Skip to content
/ CEMIG Public

A new motif prediction algorithm called CEMIG (cis regulatory motif influence using de Bruijn graph)

Notifications You must be signed in to change notification settings

OSU-BMBL/CEMIG

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

47 Commits
 
 
 
 
 
 
 
 

Repository files navigation

CEMIG (Cis rEgulatory Motif Influence using de Bruijn Graph)

Overview

CEMIG is a new motif prediction algorithm which takes k-mer as the basic unit, finds motif seeds by introducing hash table and describing the relationship between k-mers constituting motif by using de Bruijn graph model, and then merges and extends among motif seeds, so as to more accurately predict the transcriptional motifs in ATAC-seq data.

image

Figure showns the illustration of the CEMIG framework.

(A) Determines the P-values of k-mer in background data utilizing Markov models.

(B) Constructs Hamming distance graph ($G$) and DBG ($G_{DB}$) graphs using k-mer.

(C) Clusters k-mera on $G$ to form $G_C$, merging same-cluster k-mers from $G_{DB}$.

(D) Identifies motifs via path extension in $G_C$.

How CEMIG works

CEMIG framework encompasses four stages:

1. P-value calculation for k-mers

Initially, CEMIG evaluates input sequences (footprints) to determine k-mer P-values using a Poisson distribution. This is informed by nucleotide frequencies estimated via zero to second-order Markov models.

2. Construction of hamming distance graph and de Bruijn graph

CEMIG constructs a Hamming distance graph ($G$) and a de Bruijn graph ($G_{DB}$) using k-mers. The k-mers are sorted by ascending P-values and classified into three tiers, which facilitates the construction of the Hamming distance graph and the $G_{DB}$.

3. Detection of k-mer clusters

CEMIG detects k-mer clusters through graph clustering on the Hamming distance graph and constructs a secondary directed graph (digraph) by amalgamating vertices from identical clusters in the $G_{DB}$.

4. Motif identification

CEMIG forecasts motifs and their respective lengths by extending paths within the digraph. It employs a greedy algorithm for path extension, starting with an ‘uncovered’ cluster vertex with the highest f(•) value and sequentially adding vertices from edges with maximum weight. This process continues until the path reaches the desired length or three k-mer vertices have been added in the same direction. The starting cluster and other cluster vertices on the path are then considered ‘covered’. CEMIG outputs the identified paths and iterates the procedure until all clusters are covered.

Sequence set

The sequence set refers to the collection of DNA sequences that are used as input data for motif discovery algorithms. The sequence set is specifically derived from ChIP-seq data or ATAC-seq data. The ChIP-seq data usually includes a narrow peak file in FASTA format. For ATAC-seq data, either a narrow peak file or a footprint file in FASTA format is used as input for the CEMIG algorithm to identify DNA binding motifs.

Installation

Enter the folder code and type make then the compiled codes are within the same directory as the source.

cd code/
make clean && make

Motif prediction using CEMIG

cd code/
./cemig -i [INPUT_FILE]

For Example:

./cemig -i ../Example/test.fa

Enrichment Score

We use the following algorithm to calculate the enrichment score for the motif found by CEMIG.

Step 1: Tecord the number of site occurrences of a motif in all sequences, for example 2000 times assuming a total of 10000 sequences.

Step 2: Use the PWM matrix of the modules found by CMEIG to score each site in the motif. Every site in motif that appears in any input sequence is recorded, and this step is repeated for all sites and the lowest score is recorded as the threshold.

Step 3: Randomly generates a background sequence of 100 based on the frequency of specific base pairs appearing in the input sequence; Using similar steps as in the first step, scan these background sequences based on motif's PWM. If the score of the fragments reaches or exceeds the threshold, it is considered that the background sequence contains the motif site.

Step 4: Based on the steps below, the number of background sequences containing motif sites is obtained and based on both the number of motif sites in the original and background sequences. Enrichment score is defined as the number of motif sites in the original and background sequences and P-value was calculated using Fisher's exact test. The following is a table for Fisher's exact test:

Contain Not contain Row Total
Input Sequence a b a + b
Background Sequence c d c + d
Column Total a + c b + d a + b + c + d (=n)

And P-value was calculated using Fisher's exact test accompanied by the enrichment score:

$$p=\frac{\left(\begin{array}{c}a+b \ a\end{array}\right)\left(\begin{array}{c}c+d \ c\end{array}\right)}{\left(\begin{array}{c}n \ a+c\end{array}\right)}=\frac{(a+b) !(c+d) !(a+c) !(b+d) !}{a ! b ! c ! d ! n !}$$

Both the enrichment scores and P-values will be output together with motif in the output file with MEME format.

Parameters

Option Parameter Description Default
-I inputfile Specify input file. The program uses input file with standard FASTA file format in default.
-O outfile Specify the output file prefix name and location. The program uses the input file path and prefix name as default values.
-P paired-end Specify Whether the input data is paired-end. This flag is set as TRUE in default.
-M maxmotifs Maximum number of output motifs. The default number is 100.
-W Width Specify k value which determines the length of k-mer. The k-mer uses 6-mer in default. This parameter is not recommended to be modified.
-G gap The maximum number of gaps allowed for the cluster to extend on the path. The default number is 6.

Citation

Any questions, problems, bugs are welcome and should be dumped to Cankun Wang [email protected].

About

A new motif prediction algorithm called CEMIG (cis regulatory motif influence using de Bruijn graph)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published