Decreased amyloid-related structure-function coupling in preclinical Alzheimer's disease.

1. # -*- coding: utf-8 -*-
2. """
3. Functions for generating workflow methods reports
5. Note: all text contained within this module is released under a `CC-0 license
6. <https://creativecommons.org/publicdomain/zero/1.0/>`_.
7. """
9. import logging
11. import numpy as np
13. from . import __version__
14. from .datasets import check_donors, fetch_donor_info
15. from .images import coerce_atlas_to_dict
16. from .utils import first_entry
18. LGR = logging.getLogger('abagen')
19. METRICS = {
20. np.mean: 'mean',
21. np.median: 'median'
22. }
23. REFERENCES = dict(
24. A2019N=('Arnatkevic̆iūtė, A., Fulcher, B. D., & Fornito, A. (2019). '
25. 'A practical guide to linking brain-wide gene expression and '
26. 'neuroimaging data. Neuroimage, 189, 353-367.'),
27. F2016P=('Fulcher, B. D., & Fornito, A. (2016). A transcriptional '
28. 'signature of hub connectivity in the mouse connectome. '
29. 'Proceedings of the National Academy of Sciences, 113(5), '
30. '1435-1440.'),
31. F2013J=('Fulcher, B. D., Little, M. A., & Jones, N. S. (2013). '
32. 'Highly comparative time-series analysis: the empirical '
33. 'structure of time series and their methods. Journal of '
34. 'the Royal Society Interface, 10(83), 20130048.'),
35. H2012N=('Hawrylycz, M. J., Lein, E. S., Guillozet-Bongaarts, A. L., '
36. 'Shen, E. H., Ng, L., Miller, J. A., ... & Jones, A. R. '
37. '(2012). An anatomically comprehensive atlas of the adult '
38. 'human brain transcriptome. Nature, 489(7416), 391-399.'),
39. H2015N=('Hawrylycz, M., Miller, J. A., Menon, V., Feng, D., '
40. 'Dolbeare, T., Guillozet-Bongaarts, A. L., ... & Lein, E. '
41. '(2015). Canonical genetic signatures of the adult human '
42. 'brain. Nature Neuroscience, 18(12), 1832.'),
43. M2021B=('Markello, R.D., Arnatkevic̆iūtė, A., Poline, J-B., Fulcher, B. '
44. 'D., Fornito, A., & Misic, B. (2021). Standardizing workflows in '
45. 'imaging transcriptomics with the abagen toolbox. Biorxiv.'),
46. P2017G=('Parkes, L., Fulcher, B. D., Yücel, M., & Fornito, A. '
47. '(2017). Transcriptional signatures of connectomic '
48. 'subregions of the human striatum. Genes, Brain and '
49. 'Behavior, 16(7), 647-663.'),
50. Q2002N=('Quackenbush, J. (2002). Microarray data normalization and '
51. 'transformation. Nature Genetics, 32(4), 496-501.'),
52. R2018N=('Romero-Garcia, R., Whitaker, K. J., Váša, F., Seidlitz, '
53. 'J., Shinn, M., Fonagy, P., ... & NSPN Consortium. (2018). '
54. 'Structural covariance networks are coupled to expression '
55. 'of genes enriched in supragranular layers of the human '
56. 'cortex. NeuroImage, 171, 256-267.')
57. )
60. class Report:
61. """ Generates report of methods for :func:`abagen.get_expression_data()`
63. Refer to the doc-string of the workflow for an overview of paramter options
64. """
66. def __init__(self, atlas, atlas_info=None, *, ibf_threshold=0.5,
67. probe_selection='diff_stability', donor_probes='aggregate',
68. lr_mirror=None, missing=None, tolerance=2, sample_norm='srs',
69. gene_norm='srs', norm_matched=True, norm_structures=False,
70. region_agg='donors', agg_metric='mean', corrected_mni=True,
71. reannotated=True, donors='all', return_donors=False,
72. data_dir=None, counts=None, n_probes=None, n_genes=None):
74. efflevel = LGR.getEffectiveLevel()
75. LGR.setLevel(100)
77. atlas, self.group_atlas = \
78. coerce_atlas_to_dict(atlas, donors, atlas_info=atlas_info,
79. data_dir=data_dir)
80. self.atlas = first_entry(atlas)
81. self.atlas_info = self.atlas.atlas_info
82. self.ibf_threshold = ibf_threshold
83. self.probe_selection = probe_selection
84. self.donor_probes = donor_probes
85. self.lr_mirror = lr_mirror
86. self.missing = missing
87. self.tolerance = tolerance
88. self.sample_norm = sample_norm
89. self.gene_norm = gene_norm
90. self.norm_matched = norm_matched
91. self.norm_structures = norm_structures
92. self.region_agg = region_agg
93. self.agg_metric = METRICS.get(agg_metric, agg_metric)
94. self.corrected_mni = corrected_mni
95. self.reannotated = reannotated
96. self.donors = donors
97. self.return_donors = return_donors
98. self.counts = counts
99. self.n_probes = n_probes
100. self.n_genes = n_genes
101. self.body = self.gen_report()
103. LGR.setLevel(efflevel)
105. def gen_report(self):
106. """ Generates main text of report
107. """
109. report = ''
110. report += """
111. Regional microarry expression data were obtained from {n_donors}
112. post-mortem brains ({n_female} female, ages {min}--{max}, {mean:.2f}
113. +/- {std:.2f}) provided by the Allen Human Brain Atlas (AHBA,
114. https://human.brain-map.org; [H2012N]). Data were processed with the
115. abagen toolbox (version {vers}; https://github.com/rmarkello/abagen;
116. [M2021B])
117. """.format(**_get_donor_demographics(self.donors), vers=__version__)
119. if self.atlas.volumetric and self.group_atlas:
120. report += """
121. using a {n_region}-region volumetric atlas in MNI space.<br>
122. """.format(n_region=len(self.atlas.labels))
123. elif not self.atlas.volumetric and self.group_atlas:
124. report += """
125. using a {n_region}-region surface-based atlas in MNI space.<br>
126. """.format(n_region=len(self.atlas.labels))
127. elif self.atlas.volumetric and not self.group_atlas:
128. report += """
129. using a {n_region}-region volumetric atlas, independently aligned
130. to each donor's native MRI space.<br>
131. """.format(n_region=len(self.atlas.labels))
132. else:
133. report += ".<br>"
135. if self.reannotated:
136. report += """
137. First, microarray probes were reannotated using data provided by
138. [A2019N]; probes not matched to a valid Entrez ID were discarded.
139. """
140. else:
141. report += """
142. First, microarray probes not matched to a valid Entrez ID were
143. discarded.
144. """
146. if self.ibf_threshold > 0:
147. report += """
148. Next, probes were filtered based on their expression intensity
149. relative to background noise [Q2002N], such that probes with
150. intensity less than the background in >={threshold:.2f}% of samples
151. across donors were discarded
152. """.format(threshold=self.ibf_threshold * 100)
153. if self.n_probes is not None:
154. report += """
155. , yielding {n_probes:,} probes
156. """.format(n_probes=self.n_probes)
157. report += "."
159. if self.probe_selection == 'average':
160. report += """
161. When multiple probes indexed the expression of the same gene we
162. calculated the mean expression across probes.
163. """
164. elif self.probe_selection == 'diff_stability':
165. report += r"""
166. When multiple probes indexed the expression of the same gene, we
167. selected and used the probe with the most consistent pattern of
168. regional variation across donors (i.e., differential stability;
169. [H2015N]), calculated with:<br>
171. $$ \Delta_{{S}}(p) = \frac{{1}}{{\binom{{N}}{{2}}}} \,
172. \sum_{{i=1}}^{{N-1}} \sum_{{j=i+1}}^{{N}}
173. \rho[B_{{i}}(p), B_{{j}}(p)] $$<br>
175. where $ \rho $ is Spearman's rank correlation of the expression of
176. a single probe, p, across regions in two donors $B_{{i}}$ and
177. $B_{{j}}$, and N is the total number of donors. Here, regions
178. correspond to the structural designations provided in the ontology
179. from the AHBA.
180. """
181. elif self.probe_selection == "pc_loading":
182. report += """
183. When multiple probes indexed the expression of the same gene we
184. selected and used the probe with the highest loading on the first
185. principal component taken from a decomposition of probe expression
186. across samples from all donors [P2017G].
187. """
188. elif self.probe_selection == "max_intensity":
189. report += """
190. When multiple probes indexed the expression of the same gene we
191. selected and used the probe with the highest mean intensity across
192. samples.
193. """
194. elif self.probe_selection == "max_variance":
195. report += """
196. When multiple probes indexed the expression of the same gene we
197. selected and used the probe with the highest variance across
198. samples.
199. """
200. elif self.probe_selection == "corr_intensity":
201. report += """
202. When multiple probes indexed the expression of the same gene we
203. selected and used the expression profile of a single representative
204. probe. For genes where only two probes were available we selected
205. the probe with the highest mean intensity across samples; where
206. three or more probes were available, we calculated the correlation
207. of probe expression across samples and selected the probe with the
208. highest average correlation.
209. """
210. elif self.probe_selection == "corr_variance":
211. report += """
212. When multiple probes indexed the expression of the same gene we
213. selected and used the expression profile of a single representative
214. probe. For genes where only two probes were available we selected
215. the probe with the highest variance across samples; where three or
216. more probes were available, we calculated the correlation of probe
217. expression across samples and selected the probe with the highest
218. average correlation.
219. """
220. elif self.probe_selection == "rnaseq":
221. report += """
222. When multiple probes indexed the expression of the same gene we
223. selected and used the probe with the most consistent pattern of
224. regional expression to RNA-seq data (available for two donors in
225. the AHBA). That is, we calculated the Spearman’s rank correlation
226. between each probes' microarray expression and RNA-seq expression
227. data of the corresponding gene, and selected the probe with the
228. highest correspondence. Here, regions correspond to the structural
229. designations provided in the ontology from the AHBA.
230. """
232. if (self.donor_probes == "aggregate" and self.probe_selection not in
233. ['average', 'diff_stability', 'rnaseq']):
234. report += """
235. The selection of probes was performed using sample expression data
236. aggregated across all donors.<br>
237. """
238. elif (self.donor_probes == "aggregate" and self.probe_selection in
239. ['average', 'diff_stability', 'rnaseq']):
240. report += "<br>"
241. elif self.donor_probes == "independent":
242. report += """
243. The selection of probes was performed independently for each donor,
244. such that the probes chosen to represent each gene could differ
245. across donors.<br>
246. """
247. elif self.donor_probes == "common":
248. report += """
249. The selection of probes was performed independently for each donor,
250. and the probe most commonly selected across all donors was chosen
251. to represent the expression of each gene for all donors.<br>
252. """
254. if self.corrected_mni and self.group_atlas:
255. report += """
256. The MNI coordinates of tissue samples were updated to those
257. generated via non-linear registration using the Advanced
258. Normalization Tools (ANTs; https://github.com/chrisfilo/alleninf).
259. """
261. if self.lr_mirror == 'bidirectional':
262. report += """
263. To increase spatial coverage, tissue samples were mirrored
264. bilaterally across the left and right hemispheres [R2018N].
265. """
266. elif self.lr_mirror == 'leftright':
267. report += """
268. To increase spatial coverage, tissue samples in the left hemisphere
269. were mirrored into the right hemisphere [R2018N].
270. """
271. elif self.lr_mirror == 'rightleft':
272. report += """
273. To increase spatial coverage, tissue samples in the right
274. hemisphere were mirrored into the left hemisphere [R2018N].
275. """
277. if self.tolerance == 0 and not self.atlas.volumetric:
278. report += """
279. Samples were assigned to brain regions by minimizing the Euclidean
280. distance between the {space} coordinates of each sample and the
281. nearest surface vertex. Samples where the Euclidean distance to the
282. nearest vertex was greater than the mean distance for all samples
283. belonging to that donor were excluded.
284. """.format(space='MNI' if self.group_atlas else 'native voxel')
285. elif self.tolerance == 0 and self.atlas.volumetric:
286. report += """
287. Samples were assigned to brain regions in the provided atlas only
288. if their {space} coordinates were directly within a voxel belonging
289. to a parcel.
290. """.format(space='MNI' if self.group_atlas else 'native voxel')
291. elif self.tolerance > 0 and not self.atlas.volumetric:
292. report += """
293. Samples were assigned to brain regions by minimizing the Euclidean
294. distance between the {space} coordinates of each sample and the
295. nearest surface vertex. Samples where the Euclidean distance to the
296. nearest vertex was more than {tolerance} standard deviations above
297. the mean distance for all samples belonging to that donor were
298. excluded.
299. """.format(space='MNI' if self.group_atlas else 'native voxel',
300. tolerance=self.tolerance)
301. elif self.tolerance > 0 and self.atlas.volumetric:
302. report += """
303. Samples were assigned to brain regions in the provided atlas if
304. their {space} coordinates were within {tolerance} mm of a given
305. parcel.
306. """.format(space='MNI' if self.group_atlas else 'native voxel',
307. tolerance=self.tolerance)
308. elif self.tolerance < 0 and not self.atlas.volumetric:
309. report += """
310. Samples were assigned to brain regions by minimizing the Euclidean
311. distance between the {space} coordinates of each sample and the
312. nearest surface vertex. Samples where the Euclidean distance to the
313. nearest vertex was more than {tolerance}mm were excluded.
314. """.format(space='MNI' if self.group_atlas else 'native voxel',
315. tolerance=abs(self.tolerance))
317. if self.atlas_info is not None:
318. report += """
319. To reduce the potential for misassignment, sample-to-region
320. matching was constrained by hemisphere and gross structural
321. divisions (i.e., cortex, subcortex/brainstem, and cerebellum, such
322. that e.g., a sample in the left cortex could only be assigned to an
323. atlas parcel in the left cortex; [A2019N]).
324. """
326. if self.missing == 'centroids':
327. report += """
328. If a brain region was not assigned a sample from any donor based on
329. the above procedure, the tissue sample closest to the centroid of
330. that parcel was identified independently for each donor. The
331. average of these samples was taken across donors, weighted by the
332. distance between the parcel centroid and the sample, to obtain an
333. estimate of the parcellated expression values for the missing
334. region.
335. """
336. if self.counts is not None:
337. n_missing = np.sum(self.counts.sum(axis=1) == 0)
338. if n_missing > 0:
339. report += """
340. This procedure was performed for {n_missing} regions that
341. were not assigned tissue samples.
342. """.format(n_missing=n_missing)
343. elif self.missing == 'interpolate':
344. report += """
345. If a brain region was not assigned a tissue sample based on the
346. above procedure, every {node} in the region was mapped to the
347. nearest tissue sample from the donor in order to generate a dense,
348. interpolated expression map. The average of these expression values
349. was taken across all {nodes} in the region, weighted by the
350. distance between each {node} and the sample mapped to it, in order
351. to obtain an estimate of the parcellated expression values for the
352. missing region.
353. """.format(node='voxel' if self.atlas.volumetric else 'vertex',
354. nodes='voxels' if self.atlas.volumetric else 'vertices')
356. if self.norm_matched:
357. report += """
358. All tissue samples not assigned to a brain region in the provided
359. atlas were discarded.
360. """
361. report += "<br>"
363. if self.sample_norm is not None:
364. report += """
365. Inter-subject variation was addressed {}
366. """.format(_get_norm_procedure(self.sample_norm, 'sample_norm'))
368. if self.gene_norm is None:
369. pass
370. elif self.gene_norm == self.sample_norm:
371. report += """
372. Gene expression values were then normalized across tissue samples
373. using an identical procedure.
374. """
375. elif (self.gene_norm != self.sample_norm
376. and self.sample_norm is not None):
377. report += """
378. Inter-subject variation in gene expression was then addressed {}
379. """.format(_get_norm_procedure(self.gene_norm, 'gene_norm'))
380. elif self.gene_norm != self.sample_norm and self.sample_norm is None:
381. report += """
382. Inter-subject variation was addressed {}
383. """.format(_get_norm_procedure(self.gene_norm, 'gene_norm'))
385. if not self.norm_matched and not self.norm_structures:
386. report += """
387. All available tissue samples were used in the normalization process
388. regardless of whether they were assigned to a brain region.
389. """
390. elif not self.norm_matched and self.norm_structures:
391. report += """
392. All available tissue samples were used in the normalization process
393. regardless of whether they were matched to a brain region; however,
394. normalization was performed separately for samples in distinct
395. structural classes (i.e., cortex, subcortex/brainstem, cerebellum).
396. """
397. elif self.norm_matched and self.norm_structures:
398. report += """
399. Normalization was performed separately for samples in distinct
400. structural classes (i.e., cortex, subcortex/brainstem, cerebellum).
401. """
403. if not self.norm_matched and self.gene_norm is not None:
404. report += """
405. Tissue samples not matched to a brain region were discarded after
406. normalization.
407. """
409. if self.region_agg == 'donors' and self.agg_metric == 'mean':
410. report += """
411. Samples assigned to the same brain region were averaged separately
412. for each donor{donors}, yielding a regional expression matrix
413. {n_donors}
414. """.format(donors=' and then across donors' if
415. not self.return_donors else '',
416. n_donors=' for each donor ' if
417. self.return_donors else '')
418. elif self.region_agg == 'samples' and self.agg_metric == 'mean':
419. report += """
420. Samples assigned to the same brain region were averaged across all
421. donors, yielding a regional expression matrix
422. """
423. elif self.region_agg == 'donors' and self.agg_metric == 'median':
424. report += """
425. The median value of samples assigned to the same brain region was
426. computed separately for each donor{donors}, yielding a regional
427. expression matrix{n_donors}
428. """.format(donors=' and then across donors' if
429. not self.return_donors else '',
430. n_donors=' for each donor' if
431. self.return_donors else '')
432. elif self.region_agg == 'samples' and self.agg_metric == 'median':
433. report += """
434. The median value of samples assigned to the same brain region was
435. computed across donors, yielding a single regional expression
436. matrix
437. """
438. if self.n_genes is not None:
439. report += """
440. with {n_region} rows, corresponding to brain regions, and
441. {n_genes:,} columns, corresponding to the retained genes
442. """.format(n_region=len(self.atlas.labels),
443. n_genes=self.n_genes)
444. report += "\b."
446. report += str(_add_references(report))
447. return _sanitize_text(report)
450. def _sanitize_text(text):
451. """ Cleans up line and paragraph breaks in `text`
452. """
454. text = ' '.join(text.replace('\n', ' ').split())
455. return text.replace('<br> ', '\n\n') \
456. .replace('<p>', '\n')
459. def _get_donor_demographics(donors):
460. """
461. Gets demographic info about the requested `donors`
463. Parameters
464. ----------
465. donors : list, optional
466. List of donors. Can be either donor numbers or UID. If not specified
467. will use all available donors. Note that donors '9861' and '10021'
468. have samples from both left + right hemispheres; all other donors have
469. samples from the left hemisphere only. Default: 'all'
471. Returns
472. -------
473. info : dict
474. With keys ['n_donor', 'n_female','min', 'max', 'mean', 'std']. The
475. values represent features of the age distribution of requested donors.
476. """
478. donors = [int(i) for i in check_donors(donors)]
479. info = fetch_donor_info().set_index('uid').loc[donors]
480. age_info = info['age'].describe()
482. dinfo = dict(n_donors=len(info), n_female=sum(info['sex'] == 'F'))
483. dinfo.update(age_info.loc[['min', 'max', 'mean', 'std']].to_dict())
485. return dinfo
488. def _get_norm_procedure(norm, parameter):
489. """
490. Returns methods description of the provided `norm`
492. Parameters
493. ----------
494. norm : str
495. Normalization procedure
497. Returns
498. -------
499. procedure : str
500. Reporting procedures
501. """
503. if parameter not in ('sample_norm', 'gene_norm'):
504. raise ValueError(f'Invalid norm parameter {parameter}')
506. if parameter == 'sample_norm':
507. mod = ('tissue sample expression values across genes')
508. suff = ('tissue sample across genes')
509. else:
510. mod = ('gene expression values across tissue samples')
511. suff = ('gene across tissue samples')
513. sigmoid = r"""
514. independently for each donor using a sigmoid function [F2016P]:<br>
516. $$ x_{{norm}} = \frac{{1}}{{1 + \exp(-\frac{{(x - \overline{{x}})}}
517. {{\sigma_{{x}}}})}} $$<br>
519. where $\bar{{x}}$ is the arithmetic mean and $\sigma_{{x}}$ is the
520. sample standard deviation of the expression of a single
521. """
522. robust_sigmoid = r"""
523. using a robust sigmoid function [F2013J]:<br>
525. $$ x_{{norm}} = \frac{{1}}{{1 + \exp(-\frac{{(x-\langle x \rangle)}}
526. {{\text{{IQR}}_{{x}}}})}} $$<br>
528. where $\langle x \rangle$ is the median and $\text{{IQR}}_{{x}}$
529. is the normalized interquartile range of the expression of a single
530. """
531. rescale = r"""
532. Normalized expression values were then rescaled to the unit interval: <br>
534. $$ x_{{scaled}} = \frac{{x_{{norm}} - \min(x_{{norm}})}}
535. {{\max(x_{{norm}}) - \min(x_{{norm}})}} $$<br>
536. """
538. procedure = ''
539. if norm in ['center', 'demean']:
540. procedure += """
541. by demeaning {mod} independently for each donor.
542. """.format(mod=mod)
543. elif norm == 'zscore':
544. procedure += """
545. by mean- and variance-normalizing (i.e., z-scoring) {mod} independently
546. for each donor.
547. """.format(mod=mod)
548. elif norm == 'minmax':
549. procedure += r"""
550. by rescaling {mod} to the unit interval independently for each donor:
551. <br>
553. $$ x_{{{{scaled}}}} = \frac{{{{x - \min(x)}}}}{{{{\max(x) - \min(x)}}}}
554. $$<br>
555. """.format(mod=mod)
556. elif norm in ['sigmoid', 'sig']:
557. procedure += r"""
558. by normalizing {mod} {sigmoid} {suff}.
559. """.format(mod=mod, sigmoid=sigmoid, suff=suff)
560. elif norm in ['scaled_sigmoid', 'scaled_sig']:
561. procedure += r"""
562. by normalizing {mod} {sigmoid} {suff}. {rescale}
563. """.format(mod=mod, sigmoid=sigmoid, suff=suff, rescale=rescale)
564. elif norm in ['scaled_sigmoid_quantiles', 'scaled_sig_qnt']:
565. procedure += r"""
566. by normalizing {mod} {sigmoid} {suff}, calculated using only data in
567. the 5–95th percentile range to downweight the impact of outliers.
568. {rescale}
569. """.format(mod=mod, sigmoid=sigmoid, suff=suff, rescale=rescale)
570. elif norm in ['robust_sigmoid', 'rsig', 'rs']:
571. procedure += r"""
572. by normalizing {mod} {robust_sigmoid} {suff}.
573. """.format(mod=mod, robust_sigmoid=robust_sigmoid, suff=suff)
574. elif norm in ['scaled_robust_sigmoid', 'scaled_rsig', 'srs']:
575. procedure += r"""
576. by normalizing {mod} {robust_sigmoid} {suff}. {rescale}
577. """.format(mod=mod, robust_sigmoid=robust_sigmoid, suff=suff,
578. rescale=rescale)
579. elif norm in ['mixed_sigmoid', 'mixed_sig']:
580. procedure += r"""
581. by normalizing {mod} using a mixed sigmoid function [F2013J]:<br>
583. $$ x_{{{{norm}}}} = \left\{{{{\begin{{{{array}}}}{{{{r r}}}}
584. \frac{{{{1}}}}{{{{1 + \exp(-\frac{{{{(x-\overline{{{{x}}}})}}}}
585. {{{{\sigma_{{{{x}}}}}}}})}}}} ,& \text{{{{IQR}}}}_{{{{x}}}} = 0
586. \frac{{{{1}}}}{{{{1 + \exp(-\frac{{{{(x-\langle x \rangle)}}}}
587. {{{{\text{{{{IQR}}}}_{{{{x}}}}}}}})}}}} ,& \text{{{{IQR}}}}_{{{{x}}}}
588. \neq 0 \end{{{{array}}}}\right. $$<br>
590. where $\bar{{{{x}}}}$ is the arithmetic mean, $\sigma_{{{{x}}}}$ is the
591. sample standard deviation, $\langle x \rangle$ is the median, and
592. $\text{{{{IQR}}}}_{{{{x}}}}$ is the normalized interquartile range of
593. the expression value of a single {suff}. {rescale}
594. """.format(mod=mod, suff=suff, rescale=rescale)
596. return procedure
599. def _add_references(report):
600. """
601. Detects references in `report` and generates list
603. Parameters
604. ----------
605. report : str
606. Report body
608. Returns
609. -------
610. references : str
611. List of references to be appended to `report`
612. """
614. refreport = ''
615. for ref, cite in REFERENCES.items():
616. if ref in report:
617. refreport += f'[{ref}]: {cite}<p>'
619. if len(refreport) > 0:
620. refreport = '<br> REFERENCES<p>----------<p>' + refreport
622. if refreport.endswith('<p> '):
623. refreport = refreport[:-4]
625. return refreport

reporting.py at commit dc4a007, under BSD-3-Clause · at the source