def get_procrustes_results(coords_f1,coords_f2,sample_id_map=None,\
randomize=None,max_dimensions=None,\
get_eigenvalues=get_mean_eigenvalues,\
get_percent_variation_explained=get_mean_percent_variation):
""" """
# Parse the PCoA files
sample_ids1, coords1, eigvals1, pct_var1 = parse_coords(coords_f1)
sample_ids2, coords2, eigvals2, pct_var2 = parse_coords(coords_f2)
if sample_id_map:
sample_ids1 = map_sample_ids(sample_ids1,sample_id_map)
sample_ids2 = map_sample_ids(sample_ids2,sample_id_map)
# rearrange the order of coords in coords2 to correspond to
# the order of coords in coords1
order = list(set(sample_ids1) & set(sample_ids2))
coords1 = reorder_coords(coords1,sample_ids1,order)
coords2 = reorder_coords(coords2,sample_ids2,order)
# If this is a random trial, apply the shuffling function passed as
# randomize()
if randomize:
coords2 = randomize(coords2)
coords1, coords2 = pad_coords_matrices(coords1,coords2)
if max_dimensions:
coords1 = filter_coords_matrix(coords1,max_dimensions)
coords2 = filter_coords_matrix(coords2,max_dimensions)
pct_var1 = pct_var1[:max_dimensions]
pct_var2 = pct_var2[:max_dimensions]
eigvals1 = eigvals1[:max_dimensions]
eigvals2 = eigvals2[:max_dimensions]
else:
if len(pct_var1)>len(pct_var2):
pct_var2 = append(pct_var2,zeros(len(pct_var1)-len(pct_var2)))
eigvals2 = append(eigvals2,zeros(len(eigvals1)-len(eigvals2)))
elif len(pct_var1)<len(pct_var2):
pct_var1 = append(pct_var1,zeros(len(pct_var2)-len(pct_var1)))
eigvals1 = append(eigvals1,zeros(len(eigvals2)-len(eigvals1)))
# Run the Procrustes analysis
transformed_coords_m1, transformed_coords_m2, m_squared =\
procrustes(coords1,coords2)
# print coords2
#print transformed_coords_m2
eigvals = get_eigenvalues(eigvals1, eigvals2)
pct_var = get_percent_variation_explained(pct_var1,pct_var2)
transformed_coords1 = format_coords(coord_header=order,\
coords=transformed_coords_m1,\
eigvals=eigvals,\
pct_var=pct_var)
transformed_coords2 = format_coords(coord_header=order,\
coords=transformed_coords_m2,\
eigvals=eigvals,\
pct_var=pct_var)
# Return the results
return transformed_coords1, transformed_coords2, m_squared
def get_procrustes_results(coords_f1,coords_f2,sample_id_map=None,\
randomize=None,max_dimensions=None,\
get_eigenvalues=get_mean_eigenvalues,\
get_percent_variation_explained=get_mean_percent_variation):
""" """
# Parse the PCoA files
sample_ids1, coords1, eigvals1, pct_var1 = parse_coords(coords_f1)
sample_ids2, coords2, eigvals2, pct_var2 = parse_coords(coords_f2)
if sample_id_map:
sample_ids1 = map_sample_ids(sample_ids1,sample_id_map)
sample_ids2 = map_sample_ids(sample_ids2,sample_id_map)
# rearrange the order of coords in coords2 to correspond to
# the order of coords in coords1
order = list(set(sample_ids1) & set(sample_ids2))
coords1 = reorder_coords(coords1,sample_ids1,order)
coords2 = reorder_coords(coords2,sample_ids2,order)
# If this is a random trial, apply the shuffling function passed as
# randomize()
if randomize:
coords2 = randomize(coords2)
coords1, coords2 = pad_coords_matrices(coords1,coords2)
if max_dimensions:
coords1 = filter_coords_matrix(coords1,max_dimensions)
coords2 = filter_coords_matrix(coords2,max_dimensions)
pct_var1 = pct_var1[:max_dimensions]
pct_var2 = pct_var2[:max_dimensions]
eigvals1 = eigvals1[:max_dimensions]
eigvals2 = eigvals2[:max_dimensions]
# Run the Procrustes analysis
transformed_coords_m1, transformed_coords_m2, m_squared =\
procrustes(coords1,coords2)
#print transformed_coords_m2
eigvals = get_eigenvalues(eigvals1, eigvals2)
pct_var = get_percent_variation_explained(pct_var1,pct_var2)
transformed_coords1 = format_coords(coord_header=order,\
coords=transformed_coords_m1,\
eigvals=eigvals,\
pct_var=pct_var)
transformed_coords2 = format_coords(coord_header=order,\
coords=transformed_coords_m2,\
eigvals=eigvals,\
pct_var=pct_var)
# Return the results
return transformed_coords1, transformed_coords2, m_squared
def test_format_coords(self):
"""format_coords should return tab-delimited table of coords"""
a = array([[1,2,3],[4,5,6],[7,8,9]])
header = list('abc')
eigvals = [2,4,6]
pct_var = [3,2,1]
res = format_coords(header, a, eigvals, pct_var)
self.assertEqual(res, "pc vector number\t1\t2\t3\na\t1\t2\t3\nb\t4\t5\t6\nc\t7\t8\t9\n\n\neigvals\t2\t4\t6\n% variation explained\t3\t2\t1")
def test_format_coords(self):
"""format_coords should return tab-delimited table of coords"""
a = array([[1,2,3],[4,5,6],[7,8,9]])
header = list('abc')
eigvals = [2,4,6]
pct_var = [3,2,1]
res = format_coords(header, a, eigvals, pct_var)
self.assertEqual(res, "pc vector number\t1\t2\t3\na\t1\t2\t3\nb\t4\t5\t6\nc\t7\t8\t9\n\n\neigvals\t2\t4\t6\n% variation explained\t3\t2\t1")
def pcoa(file):
samples, distmtx = parse_distmat(file)
# coords, each row is an axis
coords, eigvals = ms.principal_coordinates_analysis(distmtx)
pcnts = (numpy.abs(eigvals) / sum(numpy.abs(eigvals))) * 100
idxs_descending = pcnts.argsort()[::-1]
coords = coords[idxs_descending]
eigvals = eigvals[idxs_descending]
pcnts = pcnts[idxs_descending]
return format_coords(samples, coords.T, eigvals, pcnts)
def pcoa(file):
samples, distmtx = parse_distmat(file)
# coords, each row is an axis
coords, eigvals = ms.principal_coordinates_analysis(distmtx)
pcnts = (numpy.abs(eigvals) / sum(numpy.abs(eigvals))) * 100
idxs_descending = pcnts.argsort()[::-1]
coords = coords[idxs_descending]
eigvals = eigvals[idxs_descending]
pcnts = pcnts[idxs_descending]
return format_coords(samples, coords.T, eigvals, pcnts)
def generate_3d_plots_invue(prefs, data, dir_path, filename, intp_pts, polyh_pts, offset):
""" Make files to be imported to inVUE
http://sourceforge.net/projects/invue/"""
# Validating existance of all columns
for col in prefs:
if col not in data['map'][0]:
raise ValueError, 'Column given "%s" does not exits in mapping \
file' % col
# Split matrix by labelname, groups & give colors
groups_and_colors=iter_color_groups(data['map'],prefs)
groups_and_colors=list(groups_and_colors)
smp_lbl, smp_lbl_grp, polypts = make_3d_plots_invue(data, groups_and_colors, \
intp_pts, polyh_pts, offset)
# Looping to binning result to write full and binned files
for lbl in smp_lbl:
for grp in smp_lbl_grp[lbl]:
# writting individual files
ind_path = "%s/%s_%s_%s.txt" % (dir_path, filename, lbl, grp)
smp = smp_lbl_grp[lbl][grp]
outfile = open(ind_path, 'w')
outfile.write(format_coords(smp['headrs'], smp['coords'], [], [], False))
outfile.close()
# writing full file
full_path = "%s/%s_%s.txt" % (dir_path, filename, lbl)
outfile = open(full_path, 'w')
outfile.write (format_coords(smp_lbl[lbl]['headrs'], smp_lbl[lbl]['coords'], \
[], [], False))
outfile.close()
# Writing tetraVertices.txt
ind_path = "%s/tetraVertices.txt" % (dir_path)
outfile = open(ind_path, 'w')
outfile.write('\n'.join(['\t'.join(map(str, row)) for row in polypts]))
outfile.write('\n')
outfile.close()
def compute_manifold(in_file,alg,params=None):
"""compute the specified manifold on the specified file"""
otu_table = parse_biom_table(in_file)
samples = otu_table.SampleIds
#Dense tables already have all values available
#For sparse tables we have to more or less generate missing points
if isinstance(otu_table, DenseTable):
otumtx = otu_table._data.T
else:
otumtx = asarray([v for v in otu_table.iterSampleData()])
#Setup the mapping algorithms from sklearns using specified parameters
#if a parameter in the dict is invalid for the chosen algorithm it is simply ignored
if alg=="isomap":
defaults = {"n_neighbors":5,"n_components":3,"eigen_solver":"auto",
"tol":0,"max_iter":None,"path_method":"auto","neighbors_algorithm":"auto"}
params = fill_args(defaults,params)
mapper = manifold.Isomap(
n_neighbors=params["n_neighbors"],
n_components=params["n_components"],
eigen_solver=params["eigen_solver"],
tol=params["tol"],
max_iter=params["max_iter"],
path_method=params["path_method"],
neighbors_algorithm=params["neighbors_algorithm"])
elif alg=="lle":
defaults = {"n_neighbors":5,"n_components":3,"reg":0.001,"eigen_solver":"auto",
"tol":1e-06,"max_iter":100,"method":"standard","hessian_tol":0.0001,
"modified_tol":1e-12,"neighbors_algorithm":"auto","random_state":None}
params = fill_args(defaults,params)
mapper = manifold.LocallyLinearEmbedding(
n_neighbors=params["n_neighbors"],
n_components=params["n_components"],
reg=params["reg"],
eigen_solver=params["eigen_solver"],
tol=params["tol"],
max_iter=params["max_iter"],
method=params["method"],
hessian_tol=params["hessian_tol"],
modified_tol=params["modified_tol"],
neighbors_algorithm=params["neighbors_algorithm"],
random_state=params["random_state"])
elif alg=="spectral":
defaults = {"n_components":3,"affinity":"nearest_neighbors","gamma":None,
"random_state":None,"eigen_solver":None,"n_neighbors":None}
params = fill_args(defaults,params)
mapper = manifold.SpectralEmbedding(
n_components=params["n_components"],
affinity=params["affinity"],
gamma=params["gamma"],
random_state=params["random_state"],
eigen_solver=params["eigen_solver"],
n_neighbors=params["n_neighbors"])
elif alg=="ltsa":
defaults = {"n_neighbors":5,"n_components":3,"reg":0.001,"eigen_solver":"auto",
"tol":1e-06,"max_iter":100,"method":"ltsa","hessian_tol":0.0001,
"modified_tol":1e-12,"neighbors_algorithm":"auto","random_state":None}
params = fill_args(defaults,params)
mapper = manifold.LocallyLinearEmbedding(
n_neighbors=params["n_neighbors"],
n_components=params["n_components"],
reg=params["reg"],
eigen_solver=params["eigen_solver"],
tol=params["tol"],
max_iter=params["max_iter"],
method=params["method"],
hessian_tol=params["hessian_tol"],
modified_tol=params["modified_tol"],
neighbors_algorithm=params["neighbors_algorithm"],
random_state=params["random_state"])
elif alg=="mds":
defaults = {"n_components":3,"metric":True,"n_init":4,"max_iter":300,
"verbose":0,"eps":0.001,"n_jobs":1,"random_state":None,
"dissimilarity":"euclidean"}
params = fill_args(defaults,params)
mapper = manifold.Isomap(
n_components=params["n_components"],
metric=params["metric"],
n_init=params["n_init"],
max_iter=params["max_iter"],
verbose=params["verbose"],
eps=params["eps"],
n_jobs=params["n_jobs"],
random_state=params["random_state"],
dissimilarity=params["dissimilarity"])
else:
print("arg in error, unknown algorithm '"+alg+"'")
exit(1)
#compute the fit and scale from -1 to 1
fit = mapper.fit_transform(otumtx)
fit /= abs(fit).max()
#dummy eigenvalues and percent variation explained
#"make_emperor.py" does not work if these are not supplied
eigvals = [3.0,2.0,1.0]
pcnts = [30.0,20.0,10.0]
return format_coords(samples, fit, eigvals, pcnts)
def get_procrustes_results(
coords_f1,
coords_f2,
sample_id_map=None,
randomize=None,
max_dimensions=None,
get_eigenvalues=get_mean_eigenvalues,
get_percent_variation_explained=get_mean_percent_variation):
""" """
# Parse the PCoA files
sample_ids1, coords1, eigvals1, pct_var1 = parse_coords(coords_f1)
sample_ids2, coords2, eigvals2, pct_var2 = parse_coords(coords_f2)
if sample_id_map:
sample_ids1 = map_sample_ids(sample_ids1, sample_id_map)
sample_ids2 = map_sample_ids(sample_ids2, sample_id_map)
# rearrange the order of coords in coords2 to correspond to
# the order of coords in coords1
order = list(set(sample_ids1) & set(sample_ids2))
coords1 = reorder_coords(coords1, sample_ids1, order)
coords2 = reorder_coords(coords2, sample_ids2, order)
if len(order) == 0:
raise ValueError('No overlapping samples in the two files')
# If this is a random trial, apply the shuffling function passed as
# randomize()
if randomize:
coords2 = randomize(coords2)
randomized_coords2 = format_coords(coord_header=order,
coords=coords2,
eigvals=eigvals2,
pct_var=pct_var2)
else:
randomized_coords2 = None
coords1, coords2 = pad_coords_matrices(coords1, coords2)
if max_dimensions:
coords1 = filter_coords_matrix(coords1, max_dimensions)
coords2 = filter_coords_matrix(coords2, max_dimensions)
pct_var1 = pct_var1[:max_dimensions]
pct_var2 = pct_var2[:max_dimensions]
eigvals1 = eigvals1[:max_dimensions]
eigvals2 = eigvals2[:max_dimensions]
else:
if len(pct_var1) > len(pct_var2):
pct_var2 = append(pct_var2, zeros(len(pct_var1) - len(pct_var2)))
eigvals2 = append(eigvals2, zeros(len(eigvals1) - len(eigvals2)))
elif len(pct_var1) < len(pct_var2):
pct_var1 = append(pct_var1, zeros(len(pct_var2) - len(pct_var1)))
eigvals1 = append(eigvals1, zeros(len(eigvals2) - len(eigvals1)))
# Run the Procrustes analysis
transformed_coords_m1, transformed_coords_m2, m_squared =\
procrustes(coords1, coords2)
# print coords2
# print transformed_coords_m2
eigvals = get_eigenvalues(eigvals1, eigvals2)
pct_var = get_percent_variation_explained(pct_var1, pct_var2)
transformed_coords1 = format_coords(coord_header=order,
coords=transformed_coords_m1,
eigvals=eigvals,
pct_var=pct_var)
transformed_coords2 = format_coords(coord_header=order,
coords=transformed_coords_m2,
eigvals=eigvals,
pct_var=pct_var)
# Return the results
return (transformed_coords1, transformed_coords2, m_squared,
randomized_coords2)