comparison mirgene_graphs.py @ 6:d58f050acd18 draft

Uploaded
author glogobyte
date Wed, 13 Oct 2021 16:05:08 +0000
parents
children 0cdc0c3c821e
comparison
equal deleted inserted replaced
5:810e789ffeab 6:d58f050acd18
1 import itertools
2 import pandas as pd
3 from math import pi
4 import numpy as np
5 import matplotlib.pyplot as plt
6 import math
7 import logomaker as lm
8 from fpdf import FPDF, fpdf
9 import glob
10
11 ###############################################################################################################################################################################
12 def pie_non_temp(merge_con,merge_non_con,merge_tre,merge_non_tre,c_unmap,t_unmap,c_unmap_counts,t_unmap_counts,group_name1,group_name2):
13
14 c_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_con]
15 t_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_tre]
16 c_non_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_non_con]
17 t_non_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_non_tre]
18
19 c_templ = 0
20 c_tem_counts = 0
21 c_mature = 0
22 c_mat_counts = 0
23 t_templ = 0
24 t_tem_counts = 0
25 t_mature = 0
26 t_mat_counts = 0
27
28 c_non = len(c_non_samples)
29 c_non_counts = sum(x[2] for x in c_non_samples)
30 t_non = len(t_non_samples)
31 t_non_counts = sum(x[2] for x in t_non_samples)
32
33 c_unmap = c_unmap - c_non
34 t_unmap = c_unmap - t_non
35
36 c_unmap_counts=c_unmap_counts - c_non_counts
37 t_unmap_counts=t_unmap_counts - t_non_counts
38
39
40 for x in c_samples:
41
42 if "/" not in x[0]:
43 if len(x[0].split("_"))==2:
44 c_mature+=1
45 c_mat_counts += x[2]
46 else:
47 c_templ+=1
48 c_tem_counts += x[2]
49 else:
50 f=0
51 for y in x[0].split("/"):
52 if len(y.split("_"))==2:
53 c_mature+=1
54 c_mat_counts += x[2]
55 f=1
56 break
57 if f==0:
58 c_templ+=1
59 c_tem_counts += x[2]
60
61 for x in t_samples:
62
63 if "/" not in x[0]:
64 if len(x[0].split("_"))==2:
65 t_mature+=1
66 t_mat_counts += x[2]
67 else:
68 t_templ+=1
69 t_tem_counts += x[2]
70 else:
71 f=0
72 for y in x[0].split("/"):
73 if len(y.split("_"))==2:
74 t_mature+=1
75 t_mat_counts += x[2]
76 f=1
77 break
78 if f==0:
79 t_templ+=1
80 t_tem_counts += x[2]
81
82 fig = plt.figure(figsize=(7,5))
83
84 labels = 'miRNA RefSeq','Template', 'Unassigned' ,'Non-template'
85 sizes = [c_mat_counts, c_tem_counts, c_unmap_counts,c_non_counts]
86 colors = ['gold', 'yellowgreen', 'lightcoral', 'lightskyblue']
87 # Plot
88 ax1 = plt.subplot2grid((1,2),(0,0))
89 patches, texts, autotexts=plt.pie(sizes, labels=labels, colors=colors, startangle=140,autopct='%1.1f%%',radius=0.8)
90 [x.set_fontsize(8) for x in texts]
91 plt.title(group_name1 + ' Group (reads)',fontsize=12)
92
93 labels = 'miRNA RefSeq','Template', 'Unassigned','non-template'
94 sizes = [t_mat_counts, t_tem_counts, t_unmap_counts, t_non_counts]
95 colors = ['gold', 'yellowgreen', 'lightcoral', 'lightskyblue']
96 # Plot
97 ax2 = plt.subplot2grid((1,2),(0,1))
98 patches, texts, autotexts=plt.pie(sizes, labels=labels, colors=colors, startangle=140,autopct='%1.1f%%',radius=0.8)
99 [x.set_fontsize(8) for x in texts]
100 plt.title(group_name2 + ' Group (reads)', fontsize=12)
101 plt.savefig('pie_non.png',dpi=300)
102
103 ####################################################################################################################################################################################################################
104
105 def pie_temp(merge_con,c_unmap,c_unmap_counts,merge_tre,t_unmap,t_unmap_counts,group_name1,group_name2):
106
107 c_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_con]
108 t_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_tre]
109
110 c_templ = 0
111 c_tem_counts = 0
112 c_mature = 0
113 c_mat_counts = 0
114 t_templ = 0
115 t_tem_counts = 0
116 t_mature = 0
117 t_mat_counts = 0
118
119 for x in c_samples:
120
121 if "/" not in x[0]:
122 if len(x[0].split("_"))==2:
123 c_mature+=1
124 c_mat_counts += x[2]
125 else:
126 c_templ+=1
127 c_tem_counts += x[2]
128 else:
129 f=0
130 for y in x[0].split("/"):
131 if len(y.split("_"))==2:
132 c_mature+=1
133 c_mat_counts += x[2]
134 f=1
135 break
136 if f==0:
137 c_templ+=1
138 c_tem_counts += x[2]
139
140 for x in t_samples:
141
142 if "/" not in x[0]:
143 if len(x[0].split("_"))==2:
144 t_mature+=1
145 t_mat_counts += x[2]
146 else:
147 t_templ+=1
148 t_tem_counts += x[2]
149 else:
150 f=0
151 for y in x[0].split("/"):
152 if len(y.split("_"))==2:
153 t_mature+=1
154 t_mat_counts += x[2]
155 f=1
156 break
157 if f==0:
158 t_templ+=1
159 t_tem_counts += x[2]
160
161
162 fig = plt.figure()
163 labels = 'miRNA RefSeq','Template', 'Unassigned'
164 sizes = [c_mat_counts, c_tem_counts, c_unmap_counts]
165 colors = ['gold', 'yellowgreen', 'lightskyblue']
166 explode = (0.2, 0.05, 0.1)
167 # Plot
168 ax1 = plt.subplot2grid((1,2),(0,0))
169 patches, texts, autotexts=plt.pie(sizes, labels=labels, colors=colors, startangle=140,autopct='%1.1f%%',radius=0.8)
170 [x.set_fontsize(8) for x in texts]
171 plt.title(group_name1 + ' group (reads)', fontsize=12)
172 labels = 'miRNA RefSeq','Template', 'Unassigned'
173 sizes = [t_mat_counts, t_tem_counts, t_unmap_counts]
174 colors = ['gold', 'yellowgreen', 'lightskyblue']
175 explode = (0.2, 0.05, 0.1)
176 # Plot
177 ax2 = plt.subplot2grid((1,2),(0,1))
178 patches, texts, autotexts=plt.pie(sizes, labels=labels, colors=colors, startangle=140,autopct='%1.1f%%',radius=0.8)
179 [x.set_fontsize(8) for x in texts]
180 plt.title(group_name2 + ' group (reads)',fontsize = 12)
181 plt.savefig('pie_tem.png',dpi=300)
182
183 ###################################################################################################################################################################################################################
184
185 def make_spider(merge_con,merge_tre,group_name1,group_name2):
186
187 c_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_con]
188 t_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge_tre]
189
190 c_5 = 0
191 c_5_counts = 0
192 c_3 = 0
193 c_3_counts = 0
194 c_both =0
195 c_both_counts=0
196 c_mature = 0
197 c_mat_counts = 0
198 c_exception=0
199 c_exception_counts=0
200
201
202 t_5 = 0
203 t_5_counts = 0
204 t_3 = 0
205 t_3_counts = 0
206 t_both = 0
207 t_both_counts = 0
208 t_mature = 0
209 t_mat_counts = 0
210 t_exception = 0
211 t_exception_counts=0
212
213 for x in c_samples:
214
215 if "/" not in x[0]:
216 if len(x[0].split("_"))==2:
217 c_mature+=1
218 c_mat_counts += x[2]
219 elif 0 == int(x[0].split("_")[3]):
220 c_5+=1
221 c_5_counts += x[2]
222 elif 0 == int(x[0].split("_")[4]):
223 c_3+=1
224 c_3_counts += x[2]
225 else:
226 c_both+=1
227 c_both_counts+=x[2]
228
229 else:
230 f=0
231 for y in x[0].split("/"):
232 if len(y.split("_"))==2:
233 c_mature+=1
234 c_mat_counts += x[2]
235 f=1
236 break
237 if f==0:
238 part=x[0].split("/")[0].split("_")[-2]+"_"+x[0].split("/")[0].split("_")[-1]
239 num=0
240 for y in x[0].split("/"):
241 if y.split("_")[-2]+"_"+y.split("_")[-1]==part:
242 num+=1
243 if num==len(x[0].split("/")):
244
245 if 0 == int(x[0].split("/")[0].split("_")[3]):
246 c_5+=1
247 c_5_counts += x[2]
248 elif 0 == int(x[0].split("/")[0].split("_")[4]):
249 c_3+=1
250 c_3_counts += x[2]
251 else:
252 c_both+=1
253 c_both_counts+=x[2]
254 else:
255 c_exception+=1
256 c_exception_counts += x[2]
257
258
259 for x in t_samples:
260
261 if "/" not in x[0]:
262 if len(x[0].split("_"))==2:
263 t_mature+=1
264 t_mat_counts += x[2]
265 elif 0 == int(x[0].split("_")[3]):
266 t_5+=1
267 t_5_counts += x[2]
268 elif 0 == int(x[0].split("_")[4]):
269 t_3+=1
270 t_3_counts += x[2]
271 else:
272 t_both+=1
273 t_both_counts+=x[2]
274
275 else:
276 f=0
277 for y in x[0].split("/"):
278 if len(y.split("_"))==2:
279 t_mature+=1
280 t_mat_counts += x[2]
281 f=1
282 break
283 if f==0:
284 part=x[0].split("/")[0].split("_")[-2]+"_"+x[0].split("/")[0].split("_")[-1]
285 num=0
286 for y in x[0].split("/"):
287 if y.split("_")[-2]+"_"+y.split("_")[-1]==part:
288 num+=1
289 if num==len(x[0].split("/")):
290
291 if 0 == int(x[0].split("/")[0].split("_")[3]):
292 t_5+=1
293 t_5_counts += x[2]
294 elif 0 == int(x[0].split("/")[0].split("_")[4]):
295 t_3+=1
296 t_3_counts += x[2]
297 else:
298 t_both+=1
299 t_both_counts+=x[2]
300 else:
301 t_exception+=1
302 t_exception_counts += x[2]
303
304
305 c_all = c_5+c_3+c_both+c_mature+c_exception
306 c_all_counts = c_5_counts + c_3_counts + c_both_counts + c_mat_counts + c_exception_counts
307
308 t_all = t_5+t_3+t_both+t_mature + t_exception
309 t_all_counts = t_5_counts + t_3_counts + t_both_counts + t_mat_counts + t_exception_counts
310
311 c_5 = round(c_5/c_all*100,2)
312 c_3 = round(c_3/c_all*100,2)
313 c_both = round(c_both/c_all*100,2)
314 c_mature = round(c_mature/c_all*100,2)
315 c_exception = round(c_exception/c_all*100,2)
316
317 c_5_counts = round(c_5_counts/c_all_counts*100,2)
318 c_3_counts = round(c_3_counts/c_all_counts*100,2)
319 c_both_counts = round(c_both_counts/c_all_counts*100,2)
320 c_mat_counts = round(c_mat_counts/c_all_counts*100,2)
321 c_exception_counts = round(c_exception_counts/c_all_counts*100,2)
322
323 t_5 = round(t_5/t_all*100,2)
324 t_3 = round(t_3/t_all*100,2)
325 t_both = round(t_both/t_all*100,2)
326 t_mature = round(t_mature/t_all*100,2)
327 t_exception = round(t_exception/t_all*100,2)
328
329 t_5_counts = round(t_5_counts/t_all_counts*100,2)
330 t_3_counts = round(t_3_counts/t_all_counts*100,2)
331 t_both_counts = round(t_both_counts/t_all_counts*100,2)
332 t_mat_counts = round(t_mat_counts/t_all_counts*100,2)
333 t_exception_counts = round(t_exception_counts/t_all_counts*100,2)
334
335 radar_max = max(c_5, c_3, c_both,c_mature,c_exception,t_5,t_3,t_both,t_mature,t_exception)
336 radar_max_counts = max(c_5_counts,c_3_counts,c_both_counts,c_mat_counts,c_exception_counts,t_5_counts,t_3_counts,t_both_counts,t_mat_counts,t_exception_counts)
337
338 df=pd.DataFrame({
339 'group':[group_name1,group_name2],
340 """5'3'-isomiRs""":[c_both,t_both],
341 """3'-isomiRs""":[c_3,t_3],
342 'RefSeq miRNA':[c_mature,t_mature],
343 """5'-isomiRs""":[c_5,t_5],
344 'Others*':[c_exception,t_exception]})
345
346 df1=pd.DataFrame({
347 'group':[group_name1,group_name2],
348 """5'-3'-isomiRs""":[c_both_counts,t_both_counts],
349 """3'-isomiRs""":[c_3_counts,t_3_counts],
350 'RefSeq miRNA':[c_mat_counts,t_mat_counts],
351 """5'-isomiRs""":[c_5_counts,t_5_counts],
352 'Others*':[c_exception_counts,t_exception_counts]})
353
354 spider_last(df,radar_max,1,group_name1,group_name2)
355 spider_last(df1,radar_max_counts,2,group_name1,group_name2)
356
357
358 ##############################################################################################################################################################################################################
359
360 def spider_last(df,radar_max,flag,group_name1,group_name2):
361 # ------- PART 1: Create background
362 fig = plt.figure()
363 # number of variable
364 categories=list(df)[1:]
365 N = len(categories)
366
367 # What will be the angle of each axis in the plot? (we divide the plot / number of variable)
368 angles = [n / float(N) * 2 * pi for n in range(N)]
369 angles += angles[:1]
370
371 # Initialise the spider plot
372 ax = plt.subplot(111, polar=True)
373
374 # If you want the first axis to be on top:
375 ax.set_theta_offset(pi/2)
376 ax.set_theta_direction(-1)
377
378 # Draw one axe per variable + add labels labels yet
379 plt.xticks(angles[:-1], categories, fontsize=11)
380
381 # Draw ylabels
382 radar_max=round(radar_max+radar_max*0.1)
383 mul=len(str(radar_max))-1
384 maxi=int(math.ceil(radar_max / pow(10,mul))) * pow(10,mul)
385 sep = round(maxi/4)
386 plt.yticks([sep, 2*sep, 3*sep, 4*sep, 5*sep], [str(sep)+'%', str(2*sep)+'%', str(3*sep)+'%', str(4*sep)+'%', str(5*sep)+'%'], color="grey", size=10)
387 plt.ylim(0, maxi)
388
389 # ------- PART 2: Add plots
390
391 # Plot each individual = each line of the data
392 # I don't do a loop, because plotting more than 3 groups makes the chart unreadable
393
394 # Ind1
395 values=df.loc[0].drop('group').values.flatten().tolist()
396 values += values[:1]
397 ax.plot(angles, values,'-o', linewidth=1, linestyle='solid', label=group_name1)
398 ax.fill(angles, values, 'b', alpha=0.1)
399
400 # Ind2
401 values=df.loc[1].drop('group').values.flatten().tolist()
402 values += values[:1]
403 ax.plot(angles, values, '-o' ,linewidth=1, linestyle='solid', label=group_name2)
404 ax.fill(angles, values, 'r', alpha=0.1)
405
406 # Add legend
407 if flag==1:
408 plt.legend(loc='upper right', bbox_to_anchor=(0.0, 0.1))
409 plt.savefig('spider_non_red.png',dpi=300)
410 else:
411 plt.legend(loc='upper right', bbox_to_anchor=(0.0, 0.1))
412 plt.savefig('spider_red.png',dpi=300)
413
414 #############################################################################################################################################################################################################
415
416 def hist_red(samples,flag,group):
417
418 lengths=[]
419 cat=[]
420 total_reads=0
421 seq=[]
422
423 if flag == "c":
424 title = "Length Distribution of "+group+" group (Redudant reads)"
425 if flag == "t":
426 title = "Length Distribution of "+group+" group (Redudant reads)"
427
428 for i in samples:
429 for x in i:
430 lengths.append(x[3])
431 if x[1]=="0":
432 seq.append([x[3],x[0].split("-")[1],"Mapped"])
433 cat.append("Mapped")
434 if x[1] == "4":
435 seq.append([x[3],x[0].split("-")[1],"Unassigned"])
436 cat.append("Unassigned")
437
438 uni_len=list(set(lengths))
439 uni_len=[x for x in uni_len if x<=35]
440 low=min(lengths)
441 up=max(lengths)
442 seq.sort()
443 uni_seq=list(seq for seq,_ in itertools.groupby(seq))
444 dim=up-low
445
446 if dim>20:
447 s=5
448 else:
449 s=8
450
451 total_reads+=sum([int(x[1]) for x in uni_seq])
452
453 map_reads=[]
454 unmap_reads=[]
455 length=[]
456 for y in uni_len:
457 map_temp=0
458 unmap_temp=0
459 for x in uni_seq:
460 if x[0]==y and x[2]=="Mapped":
461 map_temp+=int(x[1])
462 if x[0]==y and x[2]=="Unassigned":
463 unmap_temp+=int(x[1])
464 if y<=35:
465 length.append(y)
466 map_reads.append(round(map_temp/total_reads*100,2))
467 unmap_reads.append(round(unmap_temp/total_reads*100,2))
468
469 ylim=max([sum(x) for x in zip(unmap_reads, map_reads)])
470 ylim=ylim+ylim*20/100
471 fig, ax = plt.subplots()
472 width=0.8
473 ax.bar(length, unmap_reads, width, label='Unassigned')
474 h=ax.bar(length, map_reads, width, bottom=unmap_reads, label='Mapped')
475 plt.xticks(np.arange(length[0], length[-1]+1, 1))
476 plt.yticks(np.arange(0, ylim, 5))
477 plt.xlabel('Length (nt)',fontsize=14)
478 plt.ylabel('Percentage',fontsize=14)
479 plt.title(title,fontsize=14)
480 ax.legend()
481 plt.ylim([0, ylim])
482 ax.grid(axis='y',linewidth=0.2)
483
484 if flag=='c':
485 plt.savefig('c_hist_red.png',dpi=300)
486
487 if flag=='t':
488 plt.savefig('t_hist_red.png',dpi=300)
489
490 ##############################################################################################################################################################################################################################################
491
492 def logo_seq_red(merge, flag, group):
493
494 if flag=="c":
495 titlos = group + " group (Redundant)"
496 file_logo = "c_logo.png"
497 file_bar = "c_bar.png"
498 if flag=="t":
499 titlos = group + " group (Redundant)"
500 file_logo = "t_logo.png"
501 file_bar = "t_bar.png"
502
503 c_samples=[[x[0],x[1],sum(int(i) for i in x[2:])] for x in merge]
504
505 A=[0]*3
506 C=[0]*3
507 G=[0]*3
508 T=[0]*3
509 total_reads=0
510
511 for y in c_samples:
512 if "/" in y[0]:
513 length=[]
514 for x in y[0].split("/"):
515 length.append([len(x.split("_")[-1]),x.split("_")[-1],y[2]])
516
517 best=min(length)
518 total_reads+=best[2]
519 for i in range(3):
520 if i<len(best[1]):
521 if best[1][i] == "A":
522 A[i]+=best[2]
523 elif best[1][i] == "C":
524 C[i]+=best[2]
525 elif best[1][i] == "G":
526 G[i]+=best[2]
527 else:
528 T[i]+=best[2]
529 else:
530 total_reads+=y[2]
531 for i in range(3):
532 if i<len(y[0].split("_")[-1]):
533 if y[0].split("_")[-1][i] == "A":
534 A[i]+=(y[2])
535 elif y[0].split("_")[-1][i] == "C":
536 C[i]+=(y[2])
537 elif y[0].split("_")[-1][i] == "G":
538 G[i]+=(y[2])
539 else:
540 T[i]+=y[2]
541
542 A[:] = [round(x*100,1) / total_reads for x in A]
543 C[:] = [round(x*100,1) / total_reads for x in C]
544 G[:] = [round(x*100,1) / total_reads for x in G]
545 T[:] = [round(x*100,1) / total_reads for x in T]
546
547 data = {'A':A,'C':C,'G':G,'T':T}
548 df = pd.DataFrame(data, index=[1,2,3])
549
550 h=df.plot.bar(color=tuple(["g", "b","gold","r"]) )
551 h.grid(axis='y',linewidth=0.2)
552 plt.xticks(rotation=0, ha="right")
553 plt.ylabel("Counts (%)",fontsize=18)
554 plt.xlabel("Numbers of additional nucleotides",fontsize=18)
555 plt.title(titlos,fontsize=20)
556 plt.tight_layout()
557 plt.savefig(file_bar, dpi=300)
558
559 crp_logo = lm.Logo(df, font_name = 'DejaVu Sans')
560
561 crp_logo.style_spines(visible=False)
562 crp_logo.style_spines(spines=['left', 'bottom'], visible=True)
563 crp_logo.style_xticks(rotation=0, fmt='%d', anchor=0)
564 crp_logo.ax.set_title(titlos,fontsize=18)
565 crp_logo.ax.set_ylabel("Counts (%)", fontsize=16,labelpad=5)
566 crp_logo.ax.set_xlabel("Numbers of additional nucleotides",fontsize=16, labelpad=5)
567 crp_logo.ax.xaxis.set_ticks_position('none')
568 crp_logo.ax.xaxis.set_tick_params(pad=-1)
569 figure = plt.gcf()
570 figure.set_size_inches(6, 4)
571 crp_logo.fig.savefig(file_logo,dpi=300)
572
573 ##########################################################################################################################################################################################################
574
575 def pdf_before_DE(analysis,group_name1,group_name2):
576
577 # Image extensions
578 if analysis=="2":
579 image_extensions = ("c_hist_red.png","t_hist_red.png","pie_non.png","spider_red.png","spider_non_red.png","c_logo.png","t_logo.png","c_bar.png","t_bar.png")
580 else:
581 image_extensions = ("c_hist_red.png","t_hist_red.png","pie_tem.png","spider_red.png","spider_non_red.png")
582 # This list will hold the images file names
583 images = []
584
585 # Build the image list by merging the glob results (a list of files)
586 # for each extension. We are taking images from current folder.
587 for extension in image_extensions:
588 images.extend(glob.glob(extension))
589 #sys.exit(images)
590 # Create instance of FPDF class
591 pdf = FPDF('P', 'in', 'A4')
592 # Add new page. Without this you cannot create the document.
593 pdf.add_page()
594 # Set font to Arial, 'B'old, 16 pts
595 pdf.set_font('Arial', 'B', 20.0)
596
597 # Page header
598 pdf.cell(pdf.w-0.5, 0.5, 'IsomiR Profile Report',align='C')
599 pdf.ln(0.7)
600 pdf.set_font('Arial','', 16.0)
601 pdf.cell(pdf.w-0.5, 0.5, 'sRNA length distribution',align='C')
602
603 # Smaller font for image captions
604 pdf.set_font('Arial', '', 11.0)
605
606 # Image caption
607 pdf.ln(0.5)
608
609 yh=FPDF.get_y(pdf)
610 pdf.image(images[0],x=0.3,w=4, h=3)
611 pdf.image(images[1],x=4,y=yh, w=4, h=3)
612 pdf.ln(0.3)
613
614 # Image caption
615 pdf.cell(0.2)
616 pdf.cell(3.0, 0.0, " Mapped and unmapped reads to custom precussor arm reference DB (5p and 3p arms) in "+group_name1)
617 pdf.ln(0.2)
618 pdf.cell(0.2)
619 pdf.cell(3.0, 0.0, " (left) and "+group_name2+" (right) groups")
620
621
622 pdf.ln(0.5)
623 h1=FPDF.get_y(pdf)
624 pdf.image(images[2],x=1, w=6.5, h=5)
625 h2=FPDF.get_y(pdf)
626 FPDF.set_y(pdf,h1+0.2)
627 pdf.set_font('Arial','', 16.0)
628 pdf.cell(pdf.w-0.5, 0.5, 'Template and non-template isomiRs',align='C')
629 pdf.set_font('Arial', '', 11.0)
630 FPDF.set_y(pdf,h2)
631 FPDF.set_y(pdf,9.5)
632 # Image caption
633 pdf.cell(0.2)
634 if analysis=="2":
635 pdf.cell(3.0, 0.0, " RefSeq miRNAs, templated isomiRs, non-templated isomiRs and unassigned sequences as percentage")
636 pdf.ln(0.2)
637 pdf.cell(0.2)
638 pdf.cell(3.0, 0.0, " of total sRNA reads in "+group_name1+" (left) and "+group_name2+" (right) groups")
639
640 else:
641 pdf.cell(3.0, 0.0, " RefSeq miRNAS, Templated isomiRs and unassigned sequences as percentage of total sRNA reads in")
642 pdf.ln(0.2)
643 pdf.cell(0.2)
644 pdf.cell(3.0, 0.0, " "+group_name1+" (left) and "+group_name2+" (right) groups")
645
646 pdf.add_page()
647 pdf.set_font('Arial', 'B', 18.0)
648 pdf.cell(pdf.w-0.5, 0.5, "Templated isomiR subtypes",align='C')
649 pdf.ln(0.7)
650 pdf.set_font('Arial', 'B', 14.0)
651 pdf.cell(pdf.w-0.5, 0.5, "Templated isomiR profile (redundant reads)",align='C')
652 pdf.ln(0.5)
653 pdf.image(images[3],x=1.5, w=5.5, h=4)
654 pdf.ln(0.6)
655 pdf.cell(pdf.w-0.5, 0.0, "Templated isomiR profile (non-redundant reads)",align='C')
656 pdf.set_font('Arial', '', 12.0)
657 pdf.ln(0.2)
658 pdf.image(images[4],x=1.5, w=5.5, h=4)
659 pdf.ln(0.3)
660 pdf.set_font('Arial', '', 11.0)
661 pdf.cell(0.2)
662 pdf.cell(3.0, 0.0, " * IsomiRs potentially generated from multiple loci")
663
664 if analysis=="2":
665 pdf.add_page('L')
666
667 pdf.set_font('Arial', 'B', 18.0)
668 pdf.cell(pdf.w-0.5, 0.5, "Non-templated isomiRs",align='C')
669 pdf.ln(0.5)
670 pdf.set_font('Arial', 'B', 14.0)
671 pdf.cell(pdf.w-0.5, 0.5, "3'-end additions to RefSeq miRNAs and templated isomiRs",align='C')
672 pdf.ln(0.7)
673
674 yh=FPDF.get_y(pdf)
675 pdf.image(images[5],x=1.5,w=3.65, h=2.65)
676 pdf.image(images[7],x=6.5,y=yh, w=3.65, h=2.65)
677 pdf.ln(0.5)
678 yh=FPDF.get_y(pdf)
679 pdf.image(images[6],x=1.5,w=3.65, h=2.65)
680 pdf.image(images[8],x=6.5,y=yh, w=3.65, h=2.65)
681
682 pdf.close()
683 pdf.output('report1.pdf','F')
684
685 #############################################################################################################################################################3
686