def main(argv):
parser = argparse.ArgumentParser()
parser.add_argument("scg_cov_file", help="input core gene coverages")
parser.add_argument("gamma_star_file",
help="input MAP estimate frequencies")
parser.add_argument("cov_file", help="mean contig/genes coverages")
parser.add_argument("epsilon_file", help="predicted transition matrix")
parser.add_argument(
'-s',
'--random_seed',
default=23724839,
type=int,
help=
("specifies seed for numpy random number generator defaults to 23724839 applied after random filtering"
))
parser.add_argument(
'-e',
'--eta_max',
default=2,
type=int,
help=("maximum contig contig count for sampler defaults to 2"))
parser.add_argument('-i',
'--iter_max',
default=20,
type=int,
help=("number of Gibbs sampling iterations"))
parser.add_argument(
'-m',
'--var_max',
default=1e10,
type=int,
help=
("maximum number of variants to user per contig defaults to all (1e10) if unset"
))
parser.add_argument('-o',
'--output_stub',
type=str,
default="output",
help=("string specifying output file stubs"))
parser.add_argument(
'-g',
'--genomes',
help=("specify validation file of known genome composition"))
parser.add_argument(
'-v',
'--variant_file',
help=("specify file of called variants on genes if available"))
parser.add_argument('--assign_tau', dest='assign_tau', action='store_true')
parser.set_defaults(assign_tau=False)
args = parser.parse_args()
#import ipdb; ipdb.set_trace()
output_stub = args.output_stub
log_file_name = output_stub + "_log_file.txt"
logging.basicConfig(
filename=log_file_name,
level=logging.INFO,
filemode='w', # Overwrites old log file
format='%(asctime)s:%(levelname)s:%(name)s:%(message)s')
#seed random number generators
logging.info('Seed random number generators = %d' % (args.random_seed))
prng = RandomState(args.random_seed)
sampletau.initRNG()
sampletau.setRNG(args.random_seed)
#read in data
logging.info('Read in SCG coverages from %s' % (args.scg_cov_file))
scg_cov = p.read_csv(args.scg_cov_file, header=0, index_col=0)
logging.info('Read gamma from %s' % (args.gamma_star_file))
gamma_star = p.read_csv(args.gamma_star_file, header=0, index_col=0)
logging.info('Read gene coverages from %s' % (args.cov_file))
cov = p.read_csv(args.cov_file, header=0, index_col=0)
logging.info('Read epsilon from %s' % (args.epsilon_file))
epsilon = p.read_csv(args.epsilon_file, header=0, index_col=0)
epsilon_matrix = epsilon.as_matrix()
if args.variant_file is not None:
logging.info('Read variants from %s' % (args.variant_file))
variants = p.read_csv(args.variant_file, header=0, index_col=0)
else:
variants = None
gamma_names = gamma_star.index.values
scg_names = scg_cov.index.values
intersect_names1 = sorted(intersect(gamma_names, scg_names))
intersect_names = sorted(intersect(cov.columns.values, intersect_names1))
logging.info('Found %d samples common to all 3 input files' %
(len(intersect_names)))
scg_cov = scg_cov.reindex(intersect_names)
gamma_star = gamma_star.reindex(intersect_names)
total_mean = scg_cov['mean'].as_matrix()
total_sd = scg_cov['sd'].as_matrix()
#renormalise gamma matrix
gamma_star_matrix = gamma_star.as_matrix()
row_sums = gamma_star_matrix.sum(axis=1)
gamma_star_matrix = gamma_star_matrix / row_sums[:, np.newaxis]
delta = np.multiply(gamma_star_matrix, total_mean[:, np.newaxis])
#reorder coverage matrix
cov = cov[intersect_names]
cov_matrix = cov.as_matrix()
logging.info('Perform KL estimation of contig counts')
klassign = KLAssign(prng, cov_matrix, delta)
klassign.factorize()
if variants is not None:
expanded_names = expand_sample_names(intersect_names)
variants_intersect = variants[expanded_names]
#now apply Gaussian Gibbs sampler
#import ipdb; ipdb.set_trace()
etaD = np.rint(klassign.eta)
etaSampler = es.Eta_Sampler(prng,
variants_intersect,
cov,
gamma_star_matrix,
delta,
total_sd,
epsilon_matrix,
etaD,
max_iter=args.iter_max,
max_eta=args.eta_max,
max_var=args.var_max)
etaSampler.update()
etaSampler.update()
#Now assign tau given eta_star
contig_names = cov.index.tolist()
if args.assign_tau is True:
etaSampler.restoreFullVariants()
etaSampler.calcTauStar(etaSampler.eta_star)
(tau_star, tau_mean, pos,
contig_index) = etaSampler.getTauStar(variants)
V = tau_star.shape[0]
tau_res = np.reshape(tau_star, (V, etaSampler.G * 4))
tau_df = p.DataFrame(tau_res, index=contig_index)
tau_df['Position'] = pos
cols = tau_df.columns.tolist()
cols = cols[-1:] + cols[:-1]
tau_df = tau_df[cols]
tau_df.to_csv(output_stub + "_tau_star.csv")
logging.info("Wrote tau star haplotype predictions")
tau_mean_res = np.reshape(tau_mean, (V, etaSampler.G * 4))
tau_mean_df = p.DataFrame(tau_mean_res, index=contig_index)
tau_mean_df['Position'] = pos
cols = tau_mean_df.columns.tolist()
cols = cols[-1:] + cols[:-1]
tau_mean_df = tau_mean_df[cols]
tau_mean_df.to_csv(output_stub + "_tau_mean.csv")
logging.info("Wrote tau mean haplotype predictions")
etaD_df = p.DataFrame(etaD, index=contig_names)
etaD_df.to_csv(output_stub + "etaD_df.csv")
etaS_df = p.DataFrame(etaSampler.eta_star, index=contig_names)
etaS_df.to_csv(output_stub + "etaS_df.csv")
etaM_df = p.DataFrame(np.mean(etaSampler.eta_store, axis=0),
index=contig_names)
etaM_df.to_csv(output_stub + "etaM_df.csv")
eta_df = p.DataFrame(klassign.eta, index=contig_names)
eta_df.to_csv(output_stub + "eta_df.csv")
if args.genomes:
genomes = p.read_csv(args.genomes, header=0, index_col=0)
genomes = genomes.loc[contig_names]
genomes_M = genomes.as_matrix()
genomes_D = np.copy(genomes_M)
#genomes_D[genomes_D < 0.5] = 0.
#genomes_D[genomes_D >= 0.5] = 1.0
#etaD[etaD < 0.5] = 0.
#etaD[etaD >= 0.5] = 1.0
(dtotal, dacc, dacc_array) = compGenes(etaD, genomes_D)
(stotal, sacc, sacc_array) = compGenes(etaSampler.eta_star, genomes_D)
#np.savetxt(sys.stdout, dacc, fmt='%.4f')
#np.savetxt(sys.stdout, sacc, fmt='%.4f')
logging.info('KL accurracy = %f' % (dtotal))
logging.info('Gibbs sampler accurracy = %f' % (stotal))
def main(argv):
parser = argparse.ArgumentParser()
parser.add_argument("scg_cov_file", help="input core gene coverages")
parser.add_argument("gamma_star_file", help="input MAP estimate frequencies")
parser.add_argument("cov_file", help="mean contig/genes coverages")
parser.add_argument("epsilon_file", help="predicted transition matrix")
parser.add_argument('-s','--random_seed',default=23724839, type=int,
help=("specifies seed for numpy random number generator defaults to 23724839 applied after random filtering"))
parser.add_argument('-e','--eta_max',default=2, type=int,
help=("maximum contig contig count for sampler defaults to 2"))
parser.add_argument('-i','--iter_max',default=20, type=int,
help=("number of Gibbs sampling iterations"))
parser.add_argument('-m','--var_max',default=1e10, type=int,
help=("maximum number of variants to user per contig defaults to all (1e10) if unset"))
parser.add_argument('-o','--output_stub', type=str, default="output",
help=("string specifying output file stubs"))
parser.add_argument('-g','--genomes',
help=("specify validation file of known genome composition"))
parser.add_argument('-v','--variant_file',
help=("specify file of called variants on genes if available"))
parser.add_argument('--assign_tau', dest='assign_tau', action='store_true')
parser.set_defaults(assign_tau=False)
args = parser.parse_args()
#import ipdb; ipdb.set_trace()
output_stub = args.output_stub
log_file_name = output_stub+"_log_file.txt"
logging.basicConfig(
filename=log_file_name,
level=logging.INFO,
filemode='w', # Overwrites old log file
format='%(asctime)s:%(levelname)s:%(name)s:%(message)s'
)
#seed random number generators
logging.info('Seed random number generators = %d' %(args.random_seed))
prng = RandomState(args.random_seed)
sampletau.initRNG()
sampletau.setRNG(args.random_seed)
#read in data
logging.info('Read in SCG coverages from %s' %(args.scg_cov_file))
scg_cov = p.read_csv(args.scg_cov_file, header=0, index_col=0)
logging.info('Read gamma from %s' %(args.gamma_star_file))
gamma_star = p.read_csv(args.gamma_star_file, header=0, index_col=0)
logging.info('Read gene coverages from %s' %(args.cov_file))
cov = p.read_csv(args.cov_file, header=0, index_col=0)
logging.info('Read epsilon from %s' %(args.epsilon_file))
epsilon = p.read_csv(args.epsilon_file, header=0, index_col=0)
epsilon_matrix = epsilon.as_matrix()
if args.variant_file is not None:
logging.info('Read variants from %s' %(args.variant_file))
variants = p.read_csv(args.variant_file, header=0, index_col=0)
else:
variants = None
gamma_names = gamma_star.index.values
scg_names = scg_cov.index.values
intersect_names1 = sorted(intersect(gamma_names,scg_names))
intersect_names = sorted(intersect(cov.columns.values,intersect_names1))
logging.info('Found %d samples common to all 3 input files' %(len(intersect_names)))
scg_cov = scg_cov.reindex(intersect_names)
gamma_star = gamma_star.reindex(intersect_names)
total_mean = scg_cov['mean'].as_matrix()
total_sd = scg_cov['sd'].as_matrix()
#renormalise gamma matrix
gamma_star_matrix = gamma_star.as_matrix()
row_sums = gamma_star_matrix.sum(axis=1)
gamma_star_matrix = gamma_star_matrix / row_sums[:, np.newaxis]
delta = np.multiply(gamma_star_matrix,total_mean[:,np.newaxis])
#reorder coverage matrix
cov = cov[intersect_names]
cov_matrix = cov.as_matrix()
logging.info('Perform KL estimation of contig counts')
klassign = KLAssign(prng,cov_matrix,delta)
klassign.factorize()
if variants is not None:
expanded_names = expand_sample_names(intersect_names)
variants_intersect = variants[expanded_names]
#now apply Gaussian Gibbs sampler
#import ipdb; ipdb.set_trace()
etaD = np.rint(klassign.eta)
etaSampler = es.Eta_Sampler(prng, variants_intersect, cov, gamma_star_matrix, delta, total_sd, epsilon_matrix, etaD,
max_iter=args.iter_max,max_eta=args.eta_max, max_var=args.var_max)
etaSampler.update()
etaSampler.update()
#Now assign tau given eta_star
contig_names = cov.index.tolist()
if args.assign_tau is True:
etaSampler.restoreFullVariants()
etaSampler.calcTauStar(etaSampler.eta_star)
(tau_star,tau_mean,pos,contig_index) = etaSampler.getTauStar(variants)
V = tau_star.shape[0]
tau_res = np.reshape(tau_star,(V,etaSampler.G*4))
tau_df = p.DataFrame(tau_res,index=contig_index)
tau_df['Position'] = pos
cols = tau_df.columns.tolist()
cols = cols[-1:] + cols[:-1]
tau_df = tau_df[cols]
tau_df.to_csv(output_stub+"_tau_star.csv")
logging.info("Wrote tau star haplotype predictions")
tau_mean_res = np.reshape(tau_mean,(V,etaSampler.G*4))
tau_mean_df = p.DataFrame(tau_mean_res,index=contig_index)
tau_mean_df['Position'] = pos
cols = tau_mean_df.columns.tolist()
cols = cols[-1:] + cols[:-1]
tau_mean_df = tau_mean_df[cols]
tau_mean_df.to_csv(output_stub+"_tau_mean.csv")
logging.info("Wrote tau mean haplotype predictions")
etaD_df = p.DataFrame(etaD,index=contig_names)
etaD_df.to_csv(output_stub+"etaD_df.csv")
etaS_df = p.DataFrame(etaSampler.eta_star,index=contig_names)
etaS_df.to_csv(output_stub+"etaS_df.csv")
etaM_df = p.DataFrame(np.mean(etaSampler.eta_store,axis=0),index=contig_names)
etaM_df.to_csv(output_stub+"etaM_df.csv")
eta_df = p.DataFrame(klassign.eta,index=contig_names)
eta_df.to_csv(output_stub+"eta_df.csv")
if args.genomes:
genomes = p.read_csv(args.genomes, header=0, index_col=0)
genomes = genomes.loc[contig_names]
genomes_M = genomes.as_matrix()
genomes_D = np.copy(genomes_M)
#genomes_D[genomes_D < 0.5] = 0.
#genomes_D[genomes_D >= 0.5] = 1.0
#etaD[etaD < 0.5] = 0.
#etaD[etaD >= 0.5] = 1.0
(dtotal, dacc,dacc_array) = compGenes(etaD, genomes_D)
(stotal, sacc,sacc_array) = compGenes(etaSampler.eta_star, genomes_D)
#np.savetxt(sys.stdout, dacc, fmt='%.4f')
#np.savetxt(sys.stdout, sacc, fmt='%.4f')
logging.info('KL accurracy = %f' %(dtotal))
logging.info('Gibbs sampler accurracy = %f' %(stotal))