is interpreted as {A,C,G,T}. Mutually exclusive with-k.--label <string>: (Optional) Label used as the prefix of output files. Default is 'MOCCS_result'.--mask: (Optional) Mask lower-case characters in fasta file as 'N'.--threshold <float>: (Optional) Only print k-mers with AUCs higher than this value.--stranded: (Optional) Count k-mers on the forward strand. This is useful when you analyze sequences bound by RNA-binding proteins (RBPs) (e.g. CLIP-Seq data). By default, MOCCS sums counts of a k-mer and its reverse-complement before calculation of the AUC score.--low-count-threshold <float>: (Optional) Set the low-count threshold. Those k-mers whose counts are less than this value will be filtered. By default, MOCCS set the low-count threshold toN * (w - k + 1) / 4^k.
Note that MOCCS_visualize.r must be located on the same directory as MOCCS.pl.
${label}.auc_count.txt: Count, AUC, and MOCCS2 score for k-mer sequences.${label}.mid.gz: Aggregated appearance counts around TFBSs for k-mer sequences. For debug.${label}.crf.gz: Cumulative relative frequencies for k-mer sequences. For debug.${label}_${k}-mer_auc_plot.pdf: Plot of cumulative relative frequency curves for top 10 k-mer sequences based on AUC.${label}_${k}-mer_MOCCS2score_plot.pdf: Plot of cumulative relative frequency curves for top 10 k-mer sequences based on MOCCS2 score.${label}.MOCCS_visualize.log.txt: R log for creating${label}_${k}-mer_auc_plot.pdf.
perl MOCCS.pl -i test_data/test_701bp.fa -k 6 --label test_out_0/test_out_0
perl MOCCS.pl -i test_data/test_701bp.fa -k 6 --label test_out_1/test_out_1 --threshold 10
perl MOCCS.pl -i test_data/test_701bp.fa -k 6 --label test_out_2/test_out_2 --mask
perl MOCCS.pl -i test_data/test_701bp.fa --regex CANNTG --label test_out_3/test_out_3
perl MOCCS.pl -i test_data/test_701bp.fa --regex CANNTG --label test_out_4/test_out_4 --mask
perl MOCCS.pl -i test_data/test_701bp.fa -k 6 --label test_out_5/test_out_5 --stranded
perl MOCCS.pl -i test_data/test_701bp.fa -k 6 --label test_out_6/test_out_6 --low-count-threshold 100
MOCCS (since version 2.0) calculates AUC and MOCCS2 score for each k-mer sequence.
AUC is the area under the cumulative relative frequency curve of the appearance of each k-mer sequence against distance from TFBSs, which represents sharpness of the histogram of its appearance around TFBSs. Note that the AUC becomes larger if the shape becomes sharper.
Let be the appearance count of each k-mer sequence at the position
bp away from TFBSs and
, then the cumulative relative frequency distribution
for the k-mer sequence is calculated as follows:
and its AUC is calculated as follows:
.
MOCCS2 score is a relative value of AUC normalized by the standard deviation (SD) at its appearance count. We recommend to use MOCCS2 score since MOCCS2 score is more robust to low appearance count than AUC is.
Let be the appearance count of the k-mer sequence and let
be the size of the analyzed window where k-mer sequences are sought at around ChIP-seq peak positions, MOCCS2 score is defined as follows:
The calculation of [SD of AUC] was mathematically derived as follows:
If a k-mer sequence appears only once at a random position within the window, its coordinate follows the uniform distribution
, whose variance is known to be
. Because i) AUC is calculated by subtracting
from the coordinate and ii) constant subtraction does not affect variance of probability distributions, variance of AUC is also
.
Next, assume that a k-mer sequence appears
times at random positions within the window. The variance of the sum of their coordinates becomes
, because variance of sum of random variables that follow the same probability distribution is proportional to the numbers of the variables. Then, because AUC is calculated by dividing the sum of their coordinates by
, the variance of AUC is
if the appearance count is
In the current implementation of MOCCS2, was set to
floor((sequence_length - 1)/2) + 1 – floor(k / 2). Note that since the value of is the same for every k-mer sequence, the relative ranks based on MOCCS2 score among k-mer sequences are not affected by the choice of the definition of
.
- Ozaki H, Iwasaki W. MOCCS: Clarifying DNA-binding motif ambiguity using ChIP-seq data. Computational biology and chemistry. 2016 Aug 1;63:62-72. PubMed Preprint
- A paper describing MOCCS2 score will be published soon.
Haruka Ozaki harukao.cb@gmail.com
Copyright (c) [2016] [Haruka Ozaki] This software is released under the MIT License, see LICENSE.txt.