def run(self):
"""
2009-2-2
"""
if self.debug:
import pdb
pdb.set_trace()
db = Stock_250kDB.Stock_250kDB(drivername=self.drivername,
username=self.db_user,
password=self.db_passwd,
hostname=self.hostname,
database=self.dbname,
schema=self.schema)
db.setup(create_tables=False)
session = db.session
session.begin()
PC_data = Association.getPCFromFile(self.eigen_vector_fname)
eigen_value_ls = Association.getEigenValueFromFile(
self.eigen_value_fname)
eigen_value_ls = numpy.array(eigen_value_ls)
explained_var = eigen_value_ls / numpy.sum(eigen_value_ls)
self.putPCDataIntoDB(db, PC_data, self.call_method_id)
self.putEigenValuesIntoDB(db, eigen_value_ls, explained_var,
self.call_method_id)
if self.commit:
session.flush()
session.commit()
def make_association_to_part(part_no, filename):
if part_no == 1:
association = Association(SentenceAssociationStrategy, filename)
elif part_no == 2:
association = Association(WindowAssociationStrategy, filename, arg=2)
else:
association = Association(DependencyEdgeAssocaition, filename)
return association
def IntLMOnTwoSNPs(cls, snp1_id, gene1_id, genotype_ls1, snp2_id, gene2_id, genotype_ls2, phenotype_index, phenotype_ls, \
min_data_point=3):
"""
2009-2-8
interaction detection linear model
y = b + SNP1xSNP2 + SNP1 + SNP2 + e
interaction is the 1st term. therefore the pvalue directly returned is also for this term.
"""
return_ls = []
genotype_ls1 = numpy.resize(genotype_ls1, [len(genotype_ls1), 1])
genotype_ls2 = numpy.resize(genotype_ls2, [len(genotype_ls2), 1])
snp_int_matrix = genotype_ls1 * genotype_ls2
genotype_ls = numpy.hstack(
(snp_int_matrix, genotype_ls1,
genotype_ls2)) #interaction variable is the 1st position
pdata = Association.linear_model(genotype_ls,
phenotype_ls,
min_data_point,
snp_index=snp1_id + snp2_id)
if pdata:
pdata = PassingData(snp1_id=snp1_id, gene1_id=gene1_id, snp2_id=snp2_id, gene2_id=gene2_id, pvalue=pdata.pvalue,\
count1=pdata.count_ls[0], count2=pdata.count_ls[1], bool_type=None, phenotype_index=phenotype_index,\
var_perc=pdata.var_perc, coeff_list=pdata.coeff_list, coeff_p_value_list=pdata.coeff_p_value_list)
return_ls.append(pdata)
return return_ls
def registerCustomExecutables(self, workflow=None):
"""
2012.6.5
"""
AbstractVariationWorkflow.registerCustomExecutables(self,
workflow=workflow)
namespace = self.namespace
version = self.version
operatingSystem = self.operatingSystem
architecture = self.architecture
clusters_size = self.clusters_size
site_handler = self.site_handler
executableClusterSizeMultiplierList = [
] #2012.8.7 each cell is a tuple of (executable, clusterSizeMultipler (0 if u do not need clustering)
Association = Executable(namespace=namespace, name="Association", version=version, \
os=operatingSystem, arch=architecture, installed=True)
Association.addPFN(
PFN(
"file://" + os.path.join(self.variationSrcPath,
"association/Association.py"),
site_handler))
executableClusterSizeMultiplierList.append((Association, 0))
Results2DB_250k = Executable(namespace=namespace, name="Results2DB_250k", version=version, \
os=operatingSystem, arch=architecture, installed=True)
Results2DB_250k.addPFN(
PFN(
"file://" +
os.path.join(self.variationSrcPath, "db/Results2DB_250k.py"),
site_handler))
executableClusterSizeMultiplierList.append((Results2DB_250k, 0))
OutputPhenotype = Executable(namespace=namespace, name="OutputPhenotype", version=version, \
os=operatingSystem, arch=architecture, installed=True)
OutputPhenotype.addPFN(
PFN(
"file://" + os.path.join(self.variationSrcPath,
"db/output/OutputPhenotype.py"),
site_handler))
executableClusterSizeMultiplierList.append((OutputPhenotype, 0))
self.addExecutableAndAssignProperClusterSize(
executableClusterSizeMultiplierList,
defaultClustersSize=self.clusters_size)
def IntLMOnTwoSNPs(cls, snp1_id, gene1_id, genotype_ls1, snp2_id, gene2_id, genotype_ls2, phenotype_index, phenotype_ls, \ min_data_point=3): """ 2009-2-8 interaction detection linear model y = b + SNP1xSNP2 + SNP1 + SNP2 + e interaction is the 1st term. therefore the pvalue directly returned is also for this term. """ return_ls = [] genotype_ls1 = numpy.resize(genotype_ls1, [len(genotype_ls1),1]) genotype_ls2 = numpy.resize(genotype_ls2, [len(genotype_ls2),1]) snp_int_matrix = genotype_ls1*genotype_ls2 genotype_ls = numpy.hstack((snp_int_matrix, genotype_ls1, genotype_ls2)) #interaction variable is the 1st position pdata = Association.linear_model(genotype_ls, phenotype_ls, min_data_point, snp_index=snp1_id+snp2_id) if pdata: pdata = PassingData(snp1_id=snp1_id, gene1_id=gene1_id, snp2_id=snp2_id, gene2_id=gene2_id, pvalue=pdata.pvalue,\ count1=pdata.count_ls[0], count2=pdata.count_ls[1], bool_type=None, phenotype_index=phenotype_index,\ var_perc=pdata.var_perc, coeff_list=pdata.coeff_list, coeff_p_value_list=pdata.coeff_p_value_list) return_ls.append(pdata) return return_ls
def subplotLatLonPhenVsPC(self, ecotype_info, StrainID2PCAPosInfo, phenData, phenotype_col_index, \
phenotype_cmap, phenotype_norm, which_figure=0, sub_title='',\
ydata='latitude', which_PC_index=0, no_of_rows=2, dot_size=10, alpha=0.7):
"""
2008-12-08
one single subplot in plotLatLonPhenVsPC()
"""
ax = pylab.subplot(no_of_rows,2,which_figure, frameon=False)
pylab.ylabel(ydata)
pylab.grid(True, alpha=0.3)
x_ls = []
y_ls = []
for strain_id in StrainID2PCAPosInfo.strain_id_ls:
ecotype_id = int(strain_id)
ecotype_obj = ecotype_info.ecotype_id2ecotype_obj.get(ecotype_id)
if ecotype_obj:
lat, lon = ecotype_obj.latitude, ecotype_obj.longitude
y_value = getattr(ecotype_obj, ydata, None)
else:
sys.stderr.write("Warning: Ecotype %s not in ecotype_info (fetched from stock db).\n"%ecotype_id)
continue
#x_value = StrainID2PCAPosInfo.PC_matrix[row_index][which_PC_index]
if which_PC_index==1:
x_value = StrainID2PCAPosInfo.strain_id2pca_x[strain_id]
xlabel = 'PC%s'%(which_PC_index+1)
elif which_PC_index==0:
x_value = StrainID2PCAPosInfo.strain_id2pca_y[strain_id]
xlabel = 'PC%s'%(which_PC_index+1)
elif which_PC_index=='longitude':
x_value = lon
xlabel = which_PC_index
elif which_PC_index =='latitude':
x_value = lat
xlabel = which_PC_index
#img_y_pos = StrainID2PCAPosInfo.strain_id2img_y_pos[strain_id]
#strain color according to phenotype
phenotype_row_index = phenData.row_id2row_index[strain_id]
phenotype = phenData.data_matrix[phenotype_row_index][phenotype_col_index]
strain_fc = phenotype_cmap(phenotype_norm(phenotype))
if numpy.isnan(phenotype):
linewidth=0.5
strain_fc = 'w'
edgecolor = 'k'
_alpha = 0 #facecolor gets very transparent
else:
linewidth=0
strain_fc = strain_fc
edgecolor = 'k'
_alpha = alpha
if ydata=='phenotype':
y_value = phenotype
if numpy.isnan(phenotype): #can't do regression or plot
continue
if y_value is None or numpy.isnan(y_value):
continue
pylab.scatter([x_value],[y_value], s=dot_size, linewidth=linewidth, facecolor=strain_fc, alpha=_alpha, zorder=10)
x_ls.append(x_value)
y_ls.append(y_value)
xlim = ax.get_xlim()
ylim = ax.get_ylim()
lm_result = Association.pure_linear_model(x_ls, y_ls)
try:
pvalue = '%.2f'%-math.log10(lm_result.pvalue)
except:
pvalue = '%s'%lm_result.pvalue
beta0 = lm_result.coeff_list[0]
beta = lm_result.coeff_list[1]
#draw a line showing the trend
x_lm_ls = [min(x_ls), max(x_ls)]
lm_func = lambda x: lm_result.coeff_list[0]+lm_result.coeff_list[1]*x
y_lm_ls = map(lm_func, x_lm_ls)
ax.plot(x_lm_ls, y_lm_ls, alpha=0.5)
ax.set_xlim(xlim)
ax.set_ylim(ylim)
pylab.xlabel('%s, pvalue=%s, beta=%.3f'%(xlabel, pvalue, beta))
def __init__(self, **keywords):
"""
2009-3-20
"""
Association.__init__(self, **keywords)
def __init__(self, **keywords): """ 2009-3-20 """ Association.__init__(self, **keywords)