def test_2(self):
"""l19 data should give stress below .13"""
ptmtx = array(
[
[7, 1, 0, 0, 0, 0, 0, 0, 0],
[4, 2, 0, 0, 0, 1, 0, 0, 0],
[2, 4, 0, 0, 0, 1, 0, 0, 0],
[1, 7, 0, 0, 0, 0, 0, 0, 0],
[0, 8, 0, 0, 0, 0, 0, 0, 0],
[0, 7, 1, 0, 0, 0, 0, 0, 0], #idx 5
[0, 4, 2, 0, 0, 0, 2, 0, 0],
[0, 2, 4, 0, 0, 0, 1, 0, 0],
[0, 1, 7, 0, 0, 0, 0, 0, 0],
[0, 0, 8, 0, 0, 0, 0, 0, 0],
[0, 0, 7, 1, 0, 0, 0, 0, 0], #idx 10
[0, 0, 4, 2, 0, 0, 0, 3, 0],
[0, 0, 2, 4, 0, 0, 0, 1, 0],
[0, 0, 1, 7, 0, 0, 0, 0, 0],
[0, 0, 0, 8, 0, 0, 0, 0, 0],
[0, 0, 0, 7, 1, 0, 0, 0, 0], #idx 15
[0, 0, 0, 4, 2, 0, 0, 0, 4],
[0, 0, 0, 2, 4, 0, 0, 0, 1],
[0, 0, 0, 1, 7, 0, 0, 0, 0]
],
'float')
distmtx = dist_euclidean(ptmtx)
nm = NMDS(distmtx, verbosity=0)
self.assertLessThan(nm.getStress(), .13)
def test_3(self):
"""l19 data should give stress below .13 in multi-D"""
ptmtx = array(
[
[7, 1, 0, 0, 0, 0, 0, 0, 0],
[4, 2, 0, 0, 0, 1, 0, 0, 0],
[2, 4, 0, 0, 0, 1, 0, 0, 0],
[1, 7, 0, 0, 0, 0, 0, 0, 0],
[0, 8, 0, 0, 0, 0, 0, 0, 0],
[0, 7, 1, 0, 0, 0, 0, 0, 0], #idx 5
[0, 4, 2, 0, 0, 0, 2, 0, 0],
[0, 2, 4, 0, 0, 0, 1, 0, 0],
[0, 1, 7, 0, 0, 0, 0, 0, 0],
[0, 0, 8, 0, 0, 0, 0, 0, 0],
[0, 0, 7, 1, 0, 0, 0, 0, 0], #idx 10
[0, 0, 4, 2, 0, 0, 0, 3, 0],
[0, 0, 2, 4, 0, 0, 0, 1, 0],
[0, 0, 1, 7, 0, 0, 0, 0, 0],
[0, 0, 0, 8, 0, 0, 0, 0, 0],
[0, 0, 0, 7, 1, 0, 0, 0, 0], #idx 15
[0, 0, 0, 4, 2, 0, 0, 0, 4],
[0, 0, 0, 2, 4, 0, 0, 0, 1],
[0, 0, 0, 1, 7, 0, 0, 0, 0]
],
'float')
distmtx = dist_euclidean(ptmtx)
for dim in range(3, 18):
nm = NMDS(distmtx, verbosity=0, dimension=dim)
self.assertLessThan(nm.getStress(), .13)
print "Removing species with less than two occurrences..."
sp_io = np.where(~(sp > 0), sp, 1)
column_sums = np.sum(sp_io, 0)
to_remove = np.where(column_sums < 2)
sp = np.delete(sp, to_remove, 1)
colnames = np.delete(colnames, to_remove)
print "Removing plots with less than two species..."
pl_io = np.where(~(sp > 0), sp, 1)
row_sums = np.sum(pl_io, 1)
to_remove = np.where(row_sums < 2)
sp = np.delete(sp, to_remove, 0)
rownames = np.delete(rownames, to_remove)
#print sp.shape, len(rownames)
#print sp.shape, len(colnames)
print "Normalizing species coverage data with McCune logarithm..."
sp = log_mccune(sp)
from cogent.cluster.nmds import NMDS, metaNMDS
from cogent.maths.distance_transform import dist_bray_curtis
print "Calculating distance matrix..."
distmtx = dist_bray_curtis(sp)
nmds = NMDS(distmtx, dimension = 3)
print nmds.getPoints()
print nmds.getStress()
#nmds = NMDS()
def reduce_similarity_matrix(similarity_matrix):
#distance_matrix = dist_euclidean(similarity_matrix)
distance_matrix = 1 - similarity_matrix
return NMDS(distance_matrix).getPoints()
def setUp(self):
"""creates inputs"""
self.mtx = array(
[[0, 3, 4, 8], [3, 0, 1, 27], [4, 1, 0, 3.5], [8, 27, 3.5, 0]],
'd')
self.nm = NMDS(self.mtx, verbosity=0)
def run(self, tempDistanceMetric=None, iDims=2, strDistanceMatrixFile=None, istrmTree=None, istrmEnvr=None):
"""
Runs analysis on loaded data.
:param tempDistanceMetric: The name of the distance metric to use when performing PCoA.
None indicates a distance matrix was already given when loading and will be used.
Supports "braycurtis","canberra","chebyshev","cityblock","correlation",
"cosine","euclidean","hamming","sqeuclidean",unifrac_unweighted","unifrac_weighted"
:type: String Distance matrix name
:param iDims: How many dimension to plot the PCoA graphs.
(This can be minimally 2; all combinations of dimensions are plotted).
iDims start with 1 (not index-based).
:type: Integer Positive integer 2 or greater.
:param strDistanceMatrixFile: If the underlying distance matrix should be output, this is the file to output to.
:type: String Output file for distances of None for indicating it shoudl not be done.
:param istrmTree: One of two files needed for unifrac calculations, this is the phylogeny of the features.
:type: String Path to file
:param istrmEnvr: One of two files needed for unifrac calculations, this is the environment file for the features.
:type: String Path to file
:return boolean: Indicator of success (True)
"""
if iDims > 1:
self._iDimensions = iDims
#If distance metric is none, check to see if the matrix is a distance matrix
#If so, run NMDS on the distance matrix
#Otherwise return a false and do not run
if(tempDistanceMetric==None):
if(ValidateData.funcIsTrue(self.isRawData)):
print("PCoA:run::Error, no distance metric was specified but the previous load was not of a distance matrix.")
return False
elif(ValidateData.funcIsFalse(self.isRawData)):
self.pcoa = NMDS(dataMatrix, verbosity=0)
return True
#Make sure the distance metric was a valid string type
if(not ValidateData.funcIsValidString(tempDistanceMetric)):
print("PCoA:run::Error, distance metric was not a valid string type.")
return False
#Supported distances
distanceMatrix = None
if(tempDistanceMetric==self.c_SPEARMAN):
distanceMatrix = Metric().funcGetDissimilarity(ldSampleTaxaAbundancies=self.dataMatrix, funcDistanceFunction=lambda u,v: spearmanr(u,v)[0])
if(tempDistanceMetric in [Metric.c_strUnifracUnweighted,Metric.c_strUnifracWeighted]):
distanceMatrix,lsLabels = Metric().funcGetBetaMetric(sMetric=tempDistanceMetric, istrmTree=istrmTree, istrmEnvr=istrmEnvr)
self.lsIDs = lsLabels
else:
distanceMatrix = Metric().funcGetBetaMetric(npadAbundancies=self.dataMatrix, sMetric=tempDistanceMetric)
if(ValidateData.funcIsFalse(distanceMatrix)):
print "PCoA:run::Error, when generating distance matrix."
return False
# Make squareform
distanceMatrix = squareform(distanceMatrix)
# Writes distance measures if needed.
if strDistanceMatrixFile:
csvrDistance = csv.writer(open(strDistanceMatrixFile, 'w'))
if self.lsIDs:
csvrDistance.writerow(["ID"]+self.lsIDs)
for x in xrange(distanceMatrix.shape[0]):
strId = [self.lsIDs[x]] if self.lsIDs else []
csvrDistance.writerow(strId+distanceMatrix[x].tolist())
self.pcoa = NMDS(distanceMatrix, dimension=max(self._iDimensions,2), verbosity=0)
self.strRecentMetric = tempDistanceMetric
return True
def pcoa_coords(dist_arr):
arr = NMDS(dist_arr, verbosity=0).getPoints()
return map(list, arr)
