comparison aggregate.py @ 4:5ff57a3d0af2 draft

Uploaded
author kaymccoy
date Sun, 11 Dec 2016 17:01:25 -0500
parents
children
comparison
equal deleted inserted replaced
3:359b7cc9ad94 4:5ff57a3d0af2
1 # A translation of aggregate.pl into python! For analysis of Tn-Seq.
2 # This script requires BioPython just like calc_fitness.py, so you need it installed along with its dependencies if you want to run these scripts on your own.
3 # How to install BioPython and a list of its dependencies can be found here: http://biopython.org/DIST/docs/install/Installation.html
4 # K. McCoy
5
6
7
8
9
10
11
12
13
14 ##### ARGUMENTS #####
15
16 def print_usage():
17 print "Aggregate.py's usage is as follows:" + "\n\n"
18 print "\033[1m" + "Required" + "\033[0m" + "\n"
19 print "-o" + "\t\t" + "Output file for aggregated data." + "\n"
20 print "\n"
21 print "\033[1m" + "Optional" + "\033[0m" + "\n"
22 print "-c" + "\t\t" + "Check for missing genes in the data set - provide a reference genome in genbank format. Missing genes will be sent to stdout." + "\n"
23 print "-m" + "\t\t" + "Place a mark in an extra column for this set of genes. Provide a file with a list of genes seperated by newlines." + "\n"
24 print "-x" + "\t\t" + "Cutoff: Don't include fitness scores with average counts (c1+c2)/2 < x (default: 0)" + "\n"
25 print "-b" + "\t\t" + "Bottleneck value: The percentage of insertions randomly lost, which will be discounted for all genes (for example, 20% would be entered as 0.20; default 0.0)" + "\n"
26 print "-f" + "\t\t" + "An in-between file carrying information on the blank count found from calc_fitness or consol_fitness; one of two ways to pass a blank count to this script" + "\n"
27 print "-w" + "\t\t" + "Use weighted algorithm to calculate averages, variance, sd, se" + "\n"
28 print "-l" + "\t\t" + "Weight ceiling: maximum value to use as a weight (default: 999,999)" + "\n"
29 print "\n"
30 print "All remainder arguements will be treated as fitness files (those files created by calc_fitness.py)" + "\n"
31 print "\n"
32
33 import argparse
34 parser = argparse.ArgumentParser()
35 parser.add_argument("-o", action="store", dest="summary")
36 parser.add_argument("-c", action="store", dest="find_missing")
37 parser.add_argument("-m", action="store", dest="marked")
38 parser.add_argument("-x", action="store", dest="cutoff")
39 parser.add_argument("-b", action="store", dest="blank_pc")
40 parser.add_argument("-f", action="store", dest="blank_file")
41 parser.add_argument("-w", action="store", dest="weighted")
42 parser.add_argument("-l", action="store", dest="weight_ceiling")
43 parser.add_argument("fitnessfiles", nargs=argparse.REMAINDER)
44
45 arguments = parser.parse_args()
46
47 if not arguments.summary:
48 print "\n" + "You are missing a value for the -o flag. "
49 print_usage()
50 quit()
51
52 if not arguments.fitnessfiles:
53 print "\n" + "You are missing fitness file(s); these should be entered immediately after all the flags. "
54 print_usage()
55 quit()
56
57 # 999,999 is a trivial placeholder number
58
59 if (not arguments.weight_ceiling):
60 arguments.weight_ceiling = 999999
61
62 # Cutoff exists to discard positions with a low number of counted transcripts, because their fitness may not be as accurate - for the same reasoning that studies with low sample sizes can be innacurate.
63
64 if (not arguments.cutoff):
65 arguments.cutoff = 0
66
67 # Gets information from the txt output file of calc_fit / consol, if inputted
68
69 if arguments.blank_file:
70 with open(arguments.blank_file) as file:
71 blank_pc = file.read().splitlines()
72 arguments.blank_pc = float(blank_pc[0].split()[1])
73
74 if (not arguments.blank_pc):
75 arguments.blank_pc = 0
76
77
78
79
80
81 ##### SUBROUTINES #####
82
83 # A subroutine that calculates the average, variance, standard deviation (sd), and standard error (se) of a group of scores; for use when aggregating scores by gene later on
84
85 import math
86 def unweighted_average(scores):
87 sum = 0
88 num = 0
89 i = 0
90 while i < len(scores):
91 if not scores[i]:
92 scores[i] = 0.0
93 sum += float(scores[i])
94 num += 1
95 i += 1
96 average = sum/num
97 xminusxbars = 0
98 while i < len(scores):
99 xminusxbars += (float(scores[i]) - average)**2
100 if num <= 1:
101 variance = 0
102 else:
103 variance = xminusxbars/(num-1)
104 sd = math.sqrt(variance)
105 se = sd / math.sqrt(num)
106 return (average, variance, sd, se)
107
108 # A subroutine that calculates the weighted average, variance, standard deviation (sd), and standard error (se) of a group of scores; the weights come from the number of reads each insertion location has
109 # For use when aggregating scores by gene later on, if the weighted argument is called
110
111 def weighted_average(scores,weights):
112 sum = 0
113 weighted_average = 0
114 weighted_variance = 0
115 top = 0
116 bottom = 0
117 i = 0
118 while i < len(weights):
119 if not scores[i]:
120 scores[i] = 0.0
121 top += float(weights[i])*float(scores[i])
122 bottom += float(weights[i])
123 i += 1
124 if bottom == 0:
125 return 0
126 weighted_average = top/bottom
127 top = 0
128 bottom = 0
129 i = 0
130 while i < len(weights):
131 top += float(weights[i]) * (float(scores[i]) - weighted_average)**2
132 bottom += float(weights[i])
133 i += 1
134 weighted_variance = top/bottom
135 weighted_stdev = math.sqrt(weighted_variance)
136 weighted_stder = weighted_stdev/math.sqrt(len(scores))
137 return (weighted_average, weighted_variance, weighted_stdev, weighted_stder)
138
139
140
141
142
143
144
145
146
147
148 ##### AGGREGATION / CALCULATIONS #####
149
150 #Reads the genes which should be marked in the final aggregate file into an array
151
152 import os.path
153 if arguments.marked:
154 with open(arguments.marked) as file:
155 marked_set = file.read().splitlines()
156
157 #Creates a dictionary of dictionaries to contain a summary of all genes and their fitness values
158 #The fitness values and weights match up, so that the weight of gene_summary[locus]["w"][2] would be gene_summary[locus]["s"][2]
159
160 import csv
161 gene_summary = {}
162 for eachfile in arguments.fitnessfiles:
163 with open(eachfile) as csvfile:
164 lines = csv.reader(csvfile)
165 for line in lines:
166 locus = line[9]
167 w = line[12]
168 if w == 'nW':
169 continue
170 if not w:
171 w == 0
172 c1 = float(line[2])
173 c2 = float(line[3])
174 avg = (c1+c2)/2
175 if avg < float(arguments.cutoff):
176 continue
177 if avg > float(arguments.weight_ceiling):
178 avg = arguments.weight_ceiling
179 if locus not in gene_summary:
180 gene_summary[locus] = {"w" : [], "s": []}
181 gene_summary[locus]["w"].append(w)
182 gene_summary[locus]["s"].append(avg)
183
184 #If finding any missing gene loci is requested in the arguments, starts out by loading all the known features from a genbank file
185
186 from Bio import SeqIO
187 if (arguments.find_missing):
188 output = [["locus","mean","var","sd","se","gene","Total","Blank","Not Blank","Blank Removed","M\n"]]
189 handle = open(arguments.find_missing, "rU")
190 for record in SeqIO.parse(handle, "genbank"):
191 refname = record.id
192 features = record.features
193 handle.close()
194
195 #Goes through the features to find which are genes
196
197 for feature in features:
198 gene = ""
199 if feature.type == "gene":
200 locus = "".join(feature.qualifiers["locus_tag"])
201 if "gene" in feature.qualifiers:
202 gene = "".join(feature.qualifiers["gene"])
203 else:
204 continue
205
206 #Goes through the fitness scores of insertions within each gene, and removes whatever % of blank fitness scores were requested along with their corresponding weights
207
208 sum = 0
209 num = 0
210 avgsum = 0
211 blank_ws = 0
212 i = 0
213 if locus in gene_summary.keys():
214 for w in gene_summary[locus]["w"]:
215 if float(w) == 0:
216 blank_ws += 1
217 else:
218 sum += float(w)
219 num += 1
220 count = num + blank_ws
221 removed = 0
222 to_remove = int(float(arguments.blank_pc)*count)
223 if blank_ws > 0:
224 i = 0
225 while i < len(gene_summary[locus]["w"]):
226 w = gene_summary[locus]["w"][i]
227 if removed == to_remove:
228 break
229 if float(w) == 0:
230 del gene_summary[locus]["w"][i]
231 del gene_summary[locus]["s"][i]
232 removed += 1
233 i -= 1
234 i += 1
235
236 #If all the fitness values within a gene are empty, sets mean/var to 0.10 and Xs out sd/se; marks the gene if that's requested
237
238 if num == 0:
239 if (arguments.marked and locus in marked_set):
240 output.append([locus, "0.10", "0.10", "X", "X", gene, count, blank_ws, num, removed, "M", "\n"])
241 else:
242 output.append([locus, "0.10", "0.10", "X", "X", gene, count, blank_ws, num, removed, "\n"])
243
244 #Otherwise calls average() or weighted_average() to find the aggregate w / count / standard deviation / standard error of the insertions within each gene; marks the gene if that's requested
245
246 else:
247 if not arguments.weighted:
248 (average, variance, stdev, stderr) = unweighted_average(gene_summary[locus]["w"])
249 else:
250 (average, variance, stdev, stderr) = weighted_average(gene_summary[locus]["w"],gene_summary[locus]["s"])
251 if (arguments.marked and locus in marked_set):
252 output.append([locus, average, variance, stdev, stderr, gene, count, blank_ws, num, removed, "M", "\n"])
253 else:
254 output.append([locus, average, variance, stdev, stderr, gene, count, blank_ws, num, removed, "\n"])
255
256 #If a gene doesn't have any insertions, sets mean/var to 0.10 and Xs out sd/se, plus leaves count through removed blank because there were no reads.
257
258 else:
259 if (arguments.marked and locus in marked_set):
260 output.append([locus, "0.10", "0.10", "X", "X", gene, "", "", "", "", "M", "\n"])
261 else:
262 output.append([locus, "0.10", "0.10", "X", "X", gene, "", "", "", "", "\n"])
263
264 #Writes the aggregated fitness file
265
266 with open(arguments.summary, "wb") as csvfile:
267 writer = csv.writer(csvfile)
268 writer.writerows(output)
269
270 #If finding missing genes is not requested, just finds the aggregate w / count / standard deviation / standard error of the insertions within each gene, and writes them to a file, plus marks the genes requested
271 #This is never called through Galaxy since finding missing genes is just better than not finding them.
272
273 else:
274 output = [["Locus","W","Count","SD","SE","M\n"]]
275 for gene in gene_summary.keys():
276 sum = 0
277 num = 0
278 average = 0
279 if "w" not in gene_summary[gene]:
280 continue
281 for i in gene_summary[gene]["w"]:
282 sum += i
283 num += 1
284 average = sum/num
285 xminusxbars = 0
286 for i in w:
287 xminusxbars += (i-average)**2
288 if num > 1:
289 sd = math.sqrt(xminusxbars/(num-1))
290 se = sd / math.sqrt(num)
291 if (arguments.marked and locus in marked_set):
292 output.append([gene, average, num, sd, se, "M", "\n"])
293 else:
294 output.append([gene, average, num, sd, se, "\n"])
295 with open(arguments.summary, "wb") as csvfile:
296 writer = csv.writer(csvfile)
297 writer.writerows(output)