Mercurial > repos > kaymccoy > consolidate_fitnesses

comparison consol_fit.py @ 4:7d2f2d1a23ee draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded
author kaymccoy
date Thu, 11 Aug 2016 18:33:54 -0400
parents
children
comparison
equal deleted inserted replaced
3:98ec522f4e95 4:7d2f2d1a23ee
1 # Consol_fit! It's a script & it'll consolidate your fitness values if you got them from a looping trimming pipeline instead of the standard split-by-transposon pipeline. That's it!
2
3 # Test: python ../script/consol_fit.py -calctxt results/py_2_L3_2394eVI_Gluc.txt -wig gview/consol_L3_2394eVI_Gluc.wig -i results/py_L3_2394eVI_Gluc.csv -out results/consol_L3_2394eVI_Gluc.csv -out2 results/py_2_L3_2394eVI_Gluc.csv -normalize tigr4_normal.txt
4
5 # Test: python ../script/consol_fit.py -calctxt results/py_2_L3_2394eVI_Gluc.txt -wig gview/consol_L3_2394eVI_Gluc.wig -i results/galaxy_test.csv -out results/consol_L3_2394eVI_Gluc.csv -out2 results/py_2_L3_2394eVI_Gluc.csv -normalize tigr4_normal.txt
6
7 import math
8 import csv
9
10
11
12
13
14
15
16
17
18
19 ##### ARGUMENTS #####
20
21 def print_usage():
22 print "\n" + "You are missing one or more required flags. A complete list of flags accepted by calc_fitness is as follows:" + "\n\n"
23 print "\033[1m" + "Required" + "\033[0m" + "\n"
24 print "-i" + "\t\t" + "The calc_fit file to be consolidated" + "\n"
25 print "-out" + "\t\t" + "Name of a file to enter the .csv output." + "\n"
26 print "-out2" + "\t\t" + "Name of a file to put the percent blank score in (used in aggregate)." + "\n"
27 print "-calctxt" + "\t\t" + "The txt file output from calc_fit" + "\n"
28 print "-normalize" + "\t" + "A file that contains a list of genes that should have a fitness of 1" + "\n"
29 print "\n"
30 print "\033[1m" + "Optional" + "\033[0m" + "\n"
31 print "-cutoff" + "\t\t" + "Discard any positions where the average of counted transcripts at time 0 and time 1 is below this number (default 0)" + "\n"
32 print "-cutoff2" + "\t\t" + "Discard any positions within the normalization genes where the average of counted transcripts at time 0 and time 1 is below this number (default 0)" + "\n"
33 print "-wig" + "\t\t" + "Create a wiggle file for viewing in a genome browser. Provide a filename." + "\n"
34 print "-maxweight" + "\t" + "The maximum weight a transposon gene can have in normalization calculations" + "\n"
35 print "-multiply" + "\t" + "Multiply all fitness scores by a certain value (e.g., the fitness of a knockout). You should normalize the data." + "\n"
36 print "\n"
37
38 import argparse
39 parser = argparse.ArgumentParser()
40 parser.add_argument("-calctxt", action="store", dest="calctxt")
41 parser.add_argument("-normalize", action="store", dest="normalize")
42 parser.add_argument("-i", action="store", dest="input")
43 parser.add_argument("-out", action="store", dest="outfile")
44 parser.add_argument("-out2", action="store", dest="outfile2")
45 parser.add_argument("-cutoff", action="store", dest="cutoff")
46 parser.add_argument("-cutoff2", action="store", dest="cutoff2")
47 parser.add_argument("-wig", action="store", dest="wig")
48 parser.add_argument("-maxweight", action="store", dest="max_weight")
49 parser.add_argument("-multiply", action="store", dest="multiply")
50 arguments = parser.parse_args()
51
52 # Checks that all the required arguments have actually been entered
53
54 if (not arguments.input or not arguments.outfile or not arguments.calctxt):
55 print_usage()
56 quit()
57
58 #
59
60 if (not arguments.max_weight):
61 arguments.max_weight = 75
62
63 #
64
65 if (not arguments.cutoff):
66 arguments.cutoff = 0
67
68 # Sets the default value of cutoff2 to 10; cutoff2 exists to discard positions within normalization genes with a low number of counted transcripts, because fitnesses calculated from them similarly may not be very accurate.
69 # This only has an effect if it's larger than cutoff, since the normalization step references a list of insertions already affected by cutoff.
70
71 if (not arguments.cutoff2):
72 arguments.cutoff2 = 10
73
74 #Gets total & refname from calc_fit outfile2
75
76 with open(arguments.calctxt) as file:
77 calctxt = file.readlines()
78 total = float(calctxt[1].split()[1])
79 refname = calctxt[2].split()[1]
80
81
82
83
84
85
86
87
88
89
90 ##### CONSOLIDATING THE CALC_FIT FILE #####
91
92 with open(arguments.input) as file:
93 input = file.readlines()
94 results = [["position", "strand", "count_1", "count_2", "ratio", "mt_freq_t1", "mt_freq_t2", "pop_freq_t1", "pop_freq_t2", "gene", "D", "W", "nW"]]
95 i = 1
96 d = float(input[i].split(",")[10])
97 while i < len(input):
98 position = float(input[i].split(",")[0])
99 strands = input[i].split(",")[1]
100 c1 = float(input[i].split(",")[2])
101 c2 = float(input[i].split(",")[3])
102 gene = input[i].split(",")[9]
103 while i + 1 < len(input) and float(input[i+1].split(",")[0]) - position <= 4:
104 if i + 1 < len(input):
105 i += 1
106 c1 += float(input[i].split(",")[2])
107 c2 += float(input[i].split(",")[3])
108 strands = input[i].split(",")[1]
109 if strands[0] == 'b':
110 new_strands = 'b/'
111 elif strands[0] == '+':
112 if input[i].split(",")[1][0] == 'b':
113 new_strands = 'b/'
114 elif input[i].split(",")[1][0] == '+':
115 new_strands = '+/'
116 elif input[i].split(",")[1][0] == '-':
117 new_strands = 'b/'
118 elif strands[0] == '-':
119 if input[i].split(",")[1][0] == 'b':
120 new_strands = 'b/'
121 elif input[i].split(",")[1][0] == '+':
122 new_strands = 'b/'
123 elif input[i].split(",")[1][0] == '-':
124 new_strands = '-/'
125 if len(strands) == 3:
126 if len(input[i].split(",")[1]) < 3:
127 new_strands += strands[2]
128 elif strands[0] == 'b':
129 new_strands += 'b'
130 elif strands[0] == '+':
131 if input[i].split(",")[1][2] == 'b':
132 new_strands += 'b'
133 elif input[i].split(",")[1][2] == '+':
134 new_strands += '+'
135 elif input[i].split(",")[1][2] == '-':
136 new_strands += 'b'
137 elif strands[0] == '-':
138 if input[i].split(",")[1][2] == 'b':
139 new_strands += 'b'
140 elif input[i].split(",")[1][2] == '+':
141 new_strands += 'b'
142 elif input[i].split(",")[1][2] == '-':
143 new_strands += '-'
144 else:
145 if len(input[i].split(",")[1]) == 3:
146 new_strands += input[i].split(",")[1][2]
147 strands = new_strands
148 i +=1
149 if c2 != 0:
150 ratio = c2/c1
151 else:
152 ratio = 0
153 mt_freq_t1 = c1/total
154 mt_freq_t2 = c2/total
155 pop_freq_t1 = 1 - mt_freq_t1
156 pop_freq_t2 = 1 - mt_freq_t2
157 w = 0
158 if mt_freq_t2 != 0:
159 top_w = math.log(mt_freq_t2*(d/mt_freq_t1))
160 bot_w = math.log(pop_freq_t2*(d/pop_freq_t1))
161 w = top_w/bot_w
162 row = [position, strands, c1, c2, ratio, mt_freq_t1, mt_freq_t2, pop_freq_t1, pop_freq_t2, gene, d, w, w]
163 results.append(row)
164 with open(arguments.outfile, "wb") as csvfile:
165 writer = csv.writer(csvfile)
166 writer.writerows(results)
167
168
169
170
171
172
173
174
175
176
177 ##### REDOING NORMALIZATION #####
178
179 # If making a WIG file is requested in the arguments, starts a string to be added to and then written to the WIG file with a typical WIG file header.
180 # The header is just in a typical WIG file format; if you'd like to look into this more UCSC has notes on formatting WIG files on their site.
181
182 if (arguments.wig):
183 wigstring = "track type=wiggle_0 name=" + arguments.wig + "\n" + "variableStep chrom=" + refname + "\n"
184
185 # If a file's given for normalization, starts normalization; this corrects for anything that would cause all the fitness values to be too high or too low.
186
187 if (arguments.normalize):
188
189 # Makes a list of the genes in the normalization file, which should all be transposon genes (these naturally ought to have a fitness value of exactly 1, because transposons are generally non-coding DNA)
190
191 with open(arguments.normalize) as file:
192 transposon_genes = file.read().splitlines()
193 print "Normalize genes loaded" + "\n"
194 blank_ws = 0
195 sum = 0
196 count = 0
197 weights = []
198 scores = []
199 for list in results:
200
201 # Finds all insertions within one of the normalization genes that also have a w value; gets their c1 and c2 values (the number of insertions at t1 and t2) and takes the average of that!
202 # The average is later used as the "weight" of an insertion location's fitness - if it's had more insertions, it should weigh proportionally more towards the average fitness of insertions within the normalization genes.
203
204 if list[9] != '' and list[9] in transposon_genes and list[11]:
205 c1 = list[2]
206 c2 = list[3]
207 score = list[11]
208 avg = (c1 + c2)/2
209
210 # Skips over those insertion locations with too few insertions - their fitness values are less accurate because they're based on such small insertion numbers.
211
212 if float(c1) >= float(arguments.cutoff2):
213
214 # Sets a max weight, to prevent insertion location scores with huge weights from unbalancing the normalization.
215
216 if (avg >= float(arguments.max_weight)):
217 avg = float(arguments.max_weight)
218
219 # Tallies how many w values are 0 within the blank_ws value; you might get many transposon genes with a w value of 0 if a bottleneck occurs, which is especially common with in vivo experiments.
220 # For example, when studying a nasal infection in a mouse model, what bacteria "sticks" and is able to survive and what bacteria is swallowed and killed or otherwise flushed out tends to be a matter
221 # of chance not fitness; all mutants with an insertion in a specific transposon gene could be flushed out by chance!
222
223 if score == 0:
224 blank_ws += 1
225
226 # Adds the fitness values of the insertions within normalization genes together and increments count so their average fitness (sum/count) can be calculated later on
227
228 sum += score
229 count += 1
230
231 # Records the weights of the fitness values of the insertion locations in corresponding lists - for example, weights[2] would be the weight of the fitness value at score[2]
232
233 weights.append(avg)
234 scores.append(score)
235
236 print str(list[9]) + " " + str(score) + " " + str(c1)
237
238 # Counts and removes all "blank" fitness values of normalization genes - those that = 0 - because they most likely don't really have a fitness value of 0, and you just happened to not get any reads from that location at t2.
239
240 blank_count = 0
241 original_count = len(scores)
242 i = 0
243 while i < original_count:
244 w_value = scores[i]
245 if w_value == 0:
246 blank_count += 1
247 weights.pop[i]
248 scores.pop[i]
249 i-=1
250 i += 1
251
252 # If no normalization genes can pass the cutoff, normalization cannot occur, so this ends the script advises the user to try again and lower cutoff and/or cutoff2.
253
254 if len(scores) == 0:
255 print 'ERROR: The normalization genes do not have enough reads to pass cutoff and/or cutoff2; please lower one or both of those arguments.' + "\n"
256 quit()
257
258 # Prints the number of of blank fitness values found and removed for reference. Writes the percentage to a file so it can be referenced for aggregate analysis.
259
260 pc_blank_normals = float(blank_count) / float(original_count)
261 print "# blank out of " + str(original_count) + ": " + str(pc_blank_normals) + "\n"
262 with open(arguments.outfile2, "w") as f:
263 f.write("blanks: " + str(pc_blank_normals) + "\n" + "total: " + str(total) + "\n" + "refname: " + refname)
264
265
266 # Finds "average" - the average fitness value for an insertion within the transposon genes - and "weighted_average" - the average fitness value for an insertion within the transposon genes weighted by how many insertions each had.
267
268 average = sum / count
269 i = 0
270 weighted_sum = 0
271 weight_sum = 0
272 while i < len(weights):
273 weighted_sum += weights[i]*scores[i]
274 weight_sum += weights[i]
275 i += 1
276 weighted_average = weighted_sum/weight_sum
277
278 # Prints the regular average, weighted average, and total insertions for reference
279
280 print "Normalization step:" + "\n"
281 print "Regular average: " + str(average) + "\n"
282 print "Weighted Average: " + str(weighted_average) + "\n"
283 print "Total Insertions: " + str(count) + "\n"
284
285 # The actual normalization happens here; every fitness score is divided by the average fitness found for genes that should have a value of 1.
286 # For example, if the average fitness for genes was too low overall - let's say 0.97 within the normalization geness - every fitness would be proportionally raised.
287
288 old_ws = 0
289 new_ws = 0
290 wcount = 0
291 for list in results:
292 if list[11] == 'W':
293 continue
294 new_w = float(list[11])/weighted_average
295
296 # Sometimes you want to multiply all the fitness values by a constant; this does that.
297 # For example you might multiply all the values by a constant for a genetic interaction screen - where Tn-Seq is performed as usual except there's one background knockout all the mutants share. This is
298 # because independent mutations should have a fitness value that's equal to their individual fitness values multipled, but related mutations will deviate from that; to find those deviations you'd multiply
299 # all the fitness values from mutants from a normal library by the fitness of the background knockout and compare that to the fitness values found from the knockout library!
300
301 if arguments.multiply:
302 new_w *= float(arguments.multiply)
303
304 # Records the old w score for reference, and adds it to a total sum of all w scores (so that the old w mean and new w mean can be printed later).
305
306 if float(list[11]) > 0:
307 old_ws += float(list[11])
308 new_ws += new_w
309 wcount += 1
310
311 # Writes the new w score into the results list of lists.
312
313 list[12] = new_w
314
315 # Adds a line to wiglist for each insertion position, with the insertion position and its new w value.
316
317 if (arguments.wig):
318 wigstring += str(list[0]) + " " + str(new_w) + "\n"
319
320 # Prints the old w mean and new w mean for reference.
321
322 old_w_mean = old_ws / wcount
323 new_w_mean = new_ws / wcount
324 print "Old W Average: " + str(old_w_mean) + "\n"
325 print "New W Average: " + str(new_w_mean) + "\n"
326
327 # Overwrites the old file with the normalized file.
328
329 with open(arguments.outfile, "wb") as csvfile:
330 writer = csv.writer(csvfile)
331 writer.writerows(results)
332
333 # If a WIG file was requested, actually creates the WIG file and writes wiglist to it
334 # So what's written here is the WIG header plus each insertion position and it's new w value if normalization were called for, and each insertion position and its unnormalized w value if normalization were not called for.
335
336 if (arguments.wig):
337 if (arguments.normalize):
338 with open(arguments.wig, "wb") as wigfile:
339 wigfile.write(wigstring)
340 else:
341 for list in results:
342 wigstring += str(list[0]) + " " + str(list[11]) + "\n"
343 with open(arguments.wig, "wb") as wigfile:
344 wigfile.write(wigstring)