cpgtools 2.0.5__py3-none-any.whl

cpgtools-2.0.5.data/scripts/beta_topN.py

@@ -0,0 +1,99 @@
+#!python
+
+"""
+#=========================================================================================
+This program picks the top N rows (according to standard deviation) from the input file.
+The resulting file can be used for clustering/PCA analysis.
+
+Example of input data file
+---------------------------
+CpG_ID	Sample_01	Sample_02	Sample_03	Sample_04
+cg_001	0.831035	0.878022	0.794427	0.880911
+cg_002	0.249544	0.209949	0.234294	0.236680
+cg_003	0.845065	0.843957	0.840184	0.824286
+"""
+
+import sys,os
+import collections
+import subprocess
+import numpy as np
+from optparse import OptionParser
+from cpgmodule._version import __version__
+from cpgmodule import ireader
+from cpgmodule.utils import *
+from cpgmodule import BED
+import pandas as pd
+
+__author__ = "Liguo Wang"
+__copyright__ = "Copyleft"
+__credits__ = []
+__license__ = "GPL"
+__maintainer__ = "Liguo Wang"
+__email__ = "wang.liguo@mayo.edu"
+__status__ = "Development"
+
+	
+def main():
+	
+	usage="%prog [options]" + "\n"
+	parser = OptionParser(usage,version="%prog " + __version__)
+	parser.add_option("-i","--input_file",action="store",type="string",dest="input_file",help="Tab-separated data frame file containing beta values with the 1st row containing sample IDs and the 1st column containing CpG IDs.")
+	parser.add_option("-c","--count",action="store",type='int', dest="cpg_count", default=1000, help="Number of most variable CpGs (ranked by standard deviation) to keep. default=%default" )
+	parser.add_option("-s","--score",action="store",type='string', dest="score_type", default='std', help="The type of score used to rank CpGs. Must be one of 'std' or 'mean'. default=%default" )
+	parser.add_option("-o","--output",action="store",type='string', dest="out_file",help="The prefix of the output file.")
+	(options,args)=parser.parse_args()
+	
+	print ()
+	if not (options.input_file):
+		print (__doc__)
+		parser.print_help()
+		sys.exit(101)
+	
+	if not (options.out_file):
+		print (__doc__)
+		parser.print_help()
+		sys.exit(103)	
+	
+	printlog("Reading input file: \"%s\"" % (options.input_file))
+	df1 = pd.read_csv(options.input_file, index_col = 0, sep="\t")
+	
+	#remove any rows with NAs
+	df2 = df1.dropna(axis=0, how='any')
+	printlog("%d rows with missing values were removed." % (len(df1) - len(df2)))
+	
+
+	
+	if options.score_type.lower() == 'std':
+		#calculate stdev for each row
+		row_stds = df2.std(axis=1)
+		df2.loc[:, 'Stdev'] =  row_stds
+
+		#sorted data frame by stdev (decreasingly). Then take the top count,. Then remove Stdev column
+		printlog("Sorting by the standard deviation (decreasingly) ... ")
+		df3 = df2.sort_values(by=['Stdev'], ascending=False)
+		
+		printlog("Data frame with sorted Stdev is saved to file: %s" % options.out_file + '.sortedStdev.tsv')
+		df3.to_csv(options.out_file + '.sortedStdev.tsv', sep = "\t",float_format='%.6f')
+			
+		df4 = df3[0:options.cpg_count].drop('Stdev',axis=1)
+		printlog("Top %d rows of Data frame is saved to file: %s" % (options.cpg_count, options.out_file + '.sortedStdev.topN.tsv'))
+		df4.to_csv(options.out_file + '.sortedStdev.topN.tsv', sep="\t",float_format='%.6f')
+	elif options.score_type.lower() == 'mean':
+		#calculate mean for each row
+		row_means = df2.mean(axis=1)
+		df2['Mean'] =  row_means
+
+		#sorted data frame by mean (decreasingly). Then take the top count,. Then remove Stdev column
+		printlog("Sorting by the mean (decreasingly) ... ")
+		df3 = df2.sort_values(by=['Mean'], ascending=False)
+		
+		printlog("Data frame with sorted Mean is saved to file: %s" % options.out_file + '.sortedMean.tsv')
+		df3.to_csv(options.out_file + '.sortedMean.tsv', sep = "\t",float_format='%.6f')
+			
+		df4 = df3[0:options.cpg_count].drop('Mean',axis=1)
+		printlog("Top %d rows of Data frame is saved to file: %s" % (options.cpg_count, options.out_file + '.sortedMean.topN.tsv'))
+		df4.to_csv(options.out_file + '.sortedMean.topN.tsv', sep="\t",float_format='%.6f')
+	
+
+if __name__=='__main__':
+	main()	

cpgtools-2.0.5.data/scripts/beta_trichotmize.py

@@ -0,0 +1,190 @@
+#!python
+
+"""
+Description
+-----------
+This program uses the Bayesian Gaussian Mixture model (BGMM) to trichotmize beta values into 
+three status: 
+ * Un-methylated (labeled as "0" in result file)
+ * Semi- or particial-methylated (labeled as "1" in result file)
+ * Full-methylated (labeled as "2" in result file)
+ * unassigned (labeled as "-1" in result file)
+"""
+
+
+import sys,os
+import collections
+import numpy as np
+from optparse import OptionParser
+from sklearn import mixture
+from time import strftime
+from cpgmodule._version import __version__
+from cpgmodule import ireader
+from cpgmodule.utils import *
+import pandas as pd
+
+__author__ = "Liguo Wang"
+__copyright__ = "Copyleft"
+__credits__ = []
+__license__ = "GPL"
+__maintainer__ = "Liguo Wang"
+__email__ = "wang.liguo@mayo.edu"
+__status__ = "Development"
+
+
+def load_data(infile):
+	"""
+	Input file is tab or space separated plain text file.
+	*The first row contains sample IDs (must be unique)
+	*The first column contains probe IDs (must be unique)
+	*Each cell (except for the 1st row and 1st column) contains Beta-value
+	
+	Example:
+	
+	Probe			sample_1	sample_2	sample_3 ...
+	cg09835024		0.0547		0.1187		0.0625	...
+	cg25813447		0.428		0.3746		0.0666	...
+	cg07779434		0.3713		0.4194		0.0493	...
+	"""
+
+	printlog("Reading input file: \"%s\"" % infile)
+	df1 = pd.read_csv(infile, index_col = 0, sep="\t")
+
+	#remove any rows with NAs
+	df2 = df1.dropna(axis=0, how='any')
+	printlog("%d rows with missing values were removed." % (len(df1) - len(df2)))
+        
+	print ("\tTotal samples: %d" % (len(df2.columns)), file=sys.stderr)
+	print ("\tTotal probes: %d" % len(df2), file=sys.stderr)
+	return df2
+
+def build_GMM(d,rnd):
+	"""
+	Return means of components of Gaussian Mixture Model.
+	d is data frame returned by "load_data" function.
+	rnd is a random number. You get exactly the same results when running multiple times using the same random number. Must be integer. 
+	"""
+	
+	bgmm_models = collections.defaultdict(list)
+	for s_id in sorted(d.columns):
+		printlog ("Building Bayesian Gaussian Mixture model for subject: %s ...\r" % s_id)
+		bgmm = mixture.BayesianGaussianMixture(n_components=3, covariance_type='full',max_iter=50000,tol=0.001,random_state=rnd)
+		bgmm_models[s_id] = bgmm.fit(d[s_id].values.reshape(-1,1))
+	#print (bgmm_models)
+	return bgmm_models
+	
+
+def summary_GMM(m):
+	"""
+	Summarize BGMM models returned by "build_GMM"
+	"""
+	printlog ("Summerzie GMM models ...")
+	
+	FOUT = open('summary_report.txt','w')
+	
+	print ("\n\n#means of components", file=FOUT)
+	print ("Subject_ID\tUnmethyl\tSemiMethyl\tMethyl",file=FOUT)
+	for k,v in m.items():
+		print (k + '\t' + '\t'.join([str(i) for i in sorted(v.means_[:,0])]),file=FOUT)		
+	
+	
+	print ("\n\n#Weights of components", file=FOUT)
+	print ("Subject_ID\tUnmethyl\tSemiMethyl\tMethyl", file=FOUT)
+	for k,v in m.items():
+		print (k + '\t' + '\t'.join([str(i) for i in sorted(v.weights_)]), file=FOUT)		
+	
+	
+	print ("\n\n#Converge status and n_iter", file=FOUT)
+	print ("Subject_ID\tConverged\tn_iter", file=FOUT)
+	for k,v in m.items():
+		print (k + '\t' + '\t'.join([str(i) for i in (v.converged_, v.n_iter_)]), file=FOUT)		
+	FOUT.close()
+	
+	printlog ("Reports were saved into \"summary_report.txt\".")
+	
+def trichotmize(d,m, prob_cutoff):
+	"""
+	trichotmize beta-value into one of ('0','0.5','1')
+	0 : Un-methylation
+	0.5: Semi- or particial-methylation
+	1: Methylation
+	
+	d is beta value object returned by "load_data" function
+	m is BGMM models returned by 'build_GMM' function
+	
+	"""
+	probe_IDs = list(d.index)
+	
+	for s_id in sorted(m.keys()):
+		printlog ("Writing to \"%s\" ..." % (s_id + ".results.txt"))
+		FOUT = open(s_id + ".results.txt",'w')
+		methyl_lables = {}	#key is index (index can be 0,1 or 2 corresponding to 3 components), value is 0, 1 or 2 corresponding to Un-, Semi- and full- methylation
+		component_means = m[s_id].means_[:,0]	# list of component means
+		betas = d[s_id]
+		for idx,val in enumerate(component_means):
+			if val == max(component_means):
+				methyl_lables[idx] = '2'	# full methyl
+			elif val == min(component_means):
+				methyl_lables[idx] = '0'	# un-methyl
+			else:
+				methyl_lables[idx] = '1'	# semi-methyl
+				
+		probs = m[s_id].predict_proba(d[s_id].values.reshape(-1,1))	# list of probabilities of components: [[  4.33638063e-035   9.54842259e-001   4.51577411e-002],...]
+		
+		print ("#Prob_of_0: Probability of CpG belonging to un-methylation group", file=FOUT)
+		print ("#Prob_of_1: Probability of CpG belonging to semi- or particial-methylation group", file=FOUT)
+		print ("#Prob_of_2: Probability of CpG belonging to full-methylation group", file=FOUT)
+		print ("#Assigned_lable: -1 = 'unassigned', 0 = 'un-methylation', 1 = 'semi- or particial-methylation', 2 = 'full-methylation'", file=FOUT)
+		print ("Probe_ID" + '\tBeta_value\t' + '\t'.join(['Prob_of_' + methyl_lables[0], 'Prob_of_' + methyl_lables[1], 'Prob_of_' + methyl_lables[2]]) + '\t' + 'Assigned_lable', file=FOUT)
+		for probe_ID, beta, p in zip(probe_IDs, betas, probs):
+			p_list = list(p)
+			#print probe_ID
+			#print p_list
+			if methyl_lables[p_list.index(max(p_list))] == '1':
+				if max(p_list) >= prob_cutoff:
+					print (probe_ID + '\t' + str(beta) + '\t' + '\t'.join([str(i) for i in p_list]) + '\t' + methyl_lables[p_list.index(max(p_list))], file=FOUT)
+				else:
+					print (probe_ID + '\t' + str(beta) + '\t' + '\t'.join([str(i) for i in p_list]) + '\t' + '-1', file=FOUT)
+			else:
+				print (probe_ID + '\t' + str(beta) + '\t' + '\t'.join([str(i) for i in p_list]) + '\t' + methyl_lables[p_list.index(max(p_list))], file=FOUT)
+		FOUT.close()
+
+def main():
+	print (__doc__)
+	usage="%prog [options]" + "\n"
+	parser = OptionParser(usage,version="%prog " + __version__)
+	parser.add_option("-i","--input_file",action="store",type="string",dest="input_file",help="Input plain text file containing beta values with the 1st row containing sample IDs (must be unique) and the 1st column containing probe IDs (must be unique).")
+	parser.add_option("-c","--prob-cut",action="store",type="float",dest="prob_cutoff",default=0.95,help="Probability cutoff to assign a probe into \"semi- or particial-methylated\" class. default=%default")
+	parser.add_option("-r","--report",action="store_true",dest="report_summary",default=False, help="If True, generates \"summary_report.txt\" file.  default=%default")
+	parser.add_option("-s","--seed",action="store",type='int', dest="random_state",default=99, help="The seed used by the random number generator. default=%default")
+	(options,args)=parser.parse_args()
+	
+	print ()
+	#print (options.report_summary)
+	#sys.exit()
+	if not (options.input_file):
+		parser.print_help()
+		sys.exit(0)
+	if not os.path.exists(options.input_file):
+		print ('\n\n' + options.input_file + " does NOT exists" + '\n',file=sys.stderr)
+		sys.exit(0)
+	
+	infile = options.input_file
+	
+	#step1: read beta value file
+	dat = load_data(infile)	
+	
+	#step2: build BGMM models
+	GMMs = build_GMM(dat, rnd = options.random_state)
+	
+	#step3: Summerize BGMM models
+	if options.report_summary:
+		summary_GMM(GMMs)
+	
+	#step4: Classification
+	trichotmize(dat, GMMs, options.prob_cutoff)
+	
+
+if __name__=='__main__':
+	main()	
+