def main():
print("<div style=\"border:1px solid black;\">", end="\n\n")
print("`{bm-disable-all}`", end="\n\n")
try:
dnas = []
while True:
try:
dna = input().strip().upper()
if len(dna) > 0:
dnas.append(dna)
except EOFError:
break
hasher = md5()
hasher.update(str(dnas).encode('utf-8'))
logo_filename = 'motif_logo_' + hasher.hexdigest() + '.svg'
if path.isdir('/output'):
logo_path = '/output/' + logo_filename
else:
logo_path = '/tmp/' + logo_filename
print(f'Generating logo for the following motif matrix...\n\n')
print(f'{"<br>".join(dnas)}\n\n')
counts = motif_matrix_count(dnas)
profile = motif_matrix_profile(counts)
logo = create_logo(profile)
plt.savefig(logo_path)
print(f'Result...\n\n')
print(f'\n\n')
finally:
print("</div>", end="\n\n")
print("`{bm-enable-all}`", end="\n\n")
def gibbs_sampler_motif_search_with_psuedocounts(k: int, dnas: List[str],
cycles: int) -> List[str]:
motif_matrix = []
for dna in dnas:
start = randrange(len(dna) - k + 1)
kmer = dna[start:start + k]
motif_matrix.append(kmer)
best_motif_matrix = motif_matrix[:] # create a copy, otherwise you'll be modifying both motif and best_motif
for j in range(0, cycles):
i = randrange(len(dnas)) # pick a dna
del motif_matrix[i] # remove the kmer for that dna from the motif str
counts = motif_matrix_count(motif_matrix)
apply_psuedocounts_to_count_matrix(counts)
profile = motif_matrix_profile(counts)
new_motif_kmer_probs = determine_probabilities_of_all_kmers_in_dna(
profile, dnas[i], k)
new_motif_kmer_idx = gibbs_rand(new_motif_kmer_probs)
new_motif_kmer = dnas[i][new_motif_kmer_idx:new_motif_kmer_idx + k]
motif_matrix.insert(i, new_motif_kmer)
if score_motif(motif_matrix) < score_motif(best_motif_matrix):
best_motif_matrix = motif_matrix[:] # create a copy, otherwise you'll be modifying both motif and best_motif
return best_motif_matrix
def main():
print("<div style=\"border:1px solid black;\">", end="\n\n")
print("`{bm-disable-all}`", end="\n\n")
try:
dnas = []
while True:
try:
dna = input().strip().upper()
if len(dna) > 0:
dnas.append(dna)
except EOFError:
break
kmer = dnas[-1]
motif_matrix = dnas[:-2]
counts = motif_matrix_count(motif_matrix)
profile = motif_matrix_profile(counts)
prob = determine_probability_of_match_using_profile_matrix(profile, kmer)
print(f'Motif matrix...\n\n')
print(f'{"<br>".join(motif_matrix)}\n\n')
print(f'Probability that {kmer} matches the motif {prob}...\n\n')
finally:
print("</div>", end="\n\n")
print("`{bm-enable-all}`", end="\n\n")
def greedy_motif_search(k: int, dnas: List[str]):
best_motif_matrix = [dna[0:k] for dna in dnas]
for motif, _ in slide_window(dnas[0], k):
motif_matrix = [motif]
counts = motif_matrix_count(motif_matrix)
profile = motif_matrix_profile(counts)
for dna in dnas[1:]:
next_motif, _ = find_most_probable_kmer_using_profile_matrix(
profile, dna)
# push in closest kmer as a motif member and recompute profile for the next iteration
motif_matrix.append(next_motif)
counts = motif_matrix_count(motif_matrix)
profile = motif_matrix_profile(counts)
if score_motif(motif_matrix) < score_motif(best_motif_matrix):
best_motif_matrix = motif_matrix
return best_motif_matrix
def score_motify_entropy(motif_matrix: List[str]) -> float:
rows = len(motif_matrix)
cols = len(motif_matrix[0])
# count up each column
counts = motif_matrix_count(motif_matrix)
profile = motif_matrix_profile(counts)
# prob dist to entropy
entropy_per_col = []
for c in range(cols):
entropy = calculate_entropy([profile['A'][c], profile['C'][c], profile['G'][c], profile['T'][c]])
entropy_per_col.append(entropy)
# sum up entropies to get entropy of motif
return sum(entropy_per_col)
def score_motif_relative_entropy(motif_matrix: List[str],
source_strs: List[str]) -> float:
# calculate frequency of nucleotide across all source strings
nuc_counter = Counter()
nuc_total = 0
for source_str in source_strs:
for nuc in source_str:
nuc_counter[nuc] += 1
nuc_total += len(source_str)
nuc_freqs = dict([(k, v / nuc_total) for k, v in nuc_counter.items()])
rows = len(motif_matrix)
cols = len(motif_matrix[0])
# count up each column
counts = motif_matrix_count(motif_matrix)
profile = motif_matrix_profile(counts)
relative_entropy_per_col = []
for c in range(cols):
# get entropy of column in motif
entropy = calculate_entropy([
profile['A'][c], profile['C'][c], profile['G'][c], profile['T'][c]
])
# get cross entropy of column in motif (mixes in global nucleotide frequencies)
cross_entropy = calculate_cross_entropy([
profile['A'][c], profile['C'][c], profile['G'][c], profile['T'][c]
], [nuc_freqs['A'], nuc_freqs['C'], nuc_freqs['G'], nuc_freqs['T']])
relative_entropy = entropy - cross_entropy
# Right now relative_entropy is calculated by subtracting cross_entropy from (a negated) entropy. But, according
# to the Pevzner book, the calculation of relative_entropy can be simplified to just...
# def calculate_relative_entropy(probabilities_for_nuc: List[float], total_frequencies_for_nucs: List[float]) -> float:
# ret = 0.0
# for prob, total_freq in zip(probabilities_for_nuc, total_frequency_for_nucs):
# ret += value * (log(value / total_freq, 2.0) if value > 0.0 else 0.0)
# return ret
relative_entropy_per_col.append(relative_entropy)
# sum up entropies to get entropy of motif
ret = sum(relative_entropy_per_col)
# All of the other score_motif algorithms try to MINIMIZE score. In the case of relative entropy (this algorithm),
# the greater the score is the better of a match it is. As such, negate this score so the existing algorithms can
# still try to minimize.
return -ret
def randomized_motif_search_with_psuedocounts(k: int,
dnas: List[str]) -> List[str]:
motif_matrix = []
for dna in dnas:
start = randrange(len(dna) - k + 1)
kmer = dna[start:start + k]
motif_matrix.append(kmer)
best_motif_matrix = motif_matrix
while True:
counts = motif_matrix_count(motif_matrix)
apply_psuedocounts_to_count_matrix(counts)
profile = motif_matrix_profile(counts)
motif_matrix = [
find_most_probable_kmer_using_profile_matrix(profile, dna)[0]
for dna in dnas
]
if score_motif(motif_matrix) < score_motif(best_motif_matrix):
best_motif_matrix = motif_matrix
else:
return best_motif_matrix
dnas = [
'GGCGTTCAGGCA', 'AAGAATCAGTCA', 'CAAGGAGTTCGC', 'CACGTCAATCAC',
'CAATAATATTCG'
]
k = 3
motifs = []
for dna in dnas:
start = randrange(len(dna) - k)
motif = dna[start:start + k]
motifs.append(motif)
best_motifs = motifs
while True:
counts_matrix = motif_matrix_count(motifs)
for elem, counts in counts_matrix.items(): # add in pseudocounts
counts_matrix[elem] = [c + 1 for c in counts]
profile_matrix = motif_matrix_profile(counts_matrix)
motifs = [
find_most_probable_kmer_using_profile_matrix(profile_matrix, dna)[0]
for dna in dnas
]
if score_motif(motifs) < score_motif(best_motifs):
best_motifs = motifs
else:
break
[print(f'{m}') for m in best_motifs]
# CGGGTCAAACGACCCTAGTG
# CGGGACGTAAGTCCCTAACG
# CCGGGCTTCCAACCGTGGCC
# CGTGACCGACGTCCCCAGCC
# GAGGACCTTCGGCCCCACCC
# GGGGACTTCTGTCCCTAGCC
# TGGGACTTTCGGCCCTGTCC
# GGGGACCAACGCCCCTGGGA
# GGGGACCGAAGTCCCCGGGC
# 11
# consensus_kmer = 'CGGGACCTACGTCCCTAGCC' # this is consensus string for hte matrix it finds
best_motif_matrix_counts = motif_matrix_count(best_motif_matrix)
for elem, counts in best_motif_matrix_counts.items(): # add in pseudocounts
best_motif_matrix_counts[elem] = [c + 1 for c in counts]
best_motif_matrix_profile = motif_matrix_profile(best_motif_matrix_counts)
with open('/home/user/Downloads/GCF_000195955.2_ASM19595v2_genomic.fna', mode='r', encoding='utf-8') as f:
data = f.read()
lines = data.split('\n')
lines = [l.strip() for l in lines] # get rid of whitespace
lines = [l if not l.startswith('>') else '' for l in lines] # remove comments
dna = ''.join(lines) # concat into single dna str
for kmer, _ in slide_window(dna, k):
prob = determine_probability_of_match_using_profile_matrix(best_motif_matrix_profile, kmer)
if prob >= 0.01: # 1% or greater
print(f'{kmer} {prob}')
# Nothing is found...
#
# The strings in DosR.txt aren't matching up to the genome at the link I posted (even though the name of the organism