def __init__(self):
""" Load sequences and vaccine efficacy scores according to literature. """
print("class being initialized")
_, _ = trim("data/NA_2012_aligned_seq.fa"
) # test that sequence_processing still works
print("done trimming NA_2012 sequences")
self.circ_labels, self.circ_seqs = trim(
"data/aligned_circulating_strains.fa")
self.vax_labels, self.vax_seqs = trim(
"data/aligned_vaccine_strains.fa")
self.year = np.array([
1971, 1972, 1973, 1975, 1984, 1985, 1987, 1989, 1992, 1993, 1994,
1995, 1996, 1997, 1998, 1999, 2001, 2003
])
self.eff = np.array([
7, 15, 11, -3, -6, -2, 17, -5, 59, 38, 25, 45, 28, -17, 34, 43, 55,
12
]) # https://doi.org/10.1016/j.vaccine.2006.01.010
assert (len(self.vax_seqs) == len(self.circ_seqs) == len(self.year) ==
len(self.eff))
self.size = len(self.vax_seqs)
def test_mat(self, seqFile):
""" function is the same as "dist_mat()" except that it only looks at first 10 sequences
in order to get a proof of concept for all my functions before scaling up to the full dataset
"""
labels, sequences = trim(seqFile)
testMat = np.zeros((1000,1000))
print('calculating the test distance matrix based on PAM250')
for i in range(0,1000):
for j in range(i,1000):
testMat[i,j] = self.seq_dist(self.sequences[i], self.sequences[j])
# plug in values for mirror images
testMat[j,i] = testMat[i,j]
return testMat
def dist_mat(self, seqFile, hybrid=False, hamming=False):
""" Calculates all pairwise sequence distances for all sequences in a given file. """
labels, sequences = trim(seqFile)
numSeq = sequences.shape[0]
distMat = np.zeros((numSeq, numSeq))
print('calculating the full distance matrix')
for i in range(0,numSeq):
for j in range(i,numSeq):
if hamming: # use hamming distance if option is true
distMat[i,j] = Hamming_dist(sequences[i], sequences[j])
else: # use substitution matrix otherwise (allow for hybrid option)
distMat[i,j] = self.seq_dist(sequences[i], sequences[j], hybrid=hybrid)
distMat[j,i] = distMat[i,j] # plug in mirror image values
return distMat
def generate_MDS(dist_mat, method, out_file):
""" Runs MDS and saves figures given a specific dist_mat.
method represents the method in which the dist_mat was generated.
out_file is the name of the figure that gets saved."""
labels_2012, sequences_2012 = trim("data/NA_2012_aligned_seq.fa")
numSeq_2012 = sequences_2012.shape[0]
mds = manifold.MDS(n_components=2,
max_iter=8000,
dissimilarity="precomputed",
n_jobs=1)
results = mds.fit(dist_mat)
pos = results.embedding_
stress = results.stress_
print('stress: ' + str(stress))
cmap = mpl.cm.autumn # more options found here: https://matplotlib.org/tutorials/colors/colormaps.html
autumn_map = plt.get_cmap('autumn')
color_array = np.zeros((numSeq_2012, 4))
for ii in range(0, numSeq_2012):
color_array[ii, :] = cmap(ii / float(numSeq_2012))
plt.figure(
figsize=(8,
5)) # set up figure for plotting, width and height in inches
pts = plt.scatter(pos[:, 0],
pos[:, 1],
color=color_array,
cmap=autumn_map,
s=10,
alpha=0.5)
plt.scatter(pos[1, 0],
pos[1, 1],
color='b',
marker='*',
s=50,
alpha=0.9,
label=('vaccine target: ' + str(labels_2012[1])))
plt.legend()
plt.xlabel('Component 1')
plt.ylabel('Component 2')
plt.grid(alpha=0.3)
plt.title('HA1 sequences of H3N2 virus: North America - 2012 - ' +
str(method))
plt.savefig(out_file)
def generate_tSNE(dist_mat, method, out_file):
""" Runs tSNE and saves figures given a specific dist_mat.
method represents the method in which the dist_mat was generated.
out_file is the name of the figure that gets saved. """
labels_2012, sequences_2012 = trim("data/NA_2012_aligned_seq.fa")
numSeq_2012 = sequences_2012.shape[0]
tsne = manifold.TSNE(n_components=2, metric='precomputed')
tsne_results = tsne.fit(dist_mat)
tsne_pos = tsne_results.embedding_
tsne_divergence = tsne_results.kl_divergence_
tsne_iter = tsne_results.n_iter_
print('Kullback-Leibler divergence after optimization: ' +
str(tsne_divergence))
print('Number of iterations run: ' + str(tsne_iter))
cmap = mpl.cm.autumn # more options found here: https://matplotlib.org/tutorials/colors/colormaps.html
autumn_map = plt.get_cmap('autumn')
color_array = np.zeros((numSeq_2012, 4))
for ii in range(0, numSeq_2012):
color_array[ii, :] = cmap(ii / float(numSeq_2012))
plt.figure(
figsize=(8,
5)) # set up figure for plotting, width and height in inches
pts = plt.scatter(tsne_pos[:, 0],
tsne_pos[:, 1],
color=color_array,
cmap=autumn_map,
s=10,
alpha=0.5)
plt.scatter(tsne_pos[1, 0],
tsne_pos[1, 1],
color='b',
marker='*',
s=50,
alpha=0.9,
label=('vaccine target: ' + str(labels_2012[1])))
plt.legend()
plt.xlabel('Component 1')
plt.ylabel('Component 2')
plt.grid(alpha=0.3)
plt.title('HA1 H3N2 North America 2012 - tSNE - ' + str(method) +
' - KL divergence: ' + str(round(tsne_divergence, 3)))
plt.savefig(out_file)