def do_one(args):
matrix_type = args.matrix_type
rbp_ps = args.rbp_ps
examine_tops = args.examine_tops
result = subprocess.call("Rscript --version 2> /dev/null", shell=True)
if result == 127:
sys.stderr.write("Rscript is not found")
sys.stderr.flush()
sys.exit(1)
if not os.path.isdir(args.output):
os.mkdir(args.output)
log_file = "%s/%s" % (args.output, args.log_file)
logger = logging.getLogger("prep_fast")
logger.setLevel(logging.DEBUG)
if not logger.hasHandlers():
handler = logging.FileHandler(log_file)
handler.setFormatter(
logging.Formatter(
"%(asctime)s - %(name)s - %(levelname)s - %(message)s"))
logger.addHandler(handler)
if args.verbose:
logger.addHandler(logging.StreamHandler())
t_overwrite = args.overwrite_input_bin
#if there is no binary input file, create one
check_required = ["expr.npy", "Xcen.npy", "genes.npy"]
for rbp_p in args.rbp_ps:
check_required.append("t_matrix_%s_%.2f.npy" %
(args.matrix_type, rbp_p))
for cr in check_required:
if not os.path.isfile("%s/%s" % (args.output, cr)):
t_overwrite = True
break
expr, Xcen, genes, t_matrix = None, None, None, None
if t_overwrite:
expr, genes, Xcen = read_matrix(f_expr=args.expr,
f_Xcen=args.centroid,
logger=logger)
logger.info(
"Calculate all pairwise Euclidean distance between cells using their physical coordinates"
)
euc = squareform(pdist(Xcen, metric="euclidean"))
for rbp_p in args.rbp_ps:
logger.info("For rbp_p %.2f:" % rbp_p)
t_matrix = spatial_genes.rank_transform_matrix(
euc,
reverse=False,
rbp_p=rbp_p,
matrix_type=matrix_type,
logger=logger)
np.save(
"%s/t_matrix_%s_%.2f.npy" %
(args.output, args.matrix_type, rbp_p), t_matrix)
np.save("%s/expr.npy" % args.output, expr)
np.save("%s/Xcen.npy" % args.output, Xcen)
np.save("%s/genes.npy" % args.output, genes)
else:
logger.info("Using existing input binaries...")
expr = np.load("%s/expr.npy" % args.output)
Xcen = np.load("%s/Xcen.npy" % args.output)
genes = np.load("%s/genes.npy" % args.output)
ncell = Xcen.shape[0]
for rbp_p in args.rbp_ps:
for examine_top in args.examine_tops:
outdir = "%s/result_fast_sim_5000_%.2f_%.3f" % (args.output, rbp_p,
examine_top)
if matrix_type == "dissim":
outdir = "%s/result_fast_5000_%.2f_%.3f" % (args.output, rbp_p,
examine_top)
if not os.path.isdir(outdir):
os.mkdir(outdir)
source_path = os.path.dirname(silhouetteRank.__file__)
if not os.path.isfile("%s/do_kmeans.R" % outdir):
copyfile("%s/do_kmeans.R" % source_path,
"%s/do_kmeans.R" % outdir)
if not os.path.isfile("%s/do_gpd.R" % outdir):
copyfile("%s/do_gpd.R" % source_path, "%s/do_gpd.R" % outdir)
if not os.path.isfile("%s/qval.R" % outdir):
copyfile("%s/qval.R" % source_path, "%s/qval.R" % outdir)
def do_one(args):
matrix_type = args.matrix_type
rbp_p = args.rbp_p
if not os.path.isdir(args.output):
os.mkdir(args.output)
logdir = "%s/logs.fast" % args.output
if not os.path.isdir(logdir):
os.mkdir(logdir)
log_file = "%s/real_%.2f_%.3f.out" % (logdir, args.rbp_p, args.examine_top)
logger = logging.getLogger("real_fast_%.2f_%.3f" %
(args.rbp_p, args.examine_top))
logger.setLevel(logging.DEBUG)
if not logger.hasHandlers():
handler = logging.FileHandler(log_file, "w")
handler.setFormatter(
logging.Formatter(
"%(asctime)s - %(name)s - %(levelname)s - %(message)s"))
logger.addHandler(handler)
if args.verbose:
logger.addHandler(logging.StreamHandler())
t_overwrite = args.overwrite_input_bin
check_required = [
"expr.npy", "Xcen.npy", "genes.npy",
"t_matrix_%s_%.2f.npy" % (args.matrix_type, args.rbp_p)
]
for cr in check_required:
if not os.path.isfile("%s/%s" % (args.output, cr)):
t_overwrite = True
break
expr, Xcen, genes, t_matrix = None, None, None, None
if t_overwrite:
expr, Xcen, genes = read(f_expr=args.expr,
f_Xcen=args.centroid,
logger=logger)
logger.info(
"Calculate all pairwise Euclidean distance between cells using their physical coordinates"
)
euc = squareform(pdist(Xcen, metric="euclidean"))
logger.info(
"Rank transform euclidean distance, and then apply exponential transform"
)
t_matrix = spatial_genes.rank_transform_matrix(euc,
reverse=False,
rbp_p=rbp_p,
matrix_type=matrix_type,
logger=logger)
np.save(
"%s/t_matrix_%s_%.2f.npy" %
(args.output, args.matrix_type, args.rbp_p), t_matrix)
np.save("%s/expr.npy" % args.output, expr)
np.save("%s/Xcen.npy" % args.output, Xcen)
np.save("%s/genes.npy" % args.output, genes)
else:
logger.info("Using existing input binaries...")
expr = np.load("%s/expr.npy" % args.output)
Xcen = np.load("%s/Xcen.npy" % args.output)
genes = np.load("%s/genes.npy" % args.output)
t_matrix = np.load("%s/t_matrix_%s_%.2f.npy" %
(args.output, args.matrix_type, args.rbp_p))
logger.info("Compute silhouette metric per gene using fast method")
examine_top = args.examine_top
for t_trial in range(args.num_trials):
res = spatial_genes.calc_silhouette_per_gene_approx(
genes=genes,
expr=expr,
matrix=t_matrix,
matrix_type=matrix_type,
examine_top=examine_top,
logger=logger)
if matrix_type == "sim":
f_name = "%s/silhouette.sim.fast.rbp.%.2f.top.%.3f.%d.txt" % (
args.output, rbp_p, examine_top, t_trial)
else:
f_name = "%s/silhouette.fast.rbp.%.2f.top.%.3f.%d.txt" % (
args.output, rbp_p, examine_top, t_trial)
fw = open(f_name, "w")
for ind, v in enumerate(res):
fw.write("%d\t%s\t%.10f\n" % (ind, v[0], v[1]))
fw.close()
continue
scores = []
f = open("%s/%d" % (outdir, target))
for l in f:
l = l.rstrip("\n")
scores.append(float(l))
f.close()
if len(scores)!=5000:
size_to_do.append(target)
if size_to_do==[]:
sys.exit(0)
expr, genes, Xcen = read_matrix()
ncell = Xcen.shape[0]
sys.stdout.write("Calculate all pairwise Euclidean distance between cells using their physical coordinates\n")
euc = squareform(pdist(Xcen, metric="euclidean"))
sys.stdout.write("Rank transform euclidean distance, and then apply exponential transform\n")
t_matrix = spatial_genes.rank_transform_matrix(euc, reverse=False, rbp_p=rbp_p, matrix_type=matrix_type)
sys.stdout.write("Compute silhouette metric per gene\n")
source_path = os.path.dirname(silhouetteRank.__file__)
if not os.path.isfile("%s/do_gpd.R" % outdir):
copyfile("%s/do_gpd.R" % source_path, "%s/do_gpd.R" % outdir)
if not os.path.isfile("%s/do_kmeans.R" % outdir):
copyfile("%s/do_kmeans.R" % source_path, "%s/do_kmeans.R" % outdir)
res = spatial_genes.random_pattern(matrix=t_matrix, matrix_type=matrix_type, num_cell = ncell, sizes=size_to_do, trials_per_gene=5000, run_gpd=True, outdir=outdir)
def do_one(args):
matrix_type = args.matrix_type
rbp_p = args.rbp_p
examine_top = args.examine_top
#matrix_type = "dissim" # sim or dissim
result = subprocess.call("Rscript --version 2> /dev/null", shell=True)
if result == 127:
sys.stderr.write("Rscript is not found")
sys.stderr.flush()
sys.exit(1)
if not os.path.isdir(args.output):
os.mkdir(args.output)
logdir = "%s/logs" % args.output
if not os.path.isdir(logdir):
os.mkdir(logdir)
log_file = "%s/%.2f_%.3f.out" % (logdir, args.rbp_p, args.examine_top)
logger = logging.getLogger("random_%.2f_%.3f" %
(args.rbp_p, args.examine_top))
logger.setLevel(logging.DEBUG)
t_overwrite = args.overwrite_input_bin
#if there is no binary input file, create one
check_required = [
"expr.npy", "Xcen.npy", "genes.npy",
"t_matrix_%s_%.2f.npy" % (args.matrix_type, args.rbp_p)
]
for cr in check_required:
if not os.path.isfile("%s/%s" % (args.output, cr)):
t_overwrite = True
break
expr, Xcen, genes, t_matrix = None, None, None, None
if t_overwrite:
expr, genes, Xcen = read_matrix(f_expr=args.expr,
f_Xcen=args.centroid,
logger=logger)
logger.info(
"Calculate all pairwise Euclidean distance between cells using their physical coordinates"
)
euc = squareform(pdist(Xcen, metric="euclidean"))
logger.info(
"Rank transform euclidean distance, and then apply exponential transform"
)
t_matrix = spatial_genes.rank_transform_matrix(euc,
reverse=False,
rbp_p=rbp_p,
matrix_type=matrix_type,
logger=logger)
np.save(
"%s/t_matrix_%s_%.2f.npy" %
(args.output, args.matrix_type, args.rbp_p), t_matrix)
np.save("%s/expr.npy" % args.output, expr)
np.save("%s/Xcen.npy" % args.output, Xcen)
np.save("%s/genes.npy" % args.output, genes)
else:
logger.info("Using existing input binaries...")
expr = np.load("%s/expr.npy" % args.output)
Xcen = np.load("%s/Xcen.npy" % args.output)
genes = np.load("%s/genes.npy" % args.output)
t_matrix = np.load("%s/t_matrix_%s_%.2f.npy" %
(args.output, args.matrix_type, args.rbp_p))
#expr, genes, Xcen = read_matrix()
ncell = Xcen.shape[0]
#sys.stdout.write("Calculate all pairwise Euclidean distance between cells using their physical coordinates\n")
#euc = squareform(pdist(Xcen, metric="euclidean"))
#sys.stdout.write("Rank transform euclidean distance, and then apply exponential transform\n")
#t_matrix = spatial_genes.rank_transform_matrix(euc, reverse=False, rbp_p=rbp_p, matrix_type=matrix_type)
logger.info("Compute silhouette metric per gene")
outdir = "%s/result_sim_5000_%.2f_%.3f" % (args.output, rbp_p, examine_top)
if matrix_type == "dissim":
outdir = "%s/result_5000_%.2f_%.3f" % (args.output, rbp_p, examine_top)
if not os.path.isdir(outdir):
os.mkdir(outdir)
source_path = os.path.dirname(silhouetteRank.__file__)
if not os.path.isfile("%s/do_gpd.R" % outdir):
copyfile("%s/do_gpd.R" % source_path, "%s/do_gpd.R" % outdir)
if not os.path.isfile("%s/do_kmeans.R" % outdir):
copyfile("%s/do_kmeans.R" % source_path, "%s/do_kmeans.R" % outdir)
sizes = read_frequency(expr=expr, genes=genes, Xcen=Xcen, frequency_file=None, \
read_from_file=False, outdir=outdir, examine_top=examine_top, num_query_sizes=args.query_sizes)
res = spatial_genes.random_pattern(matrix=t_matrix,
matrix_type=matrix_type,
num_cell=ncell,
sizes=sizes,
trials_per_gene=5000,
run_gpd=True,
outdir=outdir,
logger=logger)