def score_contig(contig, par, genome, genome_index, test):
score_obj = ScoreCollection()
score_gen = Score(mn.log_probability(contig.signature, par), genome, genome, contig.contig_id)
score_obj.genome = score_gen
group_genomes = test.group.all_genomes_but_index(genome_index)
for gen in group_genomes:
score_obj.group.append(Score(mn.log_probability(contig.signature, gen.par()),genome, gen, contig.contig_id))
for group in test.rest_groups:
outside_group = []
for gen in group.genomes:
score = Score(mn.log_probability(contig.signature,gen.par()), genome, gen, contig.contig_id)
outside_group.append(score)
score_obj.other.append(outside_group)
return score_obj
def test_log_probability(self):
dna_c = dna.DNA(id="hej",seq="AAAA")
dna_c.signature = Counter({0:10,1: 4,2:1,3:8,4:20})
log_p_val = ml.log_probability(dna_c,np.array([0.2,0.1,0.05,0.2,0.45]))
real_p = np.exp(log_p_val)
assert_almost_equal(real_p, 0.001074701)
dna_c = dna.DNA(id="hej",seq="AAAA")
dna_c.signature = Counter({0:1, 1:0})
log_p_val = ml.log_probability(dna_c,np.array([0.5,0.5]))
real_p = np.exp(log_p_val)
assert_almost_equal(real_p, 0.5)
dna_c = dna.DNA(id="hej",seq="AAAA")
dna_c.signature = Counter({0:0, 1:1})
log_p_val = ml.log_probability(dna_c,np.array([0.5,0.5]))
real_p = np.exp(log_p_val)
assert_almost_equal(real_p, 0.5)
def main(contigs_file, taxonomy_file, dir_path, kmer_length, dir_structure, taxonomy_info_in_contigs):
groups = []
DNA.generate_kmer_hash(kmer_length)
contigs = read_contigs_file(contigs_file, taxonomy_info=taxonomy_info_in_contigs)
# Divide genomes into groups, one for each genus
meta_genomes = genome_info_from_parsed_taxonomy_file(taxonomy_file)
# Fetch sequence for each genome
genomes = read_FASTA_files_no_groups(meta_genomes, dir_path, dir_structure=dir_structure)
for genome in genomes:
genome.calculate_signature()
genome.pseudo_par = mn.fit_nonzero_parameters([genome])
scores = []
for contig in contigs:
contig.calculate_signature()
for genome in genomes:
if contig.id == genome.id:
temp_genome = deepcopy(genome)
temp_genome.signature.subtract(contig.signature)
temp_pseudo_par = mn.fit_nonzero_parameters([temp_genome])
p_val = mn.log_probability(contig, temp_pseudo_par)
else:
p_val = mn.log_probability(contig, genome.pseudo_par)
scores.append(Score(p_val, contig, genome, contig.contig_id, taxonomy_info=taxonomy_info_in_contigs))
if taxonomy_info_in_contigs:
sys.stdout.write(
"p_value\tcontig_family\tcontig_genus\tcontig_species\tcontig_genome\tcompare_family\tcompare_genus\tcompare_species\tcompare_genome\tcontig_id"
+ os.linesep
)
else:
sys.stdout.write(
"p_value\t\tcontig_genome\tcompare_family\tcompare_genus\tcompare_species\tcompare_genome\tcontig_id"
+ os.linesep
)
for score in scores:
sys.stdout.write(str(score) + "\n")
def test_uniform_one_contig_prob(self):
f = fileinput.input(os.path.join(data_path,"bambus2.scaffold.linear.fasta.one_contig"))
c = list(SeqIO.parse(f,"fasta"))
f.close()
dna_c = dna.DNA(id = c[0].id, seq = str(c[0].seq))
dna_c.calculate_signature()
k = 4**4
uniform_prob = np.ones((dna.DNA.kmer_hash_count))
for i,cnt in dna_c.signature.items():
uniform_prob[i] = 1./k
log_prob = ml.log_probability(dna_c,uniform_prob)
print log_prob
assert_almost_equal(log_prob, -3791.05738056)
def test_log_probability(self):
f = fileinput.input(os.path.join(data_path,"bambus2.scaffold.linear.fasta.one_contig"))
c = list(SeqIO.parse(f,"fasta"))
f.close()
dna_c = dna.DNA(id = c[0].id, seq = str(c[0].seq))
dna_c.calculate_signature()
k = 4**4
uniform_prob = np.ones((dna.DNA.kmer_hash_count))
for i,cnt in dna_c.signature.items():
uniform_prob[i] = 1./k
log_prob = ml.log_probability(dna_c,uniform_prob)
mu = np.log(np.array([0.5,3.0,5.0]))
sigma = 0.5
x = np.log(np.array([0.5,3.0,5.0]))
cov_matrix = np.array([x])
p_ig = ig.log_pdf(x,mu,sigma)
p_test = model.log_probability(dna_c,cov_matrix,uniform_prob,mu,sigma)
assert_equal(p_ig+log_prob,p_test)
def log_probability(seq,cov_matrix,prob_vector,mu,sigma,factor=1):
ig_p = isotropic_gaussian.log_pdf(cov_matrix,mu,sigma)
mu_p = multinomial.log_probability(seq,prob_vector)
return ig_p*factor+mu_p
