Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Motif search with Gibbs sampler

I am a beginner in both programming and bioinformatics. So, I would appreciate your understanding. I tried to develop a python script for motif search using Gibbs sampling as explained in Coursera class, "Finding Hidden Messages in DNA". The pseudocode provided in the course is:

GIBBSSAMPLER(Dna, k, t, N)
    randomly select k-mers Motifs = (Motif1, …, Motift) in each string
        from Dna
    BestMotifs ← Motifs
    for j ← 1 to N
        i ← Random(t)
        Profile ← profile matrix constructed from all strings in Motifs
                   except for Motifi
        Motifi ← Profile-randomly generated k-mer in the i-th sequence
        if Score(Motifs) < Score(BestMotifs)
            BestMotifs ← Motifs
    return BestMotifs

Problem description:

CODE CHALLENGE: Implement GIBBSSAMPLER.

Input: Integers k, t, and N, followed by a collection of strings Dna. Output: The strings BestMotifs resulting from running GIBBSSAMPLER(Dna, k, t, N) with 20 random starts. Remember to use pseudocounts!

Sample Input:

 8 5 100
 CGCCCCTCTCGGGGGTGTTCAGTAACCGGCCA
 GGGCGAGGTATGTGTAAGTGCCAAGGTGCCAG
 TAGTACCGAGACCGAAAGAAGTATACAGGCGT
 TAGATCAAGTTTCAGGTGCACGTCGGTGAACC
 AATCCACCAGCTCCACGTGCAATGTTGGCCTA

Sample Output:

 TCTCGGGG
 CCAAGGTG
 TACAGGCG
 TTCAGGTG
 TCCACGTG

I followed the pseudocode to the best of my knowledge. Here is my code:

def BuildProfileMatrix(dnamatrix):
    ProfileMatrix = [[1 for x in xrange(len(dnamatrix[0]))] for x in xrange(4)]
    indices = {'A':0, 'C':1, 'G': 2, 'T':3}
    for seq in dnamatrix:
    for i in xrange(len(dnamatrix[0])):            
        ProfileMatrix[indices[seq[i]]][i] += 1
    ProbMatrix = [[float(x)/sum(zip(*ProfileMatrix)[0]) for x in y] for y in ProfileMatrix]
    return ProbMatrix
def ProfileRandomGenerator(profile, dna, k, i):
    indices = {'A':0, 'C':1, 'G': 2, 'T':3}
    score_list = []
    for x in xrange(len(dna[i]) - k + 1):
        probability = 1
        window = dna[i][x : k + x]
    for y in xrange(k):
        probability *= profile[indices[window[y]]][y]
    score_list.append(probability)
    rnd = uniform(0, sum(score_list))
    current = 0
    for z, bias in enumerate(score_list):
        current += bias
        if rnd <= current:
            return dna[i][z : k + z]
def score(motifs):
    ProfileMatrix = [[0 for x in xrange(len(motifs[0]))] for x in xrange(4)]
    indices = {'A':0, 'C':1, 'G': 2, 'T':3}
    for seq in motifs:
        for i in xrange(len(motifs[0])):            
            ProfileMatrix[indices[seq[i]]][i] += 1
    score = len(motifs)*len(motifs[0]) - sum([max(x) for x in zip(*ProfileMatrix)])
    return score
from random import randint, uniform    
def GibbsSampler(k, t, N):
     dna = ['CGCCCCTCTCGGGGGTGTTCAGTAACCGGCCA',
    'GGGCGAGGTATGTGTAAGTGCCAAGGTGCCAG',
    'TAGTACCGAGACCGAAAGAAGTATACAGGCGT',
    'TAGATCAAGTTTCAGGTGCACGTCGGTGAACC',
    'AATCCACCAGCTCCACGTGCAATGTTGGCCTA']
    Motifs = []
    for i in [randint(0, len(dna[0])-k) for x in range(len(dna))]:
        j = 0
        kmer = dna[j][i : k+i]
        j += 1
        Motifs.append(kmer)
    BestMotifs = []
    s_best = float('inf')
    for i in xrange(N):
        x = randint(0, t-1)
    Motifs.pop(x)
    profile = BuildProfileMatrix(Motifs)
    Motif = ProfileRandomGenerator(profile, dna, k, x)
    Motifs.append(Motif)
    s_motifs = score(Motifs)
    if s_motifs < s_best:
        s_best = s_motifs
        BestMotifs = Motifs
return [s_best, BestMotifs]

k, t, N =8, 5, 100            
best_motifs = [float('inf'), None]

# Repeat the Gibbs sampler search 20 times.
for repeat in xrange(20):
    current_motifs = GibbsSampler(k, t, N)
    if current_motifs[0] < best_motifs[0]:
        best_motifs = current_motifs
# Print and save the answer.
print '\n'.join(best_motifs[1])            

Unfortunately, my code never gives the same output as the solved example. Besides, while trying to debug the code I found that I get weird scores that define the mismatches between motifs. However, when I tried to run the score function separately, it worked perfectly.

Each time I run the script, the output changes, but anyway here is an example of one of the outputs for the input present in the code:

Example output of my code

TATGTGTA
TATGTGTA
TATGTGTA
GGTGTTCA
TATACAGG

Could you please help me debug this code?!! I spent the whole day trying to find out what's wrong with it although I know it might be some silly mistake I made, but my eye failed to catch it.

Thank you all!!

like image 846
Ali Elbehery Avatar asked Feb 27 '16 23:02

Ali Elbehery


People also ask

What is Gibbs sampler used for?

Gibbs sampling is commonly used for statistical inference (e.g. determining the best value of a parameter, such as determining the number of people likely to shop at a particular store on a given day, the candidate a voter will most likely vote for, etc.).

Which tool is used to identify motifs?

MEME is a de novo motif finding tool, which was designed for finding un-gapped motifs in unaligned DNA or protein sequences.


1 Answers

Finally, I found out what was wrong in my code! It was in line 54:

Motifs.append(Motif)

After randomly removing one of the motifs, followed by building a profile out of these motifs then randomly selecting a new motif based on this profile, I should have added the selected motif in the same position before removal NOT appended to the end of the motif list.

Now, the correct code is:

Motifs.insert(x, Motif)

The new code worked as expected.

like image 175
Ali Elbehery Avatar answered Nov 11 '22 10:11

Ali Elbehery