def find_vstructures(bfile, pfile, gfile, cisfile, assoc0file, window, vfile,
startTraitIdx, nTraits):
"""
running association scan
input:
pfile : phenotype file
cfile : covariance file
ffile : fixed effects file
cisfile : file containing cis anchors
assoc0file : file containing the results from the initial association scan
vfile : file containing v-structures
"""
preader = phenoReader.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
model = gnetlmm.GNetLMM(preader, greader, window=window)
genecorr_reader = reader.FileReader(gfile + '.pv')
model.set_genecorr_reader(genecorr_reader)
assoc0Reader = reader.FileReader(assoc0file + '.pv')
model.set_assoc0_reader(assoc0Reader)
model.load_cis_anchors(cisfile)
model.find_vstructures(startTraitIdx, nTraits)
model.save_vstructures(vfile + '.csv')
def gene_has_anchor(bfile,
pfile,
assoc0,
anchor_thresh,
anchorfile,
window,
cis=True):
"""
tests if a gene has a cis anchor
input:
bfile : binary bed file (bfile.bed, bfile.bim and bfile.fam are required)
pfile : phenotype file
assoc0file : basefilename of initial association scan
anchor_thresh : thrshold for anchor-associations
anchor_file : filename for saving cis assocaitions
window : maximal distance between cis-snp and gene
cis : if set, look for cis-associations only (default: True)
"""
preader = phenoReaderFile.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
model = gnetlmm.GNetLMM(preader, greader, window=window)
assoc0Reader = reader.FileReader(assoc0 + '.pv')
model.set_assoc0_reader(assoc0Reader)
model.gene_has_anchor(anchor_thresh, cis)
model.save_anchors(anchorfile)
def scan(bfile, pfile, cfile, ffile, vfile, assocfile, startTraitIdx, nTraits):
"""
running association scan
input:
bfile : basefilename of plink file
pfile : phenotype file
cfile : covariance file
ffile : fixed effects file
vfile : file containing vstructures
assocfile : file for saving results
"""
K = None
if cfile is not None:
K = np.loadtxt(cfile)
Covs = None
if ffile is not None:
Covs = np.loadtxt(ffile)
preader = phenoReader.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
model = gnetlmm.GNetLMM(preader, greader, Covs=Covs, K=K)
model.load_vstructures(vfile + ".csv")
model.update_associations(startTraitIdx, nTraits)
model.save_updates(assocfile)
def merge_assoc0_scan(assoc0file, nSnps, bfile):
"""
merging associations files
input:
assoc0 : basename of assoc0 results
nSnps : number of SNPs in each block
bfile : basename of bed file
"""
greader = bedReader.BedReader(bfile)
F = greader.get_nrows()
fn_beta = []
fn_pv = []
f = 0
while f < F:
_fn = assoc0file + '.startSnp_%d' % f
_fn_beta = '%s.beta.matrix' % _fn
_fn_pv = '%s.pv.matrix' % _fn
assert os.path.exists(_fn_beta), 'File %s is missing' % (_fn_beta)
assert os.path.exists(_fn_pv), 'File %s is missing' % (_fn_pv)
fn_beta.append(_fn_beta)
fn_pv.append(_fn_pv)
f += nSnps
merge_files(fn_beta, assoc0file + '.beta.matrix')
merge_files(fn_pv, assoc0file + '.pv.matrix')
def __init__(self, bfile, dist_min=100):
"""
initializes object
input:
bfile : basefilename of plink file
dist_causal_snps : minimal distance between two causal SNPs
"""
self.dist_min = dist_min
self.bfile = bfile
self.genoreader = bedReader.BedReader(self.bfile)
self.chrom = self.genoreader.getSnpChrom()
self.pos = self.genoreader.getSnpPos()
self.rs = self.genoreader.getSnpIds()
self.N = self.genoreader.fam.shape[0]
self.F = self.chrom.shape[0]
def find_vstructures(bfile, pfile, gfile, anchorfile, assoc0file, window,
vfile, startTraitIdx, nTraits, corr_thresh, ind_thresh,
max_genes):
"""
running association scan
input:
bfile : blink file
pfile : phenotype file
gfile : correlation file
anchorfile : file containing anchors
assoc0file : file containing the results from the initial association scan
window : window size
vfile : file containing v-structures
startTraitIdx : index of first trait to be analyzed
nTraits : number of traits to be analyzed
corr_thresh : q-value for calling a correlation significant
ind_thresh : q-value for calling a correlation not significant
max_genes : maximal number of genes in the conditioning set
"""
preader = phenoReaderFile.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
model = gnetlmm.GNetLMM(preader,
greader,
window=window,
thresh_corr=corr_thresh,
thresh_ind=ind_thresh)
genecorr_reader = reader.FileReader(gfile + '.pv')
model.set_genecorr_reader(genecorr_reader)
assoc0Reader = reader.FileReader(assoc0file + '.pv')
model.set_assoc0_reader(assoc0Reader)
model.load_anchors(anchorfile)
model.find_vstructures(startTraitIdx, nTraits, max_genes)
if np.isfinite(nTraits): vfile += ".startTrait_%d" % startTraitIdx
model.save_vstructures(vfile + '.csv')
def plot_power(bfile, pfile, assoc0file, assocfile, plotfile, window):
# reading out p-values
score = {}
assoc0Reader = reader.FileReader(assoc0file + '.pv')
score['LMM'] = -np.log10(assoc0Reader.getMatrix())
assocReader = reader.FileReader(assocfile + '.pv')
score['GNetLMM'] = -np.log10(assocReader.getMatrix())
# get network
Agene = np.loadtxt(pfile + '.Agene')
# gene info
preader = phenoReader.PhenoReaderFile(pfile)
gene_start = preader.getGeneStart()
gene_end = preader.getGeneEnd()
gene_chrom = preader.getGeneChrom()
# snp info
breader = bedReader.BedReader(bfile)
snp_pos = breader.getSnpPos()
snp_chrom = breader.getSnpChrom()
P_cis, P_trans = get_groundtruth(Agene, snp_pos, snp_chrom, gene_start,
gene_chrom, window)
# delete cis-associations
for key in score.keys():
score[key][P_cis] = 0
# compute receiver operator characteristics
FPR = {}
TPR = {}
for key in score.keys():
FPR[key], TPR[key] = roc.roc(P_trans, score[key])
# plotting results
plotting.plotROCcurve(['LMM', 'GNetLMM'],
TPR,
FPR,
xlim=(0, 0.05),
ylim=(0, 0.42),
fn=plotfile)
def marginal_genecorr(bfile, pfile, gfile):
"""
running marginal gene-gene correlations
Input:
bfile : binary bed file (bfile.bed, bfile.bim and bfile.fam are required)
pfile : phenotype file
cfile : covariance matrix file
ffile : covariates file
gfile : basename of output file
"""
preader = phenoReader.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
model = gnetlmm.GNetLMM(preader, greader)
corr, pv = model.marginal_gene_correlations()
write = writer.Writer(gfile + '.pv')
write.writeMatrix(pv, fmt='%.4e')
write = writer.Writer(gfile + '.corr')
write.writeMatrix(corr, fmt='%.4f')
def initial_scan(bfile,
pfile,
cfile,
ffile,
assoc0file,
startSnpIdx=0,
nSnps=np.inf,
memory_efficient=False):
"""
running initial scan using a standard linear mixed model
Input:
bfile : binary bed file (bfile.bed, bfile.bim and bfile.fam are required)
pfile : phenotype file
cfile : covariance matrix file
ffile : covariates file
assoc0file : basename of output file
"""
K = None
if cfile is not None:
K = np.loadtxt(cfile)
Covs = None
if ffile is not None:
Covs = np.loadtxt(ffile)
if np.isfinite(nSnps): assoc0file += ".startSnp_%d" % startSnpIdx
preader = phenoReaderFile.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
model = gnetlmm.GNetLMM(preader, greader, K=K, Covs=Covs)
beta0, pv0 = model.initial_scan(startSnpIdx, nSnps, memory_efficient)
write = writer.Writer(assoc0file + '.pv')
write.writeMatrix(pv0, fmt='%.4e')
write = writer.Writer(assoc0file + '.beta')
write.writeMatrix(beta0, fmt='%.4f')
def create_nice_output(vfile, bfile, pfile, assoc0file, assocfile, blockfile, outfile):
"""
creating human readbable output file
"""
preader = phenoReaderFile.PhenoReaderFile(pfile)
greader = bedReader.BedReader(bfile)
snp_ids = greader.getSnpIds()
gene_ids = preader.getGeneIds()
vstruct = vstructures.VstructureFile(vfile + '.csv')
assoc0Reader = reader.FileReader(assoc0file + '.pv')
assocReader = assoc_results.AssocResultsList()
assocReader.load_csv(assocfile + '.csv')
blockReader = None
if blockfile is not None:
blockReader = assoc_results.AssocResultsList()
blockReader.load_csv(blockfile + '.csv')
f = open(outfile + '.csv','w')
if blockReader is None:
header = "Anchor Snp\t Anchor Gene\t Focal Gene\t Orthogonal Genes\t pv(LMM) \t pv(GNetLMM)"
else:
header = "Anchor Snp\t Anchor Gene\t Focal Gene\t Orthogonal Genes\t pv(LMM) \t pv(GNetLMM) \t pv(BlockLMM)"
if assocReader.var_snps is not None: header += '\t varSnp(GnetLMM)'
if assocReader.var_covs is not None: header += '\t varCovs(GnetLMM)'
if assocReader.var_genes is not None: header += '\t varGenes(GnetLMM)'
if blockReader is not None:
if blockReader.var_snps is not None: header += '\t varSnp(BlockLMM)'
if blockReader.var_covs is not None: header += '\t varCovs(BlockLMM)'
if blockReader.var_genes is not None: header += '\t varGenes(BlockLMM)'
header += "\n"
f.write(header)
for idx_focal_gene, idx_anchor_snp, idx_orth_gene, idx_anchor_gene in vstruct.iterator(full=True):
focal_gene = ",".join(gene_ids[idx_focal_gene])
orth_gene = ",".join(gene_ids[idx_orth_gene])
anchor_snps = snp_ids[idx_anchor_snp]
for i in range(len(anchor_snps)):
try:
anchor_gene = gene_ids[idx_anchor_gene[i]]
except:
anchor_gene = ""
pv_lmm = assoc0Reader.getRows([idx_anchor_snp[i]])[:,idx_focal_gene][0,0]
idx_gnet = np.logical_and(assocReader.focal_gene==idx_focal_gene, assocReader.snp_anchor==idx_anchor_snp[i])
if blockReader is not None:
idx_block = np.logical_and(blockReader.focal_gene==idx_focal_gene, blockReader.snp_anchor==idx_anchor_snp[i])
if blockReader is None:
line = "%s\t%s\t%s\t%s\t%.4e\t%.4e"%(anchor_snps[i],anchor_gene,focal_gene, orth_gene, pv_lmm, assocReader.pv[idx_gnet])
else:
line = "%s\t%s\t%s\t%s\t%.4e\t%.4e\t%.4e"%(anchor_snps[i],anchor_gene,focal_gene, orth_gene, pv_lmm, assocReader.pv[idx_gnet], blockReader.pv[idx_block])
if assocReader.var_snps is not None: line += '\t%.4e'%assocReader.var_snps[idx_gnet]
if assocReader.var_covs is not None: line += '\t%.4e'%assocReader.var_covs[idx_gnet]
if assocReader.var_genes is not None: line += '\t%.4e'%assocReader.var_genes[idx_gnet]
if blockReader is not None:
if blockReader.var_snps is not None: line += '\t%.4e'%blockReader.var_snps[idx_block]
if blockReader.var_covs is not None: line += '\t%.4e'%blockReader.var_covs[idx_block]
if blockReader.var_genes is not None: line += '\t%.4e'%blockReader.var_genes[idx_block]
line += "\n"
f.write(line)
f.close()
