├── DESCRIPTION
├── LICENSE
├── NAMESPACE
├── NEWS.md
├── R
├── bAccuracy.R
├── data.R
├── matPCs.R
└── multiAdaSampling.R
├── README.md
├── data
└── gse87795_subset_sce.rda
├── img
├── scReClassify.jpg
└── scReClassify_sticker.png
├── inst
├── CITATION
├── scReClassify.jpg
└── scReClassify_sticker.png
├── man
├── bAccuracy.Rd
├── gse87795_subset_sce.Rd
├── matPCs.Rd
├── multiAdaSampling.Rd
└── scReClassify.Rd
├── tests
├── testthat.R
└── testthat
│ ├── test-bAccuracy.R
│ ├── test-matPCs.R
│ └── test-multiAdaSampling.R
└── vignettes
└── scReClassify.Rmd
/DESCRIPTION:
--------------------------------------------------------------------------------
1 | Package: scReClassify
2 | Type: Package
3 | Title: scReClassify: post hoc cell type classification of single-cell RNA-seq data
4 | Version: 0.99.8
5 | Authors@R: c(
6 | person("Pengyi", "Yang", email = "pengyi.yang@sydney.edu.au",
7 | role = c("aut"), comment = c(ORCID = "0000-0003-1098-3138")),
8 | person("Taiyun", "Kim", email = "taiyun.kim91@gmail.com",
9 | role = c("aut", "cre"), comment = c(ORCID = "0000-0002-5028-836X")))
10 | Description: A post hoc cell type classification tool to fine-tune cell type
11 | annotations generated by any cell type classification procedure with
12 | semi-supervised learning algorithm AdaSampling technique.
13 | The current version of scReClassify supports Support Vector Machine and
14 | Random Forest as a base classifier.
15 | License: GPL-3 + file LICENSE
16 | BugReports: https://github.com/SydneyBioX/scReClassify/issues
17 | URL: https://github.com/SydneyBioX/scReClassify, http://www.bioconductor.org/packages/release/bioc/html/scReClassify.html
18 | Depends: R (>= 4.1)
19 | Encoding: UTF-8
20 | LazyData: false
21 | RoxygenNote: 7.1.2
22 | Roxygen: list(markdown = TRUE)
23 | Imports:
24 | randomForest,
25 | e1071,
26 | stats,
27 | SummarizedExperiment,
28 | SingleCellExperiment,
29 | methods
30 | VignetteBuilder: knitr
31 | biocViews: Software, Transcriptomics, SingleCell, Classification, SupportVectorMachine
32 | Suggests:
33 | testthat,
34 | knitr,
35 | BiocStyle,
36 | rmarkdown,
37 | DT,
38 | mclust,
39 | dplyr
40 |
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | GNU GENERAL PUBLIC LICENSE
2 | Version 3, 29 June 2007
3 |
4 | Copyright (C) 2007 Free Software Foundation, Inc.
5 | Everyone is permitted to copy and distribute verbatim copies
6 | of this license document, but changing it is not allowed.
7 |
8 | Preamble
9 |
10 | The GNU General Public License is a free, copyleft license for
11 | software and other kinds of works.
12 |
13 | The licenses for most software and other practical works are designed
14 | to take away your freedom to share and change the works. By contrast,
15 | the GNU General Public License is intended to guarantee your freedom to
16 | share and change all versions of a program--to make sure it remains free
17 | software for all its users. We, the Free Software Foundation, use the
18 | GNU General Public License for most of our software; it applies also to
19 | any other work released this way by its authors. You can apply it to
20 | your programs, too.
21 |
22 | When we speak of free software, we are referring to freedom, not
23 | price. Our General Public Licenses are designed to make sure that you
24 | have the freedom to distribute copies of free software (and charge for
25 | them if you wish), that you receive source code or can get it if you
26 | want it, that you can change the software or use pieces of it in new
27 | free programs, and that you know you can do these things.
28 |
29 | To protect your rights, we need to prevent others from denying you
30 | these rights or asking you to surrender the rights. Therefore, you have
31 | certain responsibilities if you distribute copies of the software, or if
32 | you modify it: responsibilities to respect the freedom of others.
33 |
34 | For example, if you distribute copies of such a program, whether
35 | gratis or for a fee, you must pass on to the recipients the same
36 | freedoms that you received. You must make sure that they, too, receive
37 | or can get the source code. And you must show them these terms so they
38 | know their rights.
39 |
40 | Developers that use the GNU GPL protect your rights with two steps:
41 | (1) assert copyright on the software, and (2) offer you this License
42 | giving you legal permission to copy, distribute and/or modify it.
43 |
44 | For the developers' and authors' protection, the GPL clearly explains
45 | that there is no warranty for this free software. For both users' and
46 | authors' sake, the GPL requires that modified versions be marked as
47 | changed, so that their problems will not be attributed erroneously to
48 | authors of previous versions.
49 |
50 | Some devices are designed to deny users access to install or run
51 | modified versions of the software inside them, although the manufacturer
52 | can do so. This is fundamentally incompatible with the aim of
53 | protecting users' freedom to change the software. The systematic
54 | pattern of such abuse occurs in the area of products for individuals to
55 | use, which is precisely where it is most unacceptable. Therefore, we
56 | have designed this version of the GPL to prohibit the practice for those
57 | products. If such problems arise substantially in other domains, we
58 | stand ready to extend this provision to those domains in future versions
59 | of the GPL, as needed to protect the freedom of users.
60 |
61 | Finally, every program is threatened constantly by software patents.
62 | States should not allow patents to restrict development and use of
63 | software on general-purpose computers, but in those that do, we wish to
64 | avoid the special danger that patents applied to a free program could
65 | make it effectively proprietary. To prevent this, the GPL assures that
66 | patents cannot be used to render the program non-free.
67 |
68 | The precise terms and conditions for copying, distribution and
69 | modification follow.
70 |
71 | TERMS AND CONDITIONS
72 |
73 | 0. Definitions.
74 |
75 | "This License" refers to version 3 of the GNU General Public License.
76 |
77 | "Copyright" also means copyright-like laws that apply to other kinds of
78 | works, such as semiconductor masks.
79 |
80 | "The Program" refers to any copyrightable work licensed under this
81 | License. Each licensee is addressed as "you". "Licensees" and
82 | "recipients" may be individuals or organizations.
83 |
84 | To "modify" a work means to copy from or adapt all or part of the work
85 | in a fashion requiring copyright permission, other than the making of an
86 | exact copy. The resulting work is called a "modified version" of the
87 | earlier work or a work "based on" the earlier work.
88 |
89 | A "covered work" means either the unmodified Program or a work based
90 | on the Program.
91 |
92 | To "propagate" a work means to do anything with it that, without
93 | permission, would make you directly or secondarily liable for
94 | infringement under applicable copyright law, except executing it on a
95 | computer or modifying a private copy. Propagation includes copying,
96 | distribution (with or without modification), making available to the
97 | public, and in some countries other activities as well.
98 |
99 | To "convey" a work means any kind of propagation that enables other
100 | parties to make or receive copies. Mere interaction with a user through
101 | a computer network, with no transfer of a copy, is not conveying.
102 |
103 | An interactive user interface displays "Appropriate Legal Notices"
104 | to the extent that it includes a convenient and prominently visible
105 | feature that (1) displays an appropriate copyright notice, and (2)
106 | tells the user that there is no warranty for the work (except to the
107 | extent that warranties are provided), that licensees may convey the
108 | work under this License, and how to view a copy of this License. If
109 | the interface presents a list of user commands or options, such as a
110 | menu, a prominent item in the list meets this criterion.
111 |
112 | 1. Source Code.
113 |
114 | The "source code" for a work means the preferred form of the work
115 | for making modifications to it. "Object code" means any non-source
116 | form of a work.
117 |
118 | A "Standard Interface" means an interface that either is an official
119 | standard defined by a recognized standards body, or, in the case of
120 | interfaces specified for a particular programming language, one that
121 | is widely used among developers working in that language.
122 |
123 | The "System Libraries" of an executable work include anything, other
124 | than the work as a whole, that (a) is included in the normal form of
125 | packaging a Major Component, but which is not part of that Major
126 | Component, and (b) serves only to enable use of the work with that
127 | Major Component, or to implement a Standard Interface for which an
128 | implementation is available to the public in source code form. A
129 | "Major Component", in this context, means a major essential component
130 | (kernel, window system, and so on) of the specific operating system
131 | (if any) on which the executable work runs, or a compiler used to
132 | produce the work, or an object code interpreter used to run it.
133 |
134 | The "Corresponding Source" for a work in object code form means all
135 | the source code needed to generate, install, and (for an executable
136 | work) run the object code and to modify the work, including scripts to
137 | control those activities. However, it does not include the work's
138 | System Libraries, or general-purpose tools or generally available free
139 | programs which are used unmodified in performing those activities but
140 | which are not part of the work. For example, Corresponding Source
141 | includes interface definition files associated with source files for
142 | the work, and the source code for shared libraries and dynamically
143 | linked subprograms that the work is specifically designed to require,
144 | such as by intimate data communication or control flow between those
145 | subprograms and other parts of the work.
146 |
147 | The Corresponding Source need not include anything that users
148 | can regenerate automatically from other parts of the Corresponding
149 | Source.
150 |
151 | The Corresponding Source for a work in source code form is that
152 | same work.
153 |
154 | 2. Basic Permissions.
155 |
156 | All rights granted under this License are granted for the term of
157 | copyright on the Program, and are irrevocable provided the stated
158 | conditions are met. This License explicitly affirms your unlimited
159 | permission to run the unmodified Program. The output from running a
160 | covered work is covered by this License only if the output, given its
161 | content, constitutes a covered work. This License acknowledges your
162 | rights of fair use or other equivalent, as provided by copyright law.
163 |
164 | You may make, run and propagate covered works that you do not
165 | convey, without conditions so long as your license otherwise remains
166 | in force. You may convey covered works to others for the sole purpose
167 | of having them make modifications exclusively for you, or provide you
168 | with facilities for running those works, provided that you comply with
169 | the terms of this License in conveying all material for which you do
170 | not control copyright. Those thus making or running the covered works
171 | for you must do so exclusively on your behalf, under your direction
172 | and control, on terms that prohibit them from making any copies of
173 | your copyrighted material outside their relationship with you.
174 |
175 | Conveying under any other circumstances is permitted solely under
176 | the conditions stated below. Sublicensing is not allowed; section 10
177 | makes it unnecessary.
178 |
179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
180 |
181 | No covered work shall be deemed part of an effective technological
182 | measure under any applicable law fulfilling obligations under article
183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or
184 | similar laws prohibiting or restricting circumvention of such
185 | measures.
186 |
187 | When you convey a covered work, you waive any legal power to forbid
188 | circumvention of technological measures to the extent such circumvention
189 | is effected by exercising rights under this License with respect to
190 | the covered work, and you disclaim any intention to limit operation or
191 | modification of the work as a means of enforcing, against the work's
192 | users, your or third parties' legal rights to forbid circumvention of
193 | technological measures.
194 |
195 | 4. Conveying Verbatim Copies.
196 |
197 | You may convey verbatim copies of the Program's source code as you
198 | receive it, in any medium, provided that you conspicuously and
199 | appropriately publish on each copy an appropriate copyright notice;
200 | keep intact all notices stating that this License and any
201 | non-permissive terms added in accord with section 7 apply to the code;
202 | keep intact all notices of the absence of any warranty; and give all
203 | recipients a copy of this License along with the Program.
204 |
205 | You may charge any price or no price for each copy that you convey,
206 | and you may offer support or warranty protection for a fee.
207 |
208 | 5. Conveying Modified Source Versions.
209 |
210 | You may convey a work based on the Program, or the modifications to
211 | produce it from the Program, in the form of source code under the
212 | terms of section 4, provided that you also meet all of these conditions:
213 |
214 | a) The work must carry prominent notices stating that you modified
215 | it, and giving a relevant date.
216 |
217 | b) The work must carry prominent notices stating that it is
218 | released under this License and any conditions added under section
219 | 7. This requirement modifies the requirement in section 4 to
220 | "keep intact all notices".
221 |
222 | c) You must license the entire work, as a whole, under this
223 | License to anyone who comes into possession of a copy. This
224 | License will therefore apply, along with any applicable section 7
225 | additional terms, to the whole of the work, and all its parts,
226 | regardless of how they are packaged. This License gives no
227 | permission to license the work in any other way, but it does not
228 | invalidate such permission if you have separately received it.
229 |
230 | d) If the work has interactive user interfaces, each must display
231 | Appropriate Legal Notices; however, if the Program has interactive
232 | interfaces that do not display Appropriate Legal Notices, your
233 | work need not make them do so.
234 |
235 | A compilation of a covered work with other separate and independent
236 | works, which are not by their nature extensions of the covered work,
237 | and which are not combined with it such as to form a larger program,
238 | in or on a volume of a storage or distribution medium, is called an
239 | "aggregate" if the compilation and its resulting copyright are not
240 | used to limit the access or legal rights of the compilation's users
241 | beyond what the individual works permit. Inclusion of a covered work
242 | in an aggregate does not cause this License to apply to the other
243 | parts of the aggregate.
244 |
245 | 6. Conveying Non-Source Forms.
246 |
247 | You may convey a covered work in object code form under the terms
248 | of sections 4 and 5, provided that you also convey the
249 | machine-readable Corresponding Source under the terms of this License,
250 | in one of these ways:
251 |
252 | a) Convey the object code in, or embodied in, a physical product
253 | (including a physical distribution medium), accompanied by the
254 | Corresponding Source fixed on a durable physical medium
255 | customarily used for software interchange.
256 |
257 | b) Convey the object code in, or embodied in, a physical product
258 | (including a physical distribution medium), accompanied by a
259 | written offer, valid for at least three years and valid for as
260 | long as you offer spare parts or customer support for that product
261 | model, to give anyone who possesses the object code either (1) a
262 | copy of the Corresponding Source for all the software in the
263 | product that is covered by this License, on a durable physical
264 | medium customarily used for software interchange, for a price no
265 | more than your reasonable cost of physically performing this
266 | conveying of source, or (2) access to copy the
267 | Corresponding Source from a network server at no charge.
268 |
269 | c) Convey individual copies of the object code with a copy of the
270 | written offer to provide the Corresponding Source. This
271 | alternative is allowed only occasionally and noncommercially, and
272 | only if you received the object code with such an offer, in accord
273 | with subsection 6b.
274 |
275 | d) Convey the object code by offering access from a designated
276 | place (gratis or for a charge), and offer equivalent access to the
277 | Corresponding Source in the same way through the same place at no
278 | further charge. You need not require recipients to copy the
279 | Corresponding Source along with the object code. If the place to
280 | copy the object code is a network server, the Corresponding Source
281 | may be on a different server (operated by you or a third party)
282 | that supports equivalent copying facilities, provided you maintain
283 | clear directions next to the object code saying where to find the
284 | Corresponding Source. Regardless of what server hosts the
285 | Corresponding Source, you remain obligated to ensure that it is
286 | available for as long as needed to satisfy these requirements.
287 |
288 | e) Convey the object code using peer-to-peer transmission, provided
289 | you inform other peers where the object code and Corresponding
290 | Source of the work are being offered to the general public at no
291 | charge under subsection 6d.
292 |
293 | A separable portion of the object code, whose source code is excluded
294 | from the Corresponding Source as a System Library, need not be
295 | included in conveying the object code work.
296 |
297 | A "User Product" is either (1) a "consumer product", which means any
298 | tangible personal property which is normally used for personal, family,
299 | or household purposes, or (2) anything designed or sold for incorporation
300 | into a dwelling. In determining whether a product is a consumer product,
301 | doubtful cases shall be resolved in favor of coverage. For a particular
302 | product received by a particular user, "normally used" refers to a
303 | typical or common use of that class of product, regardless of the status
304 | of the particular user or of the way in which the particular user
305 | actually uses, or expects or is expected to use, the product. A product
306 | is a consumer product regardless of whether the product has substantial
307 | commercial, industrial or non-consumer uses, unless such uses represent
308 | the only significant mode of use of the product.
309 |
310 | "Installation Information" for a User Product means any methods,
311 | procedures, authorization keys, or other information required to install
312 | and execute modified versions of a covered work in that User Product from
313 | a modified version of its Corresponding Source. The information must
314 | suffice to ensure that the continued functioning of the modified object
315 | code is in no case prevented or interfered with solely because
316 | modification has been made.
317 |
318 | If you convey an object code work under this section in, or with, or
319 | specifically for use in, a User Product, and the conveying occurs as
320 | part of a transaction in which the right of possession and use of the
321 | User Product is transferred to the recipient in perpetuity or for a
322 | fixed term (regardless of how the transaction is characterized), the
323 | Corresponding Source conveyed under this section must be accompanied
324 | by the Installation Information. But this requirement does not apply
325 | if neither you nor any third party retains the ability to install
326 | modified object code on the User Product (for example, the work has
327 | been installed in ROM).
328 |
329 | The requirement to provide Installation Information does not include a
330 | requirement to continue to provide support service, warranty, or updates
331 | for a work that has been modified or installed by the recipient, or for
332 | the User Product in which it has been modified or installed. Access to a
333 | network may be denied when the modification itself materially and
334 | adversely affects the operation of the network or violates the rules and
335 | protocols for communication across the network.
336 |
337 | Corresponding Source conveyed, and Installation Information provided,
338 | in accord with this section must be in a format that is publicly
339 | documented (and with an implementation available to the public in
340 | source code form), and must require no special password or key for
341 | unpacking, reading or copying.
342 |
343 | 7. Additional Terms.
344 |
345 | "Additional permissions" are terms that supplement the terms of this
346 | License by making exceptions from one or more of its conditions.
347 | Additional permissions that are applicable to the entire Program shall
348 | be treated as though they were included in this License, to the extent
349 | that they are valid under applicable law. If additional permissions
350 | apply only to part of the Program, that part may be used separately
351 | under those permissions, but the entire Program remains governed by
352 | this License without regard to the additional permissions.
353 |
354 | When you convey a copy of a covered work, you may at your option
355 | remove any additional permissions from that copy, or from any part of
356 | it. (Additional permissions may be written to require their own
357 | removal in certain cases when you modify the work.) You may place
358 | additional permissions on material, added by you to a covered work,
359 | for which you have or can give appropriate copyright permission.
360 |
361 | Notwithstanding any other provision of this License, for material you
362 | add to a covered work, you may (if authorized by the copyright holders of
363 | that material) supplement the terms of this License with terms:
364 |
365 | a) Disclaiming warranty or limiting liability differently from the
366 | terms of sections 15 and 16 of this License; or
367 |
368 | b) Requiring preservation of specified reasonable legal notices or
369 | author attributions in that material or in the Appropriate Legal
370 | Notices displayed by works containing it; or
371 |
372 | c) Prohibiting misrepresentation of the origin of that material, or
373 | requiring that modified versions of such material be marked in
374 | reasonable ways as different from the original version; or
375 |
376 | d) Limiting the use for publicity purposes of names of licensors or
377 | authors of the material; or
378 |
379 | e) Declining to grant rights under trademark law for use of some
380 | trade names, trademarks, or service marks; or
381 |
382 | f) Requiring indemnification of licensors and authors of that
383 | material by anyone who conveys the material (or modified versions of
384 | it) with contractual assumptions of liability to the recipient, for
385 | any liability that these contractual assumptions directly impose on
386 | those licensors and authors.
387 |
388 | All other non-permissive additional terms are considered "further
389 | restrictions" within the meaning of section 10. If the Program as you
390 | received it, or any part of it, contains a notice stating that it is
391 | governed by this License along with a term that is a further
392 | restriction, you may remove that term. If a license document contains
393 | a further restriction but permits relicensing or conveying under this
394 | License, you may add to a covered work material governed by the terms
395 | of that license document, provided that the further restriction does
396 | not survive such relicensing or conveying.
397 |
398 | If you add terms to a covered work in accord with this section, you
399 | must place, in the relevant source files, a statement of the
400 | additional terms that apply to those files, or a notice indicating
401 | where to find the applicable terms.
402 |
403 | Additional terms, permissive or non-permissive, may be stated in the
404 | form of a separately written license, or stated as exceptions;
405 | the above requirements apply either way.
406 |
407 | 8. Termination.
408 |
409 | You may not propagate or modify a covered work except as expressly
410 | provided under this License. Any attempt otherwise to propagate or
411 | modify it is void, and will automatically terminate your rights under
412 | this License (including any patent licenses granted under the third
413 | paragraph of section 11).
414 |
415 | However, if you cease all violation of this License, then your
416 | license from a particular copyright holder is reinstated (a)
417 | provisionally, unless and until the copyright holder explicitly and
418 | finally terminates your license, and (b) permanently, if the copyright
419 | holder fails to notify you of the violation by some reasonable means
420 | prior to 60 days after the cessation.
421 |
422 | Moreover, your license from a particular copyright holder is
423 | reinstated permanently if the copyright holder notifies you of the
424 | violation by some reasonable means, this is the first time you have
425 | received notice of violation of this License (for any work) from that
426 | copyright holder, and you cure the violation prior to 30 days after
427 | your receipt of the notice.
428 |
429 | Termination of your rights under this section does not terminate the
430 | licenses of parties who have received copies or rights from you under
431 | this License. If your rights have been terminated and not permanently
432 | reinstated, you do not qualify to receive new licenses for the same
433 | material under section 10.
434 |
435 | 9. Acceptance Not Required for Having Copies.
436 |
437 | You are not required to accept this License in order to receive or
438 | run a copy of the Program. Ancillary propagation of a covered work
439 | occurring solely as a consequence of using peer-to-peer transmission
440 | to receive a copy likewise does not require acceptance. However,
441 | nothing other than this License grants you permission to propagate or
442 | modify any covered work. These actions infringe copyright if you do
443 | not accept this License. Therefore, by modifying or propagating a
444 | covered work, you indicate your acceptance of this License to do so.
445 |
446 | 10. Automatic Licensing of Downstream Recipients.
447 |
448 | Each time you convey a covered work, the recipient automatically
449 | receives a license from the original licensors, to run, modify and
450 | propagate that work, subject to this License. You are not responsible
451 | for enforcing compliance by third parties with this License.
452 |
453 | An "entity transaction" is a transaction transferring control of an
454 | organization, or substantially all assets of one, or subdividing an
455 | organization, or merging organizations. If propagation of a covered
456 | work results from an entity transaction, each party to that
457 | transaction who receives a copy of the work also receives whatever
458 | licenses to the work the party's predecessor in interest had or could
459 | give under the previous paragraph, plus a right to possession of the
460 | Corresponding Source of the work from the predecessor in interest, if
461 | the predecessor has it or can get it with reasonable efforts.
462 |
463 | You may not impose any further restrictions on the exercise of the
464 | rights granted or affirmed under this License. For example, you may
465 | not impose a license fee, royalty, or other charge for exercise of
466 | rights granted under this License, and you may not initiate litigation
467 | (including a cross-claim or counterclaim in a lawsuit) alleging that
468 | any patent claim is infringed by making, using, selling, offering for
469 | sale, or importing the Program or any portion of it.
470 |
471 | 11. Patents.
472 |
473 | A "contributor" is a copyright holder who authorizes use under this
474 | License of the Program or a work on which the Program is based. The
475 | work thus licensed is called the contributor's "contributor version".
476 |
477 | A contributor's "essential patent claims" are all patent claims
478 | owned or controlled by the contributor, whether already acquired or
479 | hereafter acquired, that would be infringed by some manner, permitted
480 | by this License, of making, using, or selling its contributor version,
481 | but do not include claims that would be infringed only as a
482 | consequence of further modification of the contributor version. For
483 | purposes of this definition, "control" includes the right to grant
484 | patent sublicenses in a manner consistent with the requirements of
485 | this License.
486 |
487 | Each contributor grants you a non-exclusive, worldwide, royalty-free
488 | patent license under the contributor's essential patent claims, to
489 | make, use, sell, offer for sale, import and otherwise run, modify and
490 | propagate the contents of its contributor version.
491 |
492 | In the following three paragraphs, a "patent license" is any express
493 | agreement or commitment, however denominated, not to enforce a patent
494 | (such as an express permission to practice a patent or covenant not to
495 | sue for patent infringement). To "grant" such a patent license to a
496 | party means to make such an agreement or commitment not to enforce a
497 | patent against the party.
498 |
499 | If you convey a covered work, knowingly relying on a patent license,
500 | and the Corresponding Source of the work is not available for anyone
501 | to copy, free of charge and under the terms of this License, through a
502 | publicly available network server or other readily accessible means,
503 | then you must either (1) cause the Corresponding Source to be so
504 | available, or (2) arrange to deprive yourself of the benefit of the
505 | patent license for this particular work, or (3) arrange, in a manner
506 | consistent with the requirements of this License, to extend the patent
507 | license to downstream recipients. "Knowingly relying" means you have
508 | actual knowledge that, but for the patent license, your conveying the
509 | covered work in a country, or your recipient's use of the covered work
510 | in a country, would infringe one or more identifiable patents in that
511 | country that you have reason to believe are valid.
512 |
513 | If, pursuant to or in connection with a single transaction or
514 | arrangement, you convey, or propagate by procuring conveyance of, a
515 | covered work, and grant a patent license to some of the parties
516 | receiving the covered work authorizing them to use, propagate, modify
517 | or convey a specific copy of the covered work, then the patent license
518 | you grant is automatically extended to all recipients of the covered
519 | work and works based on it.
520 |
521 | A patent license is "discriminatory" if it does not include within
522 | the scope of its coverage, prohibits the exercise of, or is
523 | conditioned on the non-exercise of one or more of the rights that are
524 | specifically granted under this License. You may not convey a covered
525 | work if you are a party to an arrangement with a third party that is
526 | in the business of distributing software, under which you make payment
527 | to the third party based on the extent of your activity of conveying
528 | the work, and under which the third party grants, to any of the
529 | parties who would receive the covered work from you, a discriminatory
530 | patent license (a) in connection with copies of the covered work
531 | conveyed by you (or copies made from those copies), or (b) primarily
532 | for and in connection with specific products or compilations that
533 | contain the covered work, unless you entered into that arrangement,
534 | or that patent license was granted, prior to 28 March 2007.
535 |
536 | Nothing in this License shall be construed as excluding or limiting
537 | any implied license or other defenses to infringement that may
538 | otherwise be available to you under applicable patent law.
539 |
540 | 12. No Surrender of Others' Freedom.
541 |
542 | If conditions are imposed on you (whether by court order, agreement or
543 | otherwise) that contradict the conditions of this License, they do not
544 | excuse you from the conditions of this License. If you cannot convey a
545 | covered work so as to satisfy simultaneously your obligations under this
546 | License and any other pertinent obligations, then as a consequence you may
547 | not convey it at all. For example, if you agree to terms that obligate you
548 | to collect a royalty for further conveying from those to whom you convey
549 | the Program, the only way you could satisfy both those terms and this
550 | License would be to refrain entirely from conveying the Program.
551 |
552 | 13. Use with the GNU Affero General Public License.
553 |
554 | Notwithstanding any other provision of this License, you have
555 | permission to link or combine any covered work with a work licensed
556 | under version 3 of the GNU Affero General Public License into a single
557 | combined work, and to convey the resulting work. The terms of this
558 | License will continue to apply to the part which is the covered work,
559 | but the special requirements of the GNU Affero General Public License,
560 | section 13, concerning interaction through a network will apply to the
561 | combination as such.
562 |
563 | 14. Revised Versions of this License.
564 |
565 | The Free Software Foundation may publish revised and/or new versions of
566 | the GNU General Public License from time to time. Such new versions will
567 | be similar in spirit to the present version, but may differ in detail to
568 | address new problems or concerns.
569 |
570 | Each version is given a distinguishing version number. If the
571 | Program specifies that a certain numbered version of the GNU General
572 | Public License "or any later version" applies to it, you have the
573 | option of following the terms and conditions either of that numbered
574 | version or of any later version published by the Free Software
575 | Foundation. If the Program does not specify a version number of the
576 | GNU General Public License, you may choose any version ever published
577 | by the Free Software Foundation.
578 |
579 | If the Program specifies that a proxy can decide which future
580 | versions of the GNU General Public License can be used, that proxy's
581 | public statement of acceptance of a version permanently authorizes you
582 | to choose that version for the Program.
583 |
584 | Later license versions may give you additional or different
585 | permissions. However, no additional obligations are imposed on any
586 | author or copyright holder as a result of your choosing to follow a
587 | later version.
588 |
589 | 15. Disclaimer of Warranty.
590 |
591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
599 |
600 | 16. Limitation of Liability.
601 |
602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
610 | SUCH DAMAGES.
611 |
612 | 17. Interpretation of Sections 15 and 16.
613 |
614 | If the disclaimer of warranty and limitation of liability provided
615 | above cannot be given local legal effect according to their terms,
616 | reviewing courts shall apply local law that most closely approximates
617 | an absolute waiver of all civil liability in connection with the
618 | Program, unless a warranty or assumption of liability accompanies a
619 | copy of the Program in return for a fee.
620 |
621 | END OF TERMS AND CONDITIONS
622 |
623 | How to Apply These Terms to Your New Programs
624 |
625 | If you develop a new program, and you want it to be of the greatest
626 | possible use to the public, the best way to achieve this is to make it
627 | free software which everyone can redistribute and change under these terms.
628 |
629 | To do so, attach the following notices to the program. It is safest
630 | to attach them to the start of each source file to most effectively
631 | state the exclusion of warranty; and each file should have at least
632 | the "copyright" line and a pointer to where the full notice is found.
633 |
634 |
635 | Copyright (C)
636 |
637 | This program is free software: you can redistribute it and/or modify
638 | it under the terms of the GNU General Public License as published by
639 | the Free Software Foundation, either version 3 of the License, or
640 | (at your option) any later version.
641 |
642 | This program is distributed in the hope that it will be useful,
643 | but WITHOUT ANY WARRANTY; without even the implied warranty of
644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
645 | GNU General Public License for more details.
646 |
647 | You should have received a copy of the GNU General Public License
648 | along with this program. If not, see .
649 |
650 | Also add information on how to contact you by electronic and paper mail.
651 |
652 | If the program does terminal interaction, make it output a short
653 | notice like this when it starts in an interactive mode:
654 |
655 | Copyright (C)
656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
657 | This is free software, and you are welcome to redistribute it
658 | under certain conditions; type `show c' for details.
659 |
660 | The hypothetical commands `show w' and `show c' should show the appropriate
661 | parts of the General Public License. Of course, your program's commands
662 | might be different; for a GUI interface, you would use an "about box".
663 |
664 | You should also get your employer (if you work as a programmer) or school,
665 | if any, to sign a "copyright disclaimer" for the program, if necessary.
666 | For more information on this, and how to apply and follow the GNU GPL, see
667 | .
668 |
669 | The GNU General Public License does not permit incorporating your program
670 | into proprietary programs. If your program is a subroutine library, you
671 | may consider it more useful to permit linking proprietary applications with
672 | the library. If this is what you want to do, use the GNU Lesser General
673 | Public License instead of this License. But first, please read
674 | .
675 |
--------------------------------------------------------------------------------
/NAMESPACE:
--------------------------------------------------------------------------------
1 | # Generated by roxygen2: do not edit by hand
2 |
3 | export(bAccuracy)
4 | export(matPCs)
5 | export(multiAdaSampling)
6 | import(SingleCellExperiment)
7 | importFrom(SummarizedExperiment,assay)
8 | importFrom(e1071,svm)
9 | importFrom(methods,is)
10 | importFrom(randomForest,randomForest)
11 | importFrom(stats,median)
12 | importFrom(stats,prcomp)
13 | importFrom(stats,predict)
14 | importFrom(stats,sd)
15 |
--------------------------------------------------------------------------------
/NEWS.md:
--------------------------------------------------------------------------------
1 | ## scReCkassify 0.99.8
2 | * Included citation for the package
3 |
4 | ## scReCkassify 0.99.7
5 | * Removed global assignment in multiAdaSampling.
6 |
7 | ## scReCkassify 0.99.6
8 | * Revised further to address comments/feedbacks
9 |
10 | ## scReCkassify 0.99.5
11 | * Fixed NEWS to match the version bump
12 |
13 |
14 | ## scReCkassify 0.99.4
15 | * Minor change to example in `matPC` function
16 |
17 |
18 | ## scReCkassify 0.99.3
19 | * Correct for any spelling errors in all documentations
20 | * Revising the package to address the comments from Bioconductor review
21 | * Cleaned the GSE87795 subset data to a SingleCellExperiment object
22 |
23 |
24 | ## scReCkassify 0.99.2
25 | * version bump with submission to Bioconductor
26 |
27 | ## scReClassify 0.99.1
28 | * Cleaning repository to pass BiocCheck
29 |
30 | ## scReClassify 0.99.0
31 | * Initial submission for Bioconductor
32 | * Code formats/documentations have been revised to meet Bioconductor
33 | requirements
34 |
35 | ## scReClassify 0.1.1
36 |
37 | * Significant updates and addition to the documentations
38 | * Major changes to code writing style to comply with BioC
39 | * Updated example data
40 |
--------------------------------------------------------------------------------
/R/bAccuracy.R:
--------------------------------------------------------------------------------
1 | #' bAccuracy
2 | #'
3 | #' This function calculates the accuracy of the prediction to the true label.
4 | #'
5 | #' @param cls.truth A character vector of true class label.
6 | #' @param final A vector of final classified label prediction from
7 | #' \code{multiAdaSampling}.
8 | #' @return An accuracy value.
9 | #' @author Pengyi Yang, Taiyun Kim
10 | #'
11 | #' @examples
12 | #' data("gse87795_subset_sce")
13 | #'
14 | #' mat.expr <- gse87795_subset_sce
15 | #' cellTypes <- gse87795_subset_sce$cellTypes
16 | #'
17 | #' # Get dimension reduced matrix. We are using `logNorm` assay from `mat.expr`.
18 | #' mat.pc <- matPCs(mat.expr, assay = "logNorm")
19 | #'
20 | #' # Here we are using Support Vector Machine as a base classifier.
21 | #' result <- multiAdaSampling(mat.pc, cellTypes, classifier = "svm",
22 | #' percent = 1, L = 10)
23 | #'
24 | #' final <- result$final
25 | #'
26 | #' # Balanced accuracy
27 | #' bacc <- bAccuracy(cellTypes, final)
28 | #'
29 | #' @export bAccuracy
30 | bAccuracy <- function(cls.truth, final) {
31 |
32 | if (length(cls.truth) != length(final)) {
33 | stop("`cls.truth` and `final` objects have different length.")
34 | }
35 |
36 | gs <- names(table(cls.truth))
37 | acc <- c()
38 | acc <- unlist(lapply(seq_len(length(gs)), function(i) {
39 | sum((cls.truth == final)[cls.truth==gs[i]]) / sum(cls.truth==gs[i])
40 | }))
41 | mean(acc)
42 | }
43 |
--------------------------------------------------------------------------------
/R/data.R:
--------------------------------------------------------------------------------
1 | #' @title GSE827795 subset data
2 | #'
3 | #' @description A SingleCellExperiment object containing a subset expression
4 | #' matrix of GSE827795. The data contains log2 transformed FPKM expression.
5 | #'
6 | #' GSE87795 is a mouse fetal liver development data containing 1000 genes,
7 | #' 367 cells and 6 cell types.
8 | #'
9 | #' The original GSE87795 data and the study details can be found at this
10 | #' [link](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE87795)
11 | #'
12 | #'
13 | #' @name gse87795_subset_sce
14 | #' @docType data
15 | #'
16 | "gse87795_subset_sce"
17 |
18 |
19 |
20 | #' scReClassify: a package for post hoc cell type classification of single-cell
21 | #' RNA-sequencing data.
22 | #'
23 | #' @description A post hoc cell type classification tool to fine-tune cell type
24 | #' annotations generated by any cell type classification procedure with
25 | #' semi-supervised learning algorithm AdaSampling technique.
26 | #'
27 | #' The current version of scReClassify supports Support Vector Machine and
28 | #' Random Forest as a base classifier.
29 | #'
30 | #' @author
31 | #' **Maintainer**:
32 | #'
33 | #' * Taiyun Kim (ORCID:0000-0002-5028-836X)
34 | #' - Email: taiyun.kim91@gmail.com
35 | #'
36 | #' Authors:
37 | #'
38 | #' * Pengyi Yang (ORCID: 0000-0003-1098-3138)
39 | #'
40 | #'
41 | #' @seealso
42 | #'
43 | #' Useful links:
44 | #' * Vignette available at: https://sydneybiox.github.io/scdney/
45 | #'
46 | #' @docType package
47 | #' @name scReClassify
48 | NULL
49 |
50 |
--------------------------------------------------------------------------------
/R/matPCs.R:
--------------------------------------------------------------------------------
1 | #' matPCs function
2 | #'
3 | #' @description Performs PCA on a given matrix and returns a dimension reduced
4 | #' matrix which captures at least 80% (default) of overall variability.
5 | #'
6 | #' @param data An expression matrix or a SingleCellExperiment object.
7 | #' @param assay An assay to select if \code{data} is a SingleCellExperiment
8 | #' object
9 | #' @param percentVar The percentage of variance threshold. This is used to
10 | #' select number of Principal Components.
11 | #' @details This function performs PCA to reduce the dimension of the gene
12 | #' expression matrix limited from 10 to 20 PCs.
13 | #' @return Dimensionally reduced matrix.
14 | #' @author Pengyi Yang, Taiyun Kim
15 | #' @export matPCs
16 | #' @examples
17 | #' data("gse87795_subset_sce")
18 | #'
19 | #' mat.expr <- gse87795_subset_sce
20 | #'
21 | #' mat.pc <- matPCs(mat.expr, assay = "logNorm")
22 | #'
23 | #' # to capture at least 70% of overall variability in the dataset,
24 | #' mat.dim.reduct.70 <- matPCs(mat.expr, assay = "logNorm", 0.7)
25 | #'
26 | #' @importFrom stats prcomp sd
27 | #' @importFrom SummarizedExperiment assay
28 | #' @importFrom methods is
29 | #' @import SingleCellExperiment
30 | matPCs <- function(data, assay = NULL, percentVar=0.8) {
31 |
32 | if (percentVar < 0 | percentVar > 1) {
33 | stop("percentVar must be a number between 0 and 1")
34 | }
35 | # If SCE
36 | if (is(data, "SingleCellExperiment")) {
37 | if (is.null(assay)) {
38 | stop("assay parameter cannot be NULL")
39 | }
40 | mat <- SummarizedExperiment::assay(data, assay)
41 | } else {
42 | mat <- data
43 | }
44 |
45 | if (is(data, "matrix") | is(data, "data.frame")) {
46 | pca <- stats::prcomp(t(mat), center = TRUE, scale. = TRUE)
47 | } else if ((is(data, "SingleCellExperiment")) &
48 | ("PCA" %in% reducedDimNames(data))) {
49 | pca <- SingleCellExperiment::reducedDim(data, "PCA")
50 | } else if (is(data, "SingleCellExperiment")) {
51 | pca <- stats::prcomp(t(mat), center = TRUE, scale. = TRUE)
52 | if (length(SingleCellExperiment::reducedDimNames(data))) {
53 | rd <- SingleCellExperiment::reducedDims(data)
54 | rd$PCA = pca$x
55 | } else {
56 | SingleCellExperiment::reducedDims(data) <- list(PCA = pca$x)
57 | }
58 | }
59 |
60 | # genes are rows, cells are cols
61 | pcs <- c()
62 | sdev <- apply(pca$x, 2, sd)
63 | eigs <- sdev^2
64 | top <- which(cumsum(eigs)/sum(eigs) > percentVar)[1]
65 | m <- ifelse(top > 20, 20, top)
66 | m <- ifelse(m < 10, 10, m)
67 | pcs <- pca$x[,seq_len(m)]
68 |
69 | return(pcs)
70 | }
71 |
--------------------------------------------------------------------------------
/R/multiAdaSampling.R:
--------------------------------------------------------------------------------
1 | #' multi Adaptive Sampling function
2 | #'
3 | #' Performs multiple adaptive sampling to train a classifier model.
4 | #'
5 | #' @param data A dimension reduced matrix from \code{matPCs}.
6 | #' @param label A vector of label information for each sample.
7 | #' @param reducedDimName A name of the \code{reducedDim} to use. This must be
8 | #' specified if \code{data} is a SingleCellExperiment object.
9 | #' @param classifier Base classifier model, either "SVM" (\code{svm}) or "RF"
10 | #' \code{'rf'} is supported.
11 | #' @param percent Percentage of samples to select at each iteration.
12 | #' @param L Number of ensembles. Default to 10.
13 | #' @param prob logical flag to return sample's probabilities to each class.
14 | #' @param balance logical flag to if the cell types are balanced.
15 | #' If `FALSE`, down sample large cell types classes to the median of all class
16 | #' sizes.
17 | #' @param iter A number of iterations to perform adaSampling.
18 | #' @return A final prediction, probabilities for each cell type and the model
19 | #' are returned as a list.
20 | #'
21 | #' @author Pengyi Yang, Taiyun Kim
22 | #' @importFrom randomForest randomForest
23 | #' @importFrom e1071 svm
24 | #' @importFrom stats median predict
25 | #' @import SingleCellExperiment
26 | #' @examples
27 | #'
28 | #' library(SingleCellExperiment)
29 | #'
30 | #' # Loading the data
31 | #' data("gse87795_subset_sce")
32 | #'
33 | #' mat.expr <- gse87795_subset_sce
34 | #' cellTypes <- gse87795_subset_sce$cellTypes
35 | #'
36 | #' # Get dimension reduced matrix. We are using `logNorm` assay from `mat.expr`.
37 | #' reducedDim(mat.expr, "matPCs") <- matPCs(mat.expr, assay = "logNorm")
38 | #'
39 | #' # Here we are using Support Vector Machine as a base classifier.
40 | #' result <- multiAdaSampling(mat.expr, cellTypes, reducedDimName = "matPCs",
41 | #' classifier = "svm", percent = 1, L = 10)
42 | #' @export multiAdaSampling
43 | multiAdaSampling <- function(data, label, reducedDimName = NULL,
44 | classifier="svm", percent=1, L=10, prob=FALSE,
45 | balance=TRUE, iter=3) {
46 | # If SCE
47 | if (is(data, "SingleCellExperiment")) {
48 | if (is.null(reducedDimName)) {
49 | stop("reducedDimName parameter cannot be NULL if data is a",
50 | " SingleCellExperiment object")
51 | }
52 | mat <- SingleCellExperiment::reducedDim(data, reducedDimName)
53 | } else {
54 | mat <- data
55 | }
56 | mat <- t(mat)
57 |
58 | if (ncol(mat) != length(label)) {
59 | stop("Parameter `label` and number of columns of data must be equal")
60 | }
61 | if (percent < 0 | percent > 1) {
62 | stop("Parameter percent cannot must be a value between 0 and 1")
63 | }
64 |
65 | models <- lapply(seq_len(L), function(l) {
66 | X <- mat
67 | Y <- label
68 |
69 | model <- c()
70 | prob.mat <- c()
71 |
72 | for (i in seq_len(iter)) {
73 | if (classifier == "rf") {
74 | model <- randomForest::randomForest(t(X), factor(Y),
75 | ntree = 100)
76 | prob.mat <- stats::predict(model, newdata=t(mat), type="prob")
77 | }
78 | if (classifier == "svm") {
79 | tmp <- t(X)
80 | rownames(tmp) <- NULL
81 | model <- e1071::svm(tmp, factor(Y), probability = TRUE)
82 | prob.mat <- attr(stats::predict(model, t(mat),
83 | decision.values = FALSE,
84 | probability = TRUE),
85 | "probabilities")
86 | }
87 | if (!(classifier %in% c("svm", "rf"))) {
88 | stop("Classifier ", classifier, " is unknown.")
89 | }
90 |
91 | X <- c()
92 | Y <- c()
93 |
94 | XY_update <- lapply(seq_len(ncol(prob.mat)), function(j) {
95 | voteClass <- prob.mat[label==colnames(prob.mat)[j],]
96 | idx <- c()
97 | if (balance) {
98 | idx <- sample(seq_len(nrow(voteClass)),
99 | size=nrow(voteClass)*percent, replace = TRUE,
100 | prob=voteClass[,j])
101 | } else {
102 | sampleSize <- round(stats::median(table(label)))
103 | if (nrow(voteClass) > sampleSize) {
104 | idx <- sample(seq_len(nrow(voteClass)), size=sampleSize*
105 | percent, replace = TRUE,
106 | prob=voteClass[,j])
107 | } else {
108 | idx <- sample(seq_len(nrow(voteClass)),
109 | size=nrow(voteClass)*percent,
110 | replace = TRUE, prob=voteClass[,j])
111 | }
112 | }
113 |
114 | list(
115 | X = mat[, rownames(voteClass)[idx]],
116 | Y = label[rownames(voteClass)[idx]]
117 | )
118 | })
119 | cur_X = lapply(XY_update, function(xy) {
120 | xy$X
121 | })
122 | cur_Y = lapply(XY_update, function(xy) {
123 | xy$Y
124 | })
125 | X <- do.call(cbind, cur_X)
126 | Y <- do.call(c, cur_Y)
127 | }
128 |
129 | model
130 | })
131 |
132 | predictMat <- matrix(0, nrow=ncol(mat), ncol=length(table(label)))
133 | final <- c()
134 | predmat <- lapply(models, function(m) {
135 | if (classifier == "svm") {
136 | tmp <- attr(predict(m,
137 | newdata=t(mat), probability = TRUE),
138 | "prob")[,names(table(label))]
139 | } else {
140 | tmp <- predict(m, newdata=t(mat),
141 | type="prob")[,names(table(label))]
142 | }
143 | predictMat <<- predictMat + tmp
144 | })
145 |
146 | if(prob==TRUE) {
147 | final <- apply(predictMat, 1, max)
148 | names(final) <- names(table(label))[apply(predictMat, 1, which.max)]
149 | } else {
150 | final <- names(table(label))[apply(predictMat, 1, which.max)]
151 | }
152 |
153 | return(
154 | list(
155 | final = final,
156 | models = models,
157 | prob = predictMat
158 | )
159 | )
160 | }
161 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 |
2 | # scReClassify
3 |
4 | `scReClassify` is a post hoc cell type classification of single-cell
5 | RNA-sequencing data. Using semi-supervised learning algorithm,
6 | adaSampling, to correct cell type annotation from noise.
7 |
8 | # Getting started
9 |
10 | ## Vignette
11 |
12 | For more detailed instuction, you can find the vignette at our website:
13 | .
14 |
15 | ### Installation
16 |
17 | #### Bioconductor
18 |
19 | ```r
20 | if (!requireNamespace("BiocManager", quietly = TRUE))
21 | install.packages("BiocManager")
22 |
23 | BiocManager::install("scReClassify")
24 | ```
25 |
26 | #### GitHub
27 |
28 | ``` r
29 | devtools::install_github("SydneyBioX/scReClassify", build_opts = c("--no-resave-data", "--no-manual"))
30 | library(scReClassify)
31 | ```
32 |
33 | For `devtool version
34 | < 2.0.0`,
35 |
36 | ``` r
37 | devtools::install_github("SydneyBioX/scReClassify", build_vignettes = TRUE)
38 | library(scReClassify)
39 | ```
40 |
41 |
42 | ``` r
43 | suppressPackageStartupMessages({
44 | library(scReClassify)
45 | library(SingleCellExperiment)
46 | library(SummarizedExperiment)
47 | })
48 | ```
49 |
50 |
51 | # Usage
52 |
53 | 
54 |
55 |
56 | Current version of this package is implemented to run with `svm` and `randomForest` classifiers.
57 |
58 | ## Load data
59 |
60 | ``` r
61 | data(gse87795_subset_sce)
62 | dat <- gse87795_subset_sce
63 | cellTypes <- gse87795_subset_sce$cellTypes
64 |
65 | # number of clusters
66 | nCs <- length(table(cellTypes))
67 |
68 | # This demo dataset is already pre-processed
69 | dim(dat)
70 | ```
71 |
72 | ## Part A. scReClassify (Demonstration)
73 |
74 | ### Dimension reduction
75 |
76 | ``` r
77 | reducedDim(dat, "matPCs") = matPCs(dat, assay = "logNorm", 0.7)
78 | ```
79 |
80 | ### Synthetic noise (Demonstration purpose)
81 |
82 | Here in this example, we will synthetically generate varying degree of noise in sample labels.
83 |
84 | ``` r
85 | lab <- cellTypes
86 |
87 | set.seed(1)
88 | noisyCls <- function(dat, rho, cls.truth){
89 | cls.noisy <- cls.truth
90 | names(cls.noisy) <- colnames(dat)
91 | for(i in seq_len(length(table(cls.noisy)))) {
92 | # class label starts from 0
93 | if (i != length(table(cls.noisy))) {
94 | cls.noisy[sample(which(cls.truth == names(table(cls.noisy))[i]), floor(sum(cls.truth == names(table(cls.noisy))[i]) * rho))] <- names(table(cls.noisy))[i+1]
95 | } else {
96 | cls.noisy[sample(which(cls.truth == names(table(cls.noisy))[i]), floor(sum(cls.truth == names(table(cls.noisy))[i]) * rho))] <- names(table(cls.noisy))[1]
97 | }
98 | }
99 |
100 | print(sum(cls.truth != cls.noisy))
101 | return(cls.noisy)
102 | }
103 |
104 | cls.noisy01 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.1, lab)
105 | cls.noisy02 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.2, lab)
106 | cls.noisy03 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.3, lab)
107 | cls.noisy04 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.4, lab)
108 | cls.noisy05 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.5, lab)
109 | ```
110 |
111 | ### Use scReClassify to correct mislabeled cell types.
112 |
113 | Here in this example, we will only use `Support Vector machine (svm)` as base classifier.
114 |
115 | #### Benchmark evaluation
116 |
117 | ``` r
118 | ###################################
119 | # SVM
120 | ###################################
121 | base <- "svm"
122 | set.seed(1)
123 | result = lapply(seq_len(10), function(j) {
124 | final <- multiAdaSampling(dat, cls.noisy01, reducedDimName = "matPCs",
125 | classifier=base, percent=1, L=10)$final
126 | ari01 <- mclust::adjustedRandIndex(lab, final)
127 | acc01 <- bAccuracy(lab, final)
128 |
129 | final <- multiAdaSampling(dat, cls.noisy02, reducedDimName = "matPCs",
130 | classifier=base, percent=1, L=10)$final
131 | ari02 <- mclust::adjustedRandIndex(lab, final)
132 | acc02 <- bAccuracy(lab, final)
133 |
134 | final <- multiAdaSampling(dat, cls.noisy03, reducedDimName = "matPCs",
135 | classifier=base, percent=1, L=10)$final
136 | ari03 <- mclust::adjustedRandIndex(lab, final)
137 | acc03 <- bAccuracy(lab, final)
138 |
139 | final <- multiAdaSampling(dat, cls.noisy04, reducedDimName = "matPCs",
140 | classifier=base, percent=1, L=10)$final
141 | ari04 <- mclust::adjustedRandIndex(lab, final)
142 | acc04 <- bAccuracy(lab, final)
143 |
144 | final <- multiAdaSampling(dat, cls.noisy05, reducedDimName = "matPCs",
145 | classifier=base, percent=1, L=10)$final
146 | ari05 <- mclust::adjustedRandIndex(lab, final)
147 | acc05 <- bAccuracy(lab, final)
148 |
149 | c(
150 | acc01 = acc01,
151 | acc02 = acc02,
152 | acc03 = acc03,
153 | acc04 = acc04,
154 | acc05 = acc05,
155 | ari01 = ari01,
156 | ari02 = ari02,
157 | ari03 = ari03,
158 | ari04 = ari04,
159 | ari05 = ari05
160 | )
161 | })
162 |
163 | result = do.call(rbind, result)
164 | acc = result[,seq_len(5)]
165 | colnames(acc) = seq(from=0.1,to=0.5,by=0.1)
166 |
167 | ari = result[,seq(from= 6, to = 10)]
168 | colnames(ari) = seq(from=0.1,to=0.5,by=0.1)
169 |
170 |
171 | plot.new()
172 | par(mfrow = c(1,2))
173 | boxplot(acc, col="lightblue", main="SVM Accuracy",
174 | ylim=c(0.45, 1), xlab = "rho", ylab = "Accuracy")
175 | points(x=seq_len(5), y=c(
176 | bAccuracy(lab, cls.noisy01),
177 | bAccuracy(lab, cls.noisy02),
178 | bAccuracy(lab, cls.noisy03),
179 | bAccuracy(lab, cls.noisy04),
180 | bAccuracy(lab, cls.noisy05)),
181 | col="red3", pch=c(2,3,4,5,6), cex=1)
182 | boxplot(ari, col="lightblue", main="SVM ARI",
183 | ylim=c(0.25, 1), xlab = "rho", ylab = "ARI")
184 | points(x=seq_len(5), y=c(
185 | mclust::adjustedRandIndex(lab, cls.noisy01),
186 | mclust::adjustedRandIndex(lab, cls.noisy02),
187 | mclust::adjustedRandIndex(lab, cls.noisy03),
188 | mclust::adjustedRandIndex(lab, cls.noisy04),
189 | mclust::adjustedRandIndex(lab, cls.noisy05)),
190 | col="red3", pch=c(2,3,4,5,6), cex=1)
191 | ```
192 |
193 |
194 | ## Part B. scReClassify (mislabeled cell type correction)
195 |
196 | ``` r
197 | # PCA procedure
198 | reducedDim(dat, "matPCs") = matPCs(dat, assay = "logNorm", 0.7)
199 |
200 |
201 | # run scReClassify
202 | set.seed(1)
203 | cellTypes.reclassify <- multiAdaSampling(dat, cellTypes,
204 | reducedDimName = "matPCs",
205 | classifier = "svm", percent = 1, L = 10)
206 |
207 | # Verification by marker genes
208 | End <- c("ITGA2B", "ITGB3")
209 |
210 | # check examples
211 | idx <- which(cellTypes.reclassify$final != cellTypes)
212 |
213 | cbind(original=cellTypes[idx], reclassify=cellTypes.reclassify$final[idx]) %>%
214 | DT::datatable()
215 |
216 | mat <- assay(dat, "logNorm")
217 |
218 | c1 <- mat[, which(cellTypes=="Endothelial Cell")]
219 | c2 <- mat[, which(cellTypes=="Erythrocyte")]
220 | c3 <- mat[, which(cellTypes=="Hepatoblast")]
221 | c4 <- mat[, which(cellTypes=="Macrophage")]
222 | c5 <- mat[, which(cellTypes=="Megakaryocyte")]
223 | c6 <- mat[, which(cellTypes=="Mesenchymal Cell")]
224 | cs <- rainbow(length(table(cellTypes)))
225 |
226 |
227 | # (example 1 E13.5_C14)
228 | #####
229 | par(mfrow=c(1,2))
230 | marker <- End[1]
231 | boxplot(c1[marker,], c2[marker,], c3[marker,],
232 | c4[marker,], c5[marker,], c6[marker,],
233 | col=cs, main=marker,
234 | names=c("Others", "Others", "Others", "Orignal",
235 | "Reclassified", "Others"), las=2, xlab = "Labels",
236 | ylab = "log2FPKM")
237 | points(5, mat[marker, which(colnames(mat) %in% "E13.5_C14")],
238 | pch=16, col="red", cex=2)
239 |
240 | marker <- End[2]
241 | boxplot(c1[marker,], c2[marker,], c3[marker,],
242 | c4[marker,], c5[marker,], c6[marker,],
243 | col=cs, main=marker,
244 | names=c("Others", "Others", "Others", "Orignal",
245 | "Reclassified", "Others"), las=2, xlab = "Labels",
246 | ylab = "log2FPKM")
247 | points(5, mat[marker, which(colnames(mat) %in% "E13.5_C14")],
248 | pch=16, col="red", cex=2)
249 | ```
250 |
251 |
252 | # Reference
253 | scReClassify is published in BMC Genomics. Please refer to the following article for more details on method implementation and evaluation.
254 |
255 | Taiyun Kim, Kitty Lo, Thomas A. Geddes, Hani Jieun Kim, Jean Yee Hwa Yang & Pengyi Yang (2019) scReClassify: post hoc cell type classification of single-cell RNA-seq data. BMC Genomics, 20:913. https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-019-6305-x
256 |
--------------------------------------------------------------------------------
/data/gse87795_subset_sce.rda:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/SydneyBioX/scReClassify/679abbb63ce3c141f2638b35a563e21b8a51ce4d/data/gse87795_subset_sce.rda
--------------------------------------------------------------------------------
/img/scReClassify.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/SydneyBioX/scReClassify/679abbb63ce3c141f2638b35a563e21b8a51ce4d/img/scReClassify.jpg
--------------------------------------------------------------------------------
/img/scReClassify_sticker.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/SydneyBioX/scReClassify/679abbb63ce3c141f2638b35a563e21b8a51ce4d/img/scReClassify_sticker.png
--------------------------------------------------------------------------------
/inst/CITATION:
--------------------------------------------------------------------------------
1 | citHeader("To cite scReClassify in publications use:")
2 | citEntry(entry = "Article",
3 | title = "scReClassify: post hoc cell type classification of single-cell RNA-seq data",
4 | author = personList(as.person("Taiyun Kim"),
5 | as.person("Kitty Lo"),
6 | as.person("Thomas A Geddes"),
7 | as.person("Hani Jieun Kim"),
8 | as.person("Jean Yee Hwa Yang"),
9 | as.person("Pengyi Yang")),
10 | journal = "BMC Genomics",
11 | year = "2019",
12 | volume = "20",
13 | number = "913",
14 | url = "https://doi.org/10.1186/s12864-019-6305-x",
15 | textVersion =
16 | paste("Kim, T., Lo, K., Geddes, T. A., Kim, H. J., Yang, J. Y. H., Yang, P.",
17 | "scReClassify: post hoc cell type classification of single-cell RNA-seq data.",
18 | "BMC Genomics, 20, 913 (2019)",
19 | "doi: https://doi.org/10.1186/s12864-019-6305-x")
20 | )
--------------------------------------------------------------------------------
/inst/scReClassify.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/SydneyBioX/scReClassify/679abbb63ce3c141f2638b35a563e21b8a51ce4d/inst/scReClassify.jpg
--------------------------------------------------------------------------------
/inst/scReClassify_sticker.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/SydneyBioX/scReClassify/679abbb63ce3c141f2638b35a563e21b8a51ce4d/inst/scReClassify_sticker.png
--------------------------------------------------------------------------------
/man/bAccuracy.Rd:
--------------------------------------------------------------------------------
1 | % Generated by roxygen2: do not edit by hand
2 | % Please edit documentation in R/bAccuracy.R
3 | \name{bAccuracy}
4 | \alias{bAccuracy}
5 | \title{bAccuracy}
6 | \usage{
7 | bAccuracy(cls.truth, final)
8 | }
9 | \arguments{
10 | \item{cls.truth}{A character vector of true class label.}
11 |
12 | \item{final}{A vector of final classified label prediction from
13 | \code{multiAdaSampling}.}
14 | }
15 | \value{
16 | An accuracy value.
17 | }
18 | \description{
19 | This function calculates the accuracy of the prediction to the true label.
20 | }
21 | \examples{
22 | data("gse87795_subset_sce")
23 |
24 | mat.expr <- gse87795_subset_sce
25 | cellTypes <- gse87795_subset_sce$cellTypes
26 |
27 | # Get dimension reduced matrix. We are using `logNorm` assay from `mat.expr`.
28 | mat.pc <- matPCs(mat.expr, assay = "logNorm")
29 |
30 | # Here we are using Support Vector Machine as a base classifier.
31 | result <- multiAdaSampling(mat.pc, cellTypes, classifier = "svm",
32 | percent = 1, L = 10)
33 |
34 | final <- result$final
35 |
36 | # Balanced accuracy
37 | bacc <- bAccuracy(cellTypes, final)
38 |
39 | }
40 | \author{
41 | Pengyi Yang, Taiyun Kim
42 | }
43 |
--------------------------------------------------------------------------------
/man/gse87795_subset_sce.Rd:
--------------------------------------------------------------------------------
1 | % Generated by roxygen2: do not edit by hand
2 | % Please edit documentation in R/data.R
3 | \docType{data}
4 | \name{gse87795_subset_sce}
5 | \alias{gse87795_subset_sce}
6 | \title{GSE827795 subset data}
7 | \format{
8 | An object of class \code{SingleCellExperiment} with 1000 rows and 367 columns.
9 | }
10 | \usage{
11 | gse87795_subset_sce
12 | }
13 | \description{
14 | A SingleCellExperiment object containing a subset expression
15 | matrix of GSE827795. The data contains log2 transformed FPKM expression.
16 |
17 | GSE87795 is a mouse fetal liver development data containing 1000 genes,
18 | 367 cells and 6 cell types.
19 |
20 | The original GSE87795 data and the study details can be found at this
21 | \href{https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE87795}{link}
22 | }
23 | \keyword{datasets}
24 |
--------------------------------------------------------------------------------
/man/matPCs.Rd:
--------------------------------------------------------------------------------
1 | % Generated by roxygen2: do not edit by hand
2 | % Please edit documentation in R/matPCs.R
3 | \name{matPCs}
4 | \alias{matPCs}
5 | \title{matPCs function}
6 | \usage{
7 | matPCs(data, assay = NULL, percentVar = 0.8)
8 | }
9 | \arguments{
10 | \item{data}{An expression matrix or a SingleCellExperiment object.}
11 |
12 | \item{assay}{An assay to select if \code{data} is a SingleCellExperiment
13 | object}
14 |
15 | \item{percentVar}{The percentage of variance threshold. This is used to
16 | select number of Principal Components.}
17 | }
18 | \value{
19 | Dimensionally reduced matrix.
20 | }
21 | \description{
22 | Performs PCA on a given matrix and returns a dimension reduced
23 | matrix which captures at least 80\% (default) of overall variability.
24 | }
25 | \details{
26 | This function performs PCA to reduce the dimension of the gene
27 | expression matrix limited from 10 to 20 PCs.
28 | }
29 | \examples{
30 | data("gse87795_subset_sce")
31 |
32 | mat.expr <- gse87795_subset_sce
33 |
34 | mat.pc <- matPCs(mat.expr, assay = "logNorm")
35 |
36 | # to capture at least 70\% of overall variability in the dataset,
37 | mat.dim.reduct.70 <- matPCs(mat.expr, assay = "logNorm", 0.7)
38 |
39 | }
40 | \author{
41 | Pengyi Yang, Taiyun Kim
42 | }
43 |
--------------------------------------------------------------------------------
/man/multiAdaSampling.Rd:
--------------------------------------------------------------------------------
1 | % Generated by roxygen2: do not edit by hand
2 | % Please edit documentation in R/multiAdaSampling.R
3 | \name{multiAdaSampling}
4 | \alias{multiAdaSampling}
5 | \title{multi Adaptive Sampling function}
6 | \usage{
7 | multiAdaSampling(
8 | data,
9 | label,
10 | reducedDimName = NULL,
11 | classifier = "svm",
12 | percent = 1,
13 | L = 10,
14 | prob = FALSE,
15 | balance = TRUE,
16 | iter = 3
17 | )
18 | }
19 | \arguments{
20 | \item{data}{A dimension reduced matrix from \code{matPCs}.}
21 |
22 | \item{label}{A vector of label information for each sample.}
23 |
24 | \item{reducedDimName}{A name of the \code{reducedDim} to use. This must be
25 | specified if \code{data} is a SingleCellExperiment object.}
26 |
27 | \item{classifier}{Base classifier model, either "SVM" (\code{svm}) or "RF"
28 | \code{'rf'} is supported.}
29 |
30 | \item{percent}{Percentage of samples to select at each iteration.}
31 |
32 | \item{L}{Number of ensembles. Default to 10.}
33 |
34 | \item{prob}{logical flag to return sample's probabilities to each class.}
35 |
36 | \item{balance}{logical flag to if the cell types are balanced.
37 | If \code{FALSE}, down sample large cell types classes to the median of all class
38 | sizes.}
39 |
40 | \item{iter}{A number of iterations to perform adaSampling.}
41 | }
42 | \value{
43 | A final prediction, probabilities for each cell type and the model
44 | are returned as a list.
45 | }
46 | \description{
47 | Performs multiple adaptive sampling to train a classifier model.
48 | }
49 | \examples{
50 |
51 | library(SingleCellExperiment)
52 |
53 | # Loading the data
54 | data("gse87795_subset_sce")
55 |
56 | mat.expr <- gse87795_subset_sce
57 | cellTypes <- gse87795_subset_sce$cellTypes
58 |
59 | # Get dimension reduced matrix. We are using `logNorm` assay from `mat.expr`.
60 | reducedDim(mat.expr, "matPCs") <- matPCs(mat.expr, assay = "logNorm")
61 |
62 | # Here we are using Support Vector Machine as a base classifier.
63 | result <- multiAdaSampling(mat.expr, cellTypes, reducedDimName = "matPCs",
64 | classifier = "svm", percent = 1, L = 10)
65 | }
66 | \author{
67 | Pengyi Yang, Taiyun Kim
68 | }
69 |
--------------------------------------------------------------------------------
/man/scReClassify.Rd:
--------------------------------------------------------------------------------
1 | % Generated by roxygen2: do not edit by hand
2 | % Please edit documentation in R/data.R
3 | \docType{package}
4 | \name{scReClassify}
5 | \alias{scReClassify}
6 | \title{scReClassify: a package for post hoc cell type classification of single-cell
7 | RNA-sequencing data.}
8 | \description{
9 | A post hoc cell type classification tool to fine-tune cell type
10 | annotations generated by any cell type classification procedure with
11 | semi-supervised learning algorithm AdaSampling technique.
12 |
13 | The current version of scReClassify supports Support Vector Machine and
14 | Random Forest as a base classifier.
15 | }
16 | \seealso{
17 | Useful links:
18 | \itemize{
19 | \item Vignette available at: https://sydneybiox.github.io/scdney/
20 | }
21 | }
22 | \author{
23 | \strong{Maintainer}:
24 | \itemize{
25 | \item Taiyun Kim (ORCID:0000-0002-5028-836X)
26 | \itemize{
27 | \item Email: taiyun.kim91@gmail.com
28 | }
29 | }
30 |
31 | Authors:
32 | \itemize{
33 | \item Pengyi Yang (ORCID: 0000-0003-1098-3138)
34 | }
35 | }
36 |
--------------------------------------------------------------------------------
/tests/testthat.R:
--------------------------------------------------------------------------------
1 | library(testthat)
2 | library(scReClassify)
3 |
4 | test_check("scReClassify")
5 |
--------------------------------------------------------------------------------
/tests/testthat/test-bAccuracy.R:
--------------------------------------------------------------------------------
1 | test_that(
2 | "Case bAccuracy.1: Missing parameters",
3 | {
4 | cls = 1:12
5 | label = 1:12
6 |
7 | expect_error(bAccuracy())
8 | expect_error(bAccuracy(cls.truth = cls))
9 | expect_error(bAccuracy(final = label))
10 | }
11 | )
12 |
13 |
14 | test_that(
15 | "Case bAccuracy.2: Different parameter lengths",
16 | {
17 | cls = 1:12
18 | label = 1:10
19 |
20 | expect_error(bAccuracy(cls, label))
21 | }
22 | )
23 |
24 | test_that(
25 | "Case bAccuracy.3: Expected outputs",
26 | {
27 | cls = 1:12
28 | label = 12:1
29 | exp_result1 = 0
30 | exp_result2 = 1
31 |
32 | expect_identical(exp_result1, bAccuracy(cls, label))
33 | expect_identical(exp_result2, bAccuracy(cls, cls))
34 |
35 | }
36 | )
37 |
--------------------------------------------------------------------------------
/tests/testthat/test-matPCs.R:
--------------------------------------------------------------------------------
1 | test_that(
2 | "Case matPCs.1: Missing parameters",
3 | {
4 | expect_error(matPCs())
5 | expect_error(matPCs(percentVar = 0.8))
6 | }
7 | )
8 |
9 | test_that(
10 | "Case matPCs.2: Expected outputs",
11 | {
12 | set.seed(123)
13 | # 50 cells 400 genes
14 | mat = matrix(rnorm(20000), nrow = 50)
15 | pc = prcomp(mat)
16 | eigs = pc$sdev^2
17 | p = cumsum(eigs)/sum(eigs)
18 |
19 | result1 = min(which(p > 0.5))
20 | result2 = min(which(p > 0.8))
21 | # Since result2 > 20
22 | result2 = 20L
23 |
24 | # Number of PCs returned
25 | expect_identical(result1, ncol(matPCs(t(mat), percentVar = 0.5)))
26 | expect_identical(result2, ncol(matPCs(t(mat), percentVar = 0.8)))
27 | }
28 | )
29 |
--------------------------------------------------------------------------------
/tests/testthat/test-multiAdaSampling.R:
--------------------------------------------------------------------------------
1 | test_that(
2 | "Case multiAdaSampling.1: Missing parameters",
3 | {
4 | dat = matrix(rnorm(1200), nrow= 12)
5 | label = 1:ncol(dat)
6 | expect_error(multiAdaSampling())
7 | expect_error(multiAdaSampling(data = dat))
8 | expect_error(multiAdaSampling(label = label))
9 | }
10 | )
11 |
12 | test_that(
13 | "Case multiAdaSampling.2: Expected output",
14 | {
15 | set.seed(123)
16 | # very basic data with strong signal with 2 classes
17 | dat = matrix(rnorm(1200), ncol = 12)
18 | colnames(dat) = paste0("Cell", seq_len(12))
19 | rownames(dat) = paste0("Gene", seq_len(100))
20 | label = rep(c("CT_A", "CT_B"), each = 6)
21 | names(label) = colnames(dat)
22 | noise_label = label
23 |
24 | pc = matPCs(dat, percentVar = 0.8)
25 | result = multiAdaSampling(
26 | data = pc,
27 | label = noise_label,
28 | classifier = "svm"
29 | )
30 |
31 | object_names = c("final", "models", "prob")
32 | prob_dim = c(12L, 2L) # 100 cells, 2 cell types
33 |
34 | expect_identical(object_names, names(result))
35 | expect_identical(prob_dim, dim(result$prob))
36 | expect_s3_class(result$models[[1]], "svm")
37 | expect_identical(12L, length(result$final))
38 | }
39 | )
40 |
41 |
42 |
43 |
--------------------------------------------------------------------------------
/vignettes/scReClassify.Rmd:
--------------------------------------------------------------------------------
1 | ---
2 | title: "An introduction to scReClassify package"
3 | author:
4 | - name: Taiyun Kim
5 | affiliation:
6 | - School of Mathematics and Statistics, The University of Sydney
7 | - Computational Systems Biology Group, Children’s Medical Research Institute, Faculty of Medicine and Health, The University of Sydney
8 | - Charles Perkins Centre, The University of Sydney
9 | output:
10 | BiocStyle::html_document:
11 | toc_newpage: true
12 | fig_retina: NULL
13 | package: BiocStyle
14 | vignette: >
15 | %\VignetteIndexEntry{An introduction to scReClassify package}
16 | %\VignetteEngine{knitr::rmarkdown}
17 | %\VignetteEncoding{UTF-8}
18 | ---
19 |
20 | ```{r setup, include=FALSE}
21 | knitr::opts_chunk$set(echo = TRUE)
22 | ```
23 |
24 |
25 | # Introduction
26 |
27 |
28 |
29 | `scReClassify` is a post hoc cell type classification of single-cell
30 | RNA-sequencing data to fine-tune cell type annotations generated by any cell
31 | type classification procedure. Typically, cell type identification relies on
32 | human inspection using combination of prior biological knowledge and
33 | computational techniques. Due to incompleteness of our current knowledge and the
34 | subjectivity involved in this process, a small amount of cells may be subject to
35 | mislabelling. Using semi-supervised learning algorithm, adaSampling, we are able
36 | to correct cell type annotations from various degree of noise.
37 |
38 |
39 | 
40 |
41 | # Installation
42 |
43 | Install the latest development version from GitHub using the `devtools` package:
44 |
45 | ```{r, eval = FALSE}
46 | if (!("devtools" %in% rownames(installed.packages())))
47 | install.packages("devtools")
48 |
49 | library(devtools)
50 | devtools::install_github("SydneyBioX/scReClassify")
51 | ```
52 |
53 | To install the Bioconductor version of scReClassify, enter the following to your
54 | R console.
55 |
56 | ```{r, eval = FALSE}
57 | if (!requireNamespace("BiocManager", quietly = TRUE))
58 | install.packages("BiocManager")
59 | BiocManager::install("scReClassify")
60 | ```
61 |
62 | # Loading packages and data
63 |
64 | ```{r}
65 | suppressPackageStartupMessages({
66 | library(scReClassify)
67 | library(DT)
68 | library(mclust)
69 | library(dplyr)
70 | library(SummarizedExperiment)
71 | library(SingleCellExperiment)
72 | })
73 |
74 | data("gse87795_subset_sce")
75 |
76 | dat <- gse87795_subset_sce
77 | cellTypes <- gse87795_subset_sce$cellTypes
78 | ```
79 |
80 | `gse87795_subset_sce` is a `SingleCellExperiment` object of a mouse fetal liver
81 | development data deposited at Gene Expression Omnibus respository with accession
82 | ID GSE87795. The cell type information can be found on the `colData` of the
83 | `SingleCellExperiment` object.
84 |
85 |
86 | ```{r}
87 | # Cell types
88 | table(cellTypes)
89 |
90 | # We set the number of clusters
91 | nCs <- length(table(cellTypes))
92 | nCs
93 |
94 | # This demo dataset is already pre-processed
95 | dim(dat)
96 | ```
97 |
98 | There are `r nCs` cell types, `r ncol(dat)` cells and `r nrow(dat)` number of
99 | genes.
100 |
101 |
102 | # Part A. scReClassify (Demonstration with synthetic mislabels)
103 |
104 | ## Dimension reduction
105 |
106 | Prior to running scReClassify, we perform dimension reduction. `matPCs` is a
107 | tool in scReClassify to simplify this process. In this function, a dimension
108 | reduced matrix is returned with `n` principal components (PCs), where `n` is the
109 | number of principal components (PCs) that by sum explains at least 70% variance.
110 |
111 | The function accepts either a `matrix` or a `SingleCellExperiment` object. If
112 | the `data` parameter is a `SingleCellExperiment` object, an `assay` variable
113 | must be specified to perform dimension reduction on the correct assay. If the
114 | `SingleCellExperiment` object `data` already has a 'PCA' in `reducedDimNames()`,
115 | the 'PCA' matrix of `n` columns are returned.
116 |
117 |
118 | ```{r}
119 | reducedDim(dat, "matPCs") <- matPCs(dat, assay = "logNorm", 0.7)
120 | ```
121 |
122 | ## Synthetic noise (Demonstration purpose)
123 |
124 | Here in this example, we will synthetically generate varying degree of
125 | noise (10-50%) in sample labels. The purpose here is to simulate different level
126 | of mislabeling in the data. Given a cell type label `cls.truth`, `noisyCls`
127 | function will randomly select a `rho` percentage of cells from a given cell type
128 | and relabel to other cell types.
129 |
130 | Here, we create different degree of noise from 10% to 50%.
131 |
132 | ```{r}
133 | lab <- cellTypes
134 |
135 | set.seed(1)
136 | # Function to create noise in the cell type label
137 | noisyCls <- function(dat, rho, cls.truth){
138 | cls.noisy <- cls.truth
139 | names(cls.noisy) <- colnames(dat)
140 |
141 | for(i in seq_len(length(table(cls.noisy)))) {
142 | # class label starts from 0
143 | if (i != length(table(cls.noisy))) {
144 | cls.noisy[sample(which(cls.truth == names(table(cls.noisy))[i]),
145 | floor(sum(cls.truth == names(table(cls.noisy))[i])*
146 | rho))] <- names(table(cls.noisy))[i+1]
147 | } else {
148 | cls.noisy[sample(which(cls.truth == names(table(cls.noisy))[i]),
149 | floor(sum(cls.truth == names(table(cls.noisy))[i])*
150 | rho))] <- names(table(cls.noisy))[1]
151 | }
152 | }
153 |
154 | print(sum(cls.truth != cls.noisy))
155 | return(cls.noisy)
156 | }
157 |
158 | cls.noisy01 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.1, lab)
159 | cls.noisy02 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.2, lab)
160 | cls.noisy03 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.3, lab)
161 | cls.noisy04 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.4, lab)
162 | cls.noisy05 <- noisyCls(t(reducedDim(dat, "matPCs")), rho=0.5, lab)
163 | ```
164 |
165 | With `noisyCls` function, we have relabeled `r cls.noisy01`, `r cls.noisy01`,
166 | `r cls.noisy01`, `r cls.noisy01` and `r cls.noisy01` number of cells for `rho`
167 | equal to 0.1, 0.2, 0.3, 0.4 and 0.5 respectively.
168 |
169 |
170 | ## Use scReClassify to correct mislabeled cell types.
171 |
172 | Here in this example, we will only use `Support Vector machine (svm)` as
173 | base classifier.
174 |
175 | #### Benchmark evaluation
176 |
177 | To benchmark scReClassify, we perform scReclassify to all degree of noise with
178 | 10 repeats. We measure the accuracy of scReClassify and the Adjusted Rand Index
179 | (ARI) to measure the concordance of the reclassified cell type to the true cell
180 | type label.
181 |
182 | ```{r}
183 | ###################################
184 | # SVM
185 | ###################################
186 | base <- "svm"
187 | set.seed(1)
188 | result = lapply(seq_len(10), function(j) {
189 | final <- multiAdaSampling(dat, cls.noisy01, reducedDimName = "matPCs",
190 | classifier=base, percent=1, L=10)$final
191 | ari01 <- mclust::adjustedRandIndex(lab, final)
192 | acc01 <- bAccuracy(lab, final)
193 |
194 | final <- multiAdaSampling(dat, cls.noisy02, reducedDimName = "matPCs",
195 | classifier=base, percent=1, L=10)$final
196 | ari02 <- mclust::adjustedRandIndex(lab, final)
197 | acc02 <- bAccuracy(lab, final)
198 |
199 | final <- multiAdaSampling(dat, cls.noisy03, reducedDimName = "matPCs",
200 | classifier=base, percent=1, L=10)$final
201 | ari03 <- mclust::adjustedRandIndex(lab, final)
202 | acc03 <- bAccuracy(lab, final)
203 |
204 | final <- multiAdaSampling(dat, cls.noisy04, reducedDimName = "matPCs",
205 | classifier=base, percent=1, L=10)$final
206 | ari04 <- mclust::adjustedRandIndex(lab, final)
207 | acc04 <- bAccuracy(lab, final)
208 |
209 | final <- multiAdaSampling(dat, cls.noisy05, reducedDimName = "matPCs",
210 | classifier=base, percent=1, L=10)$final
211 | ari05 <- mclust::adjustedRandIndex(lab, final)
212 | acc05 <- bAccuracy(lab, final)
213 |
214 | c(
215 | acc01 = acc01,
216 | acc02 = acc02,
217 | acc03 = acc03,
218 | acc04 = acc04,
219 | acc05 = acc05,
220 | ari01 = ari01,
221 | ari02 = ari02,
222 | ari03 = ari03,
223 | ari04 = ari04,
224 | ari05 = ari05
225 | )
226 | })
227 |
228 | result = do.call(rbind, result)
229 | acc = result[,seq_len(5)]
230 | colnames(acc) = seq(from=0.1,to=0.5,by=0.1)
231 |
232 | ari = result[,seq(from= 6, to = 10)]
233 | colnames(ari) = seq(from=0.1,to=0.5,by=0.1)
234 |
235 | ```
236 |
237 |
238 | We can visualise the performance of the scReClassify. The boxes represent the
239 | accuracy and the ARI after scReClassify. The red markers indicate the baseline
240 | (prior to scReClassify).
241 |
242 |
243 | ```{r}
244 | plot.new()
245 | par(mfrow = c(1,2))
246 | boxplot(acc, col="lightblue", main="SVM Accuracy",
247 | ylim=c(0.45, 1), xlab = "rho", ylab = "Accuracy")
248 | points(x=seq_len(5), y=c(
249 | bAccuracy(lab, cls.noisy01),
250 | bAccuracy(lab, cls.noisy02),
251 | bAccuracy(lab, cls.noisy03),
252 | bAccuracy(lab, cls.noisy04),
253 | bAccuracy(lab, cls.noisy05)),
254 | col="red3", pch=c(2,3,4,5,6), cex=1)
255 | boxplot(ari, col="lightblue", main="SVM ARI",
256 | ylim=c(0.25, 1), xlab = "rho", ylab = "ARI")
257 | points(x=seq_len(5), y=c(
258 | mclust::adjustedRandIndex(lab, cls.noisy01),
259 | mclust::adjustedRandIndex(lab, cls.noisy02),
260 | mclust::adjustedRandIndex(lab, cls.noisy03),
261 | mclust::adjustedRandIndex(lab, cls.noisy04),
262 | mclust::adjustedRandIndex(lab, cls.noisy05)),
263 | col="red3", pch=c(2,3,4,5,6), cex=1)
264 | ```
265 |
266 | The plot shows that with scReClassify, cell type information have been refined
267 | (boxes are higher than the red markers). The scReClassified results show
268 | higher accuracy across noise levels 0.1 - 0.4 (i.e. closer to the true label).
269 | With the noise level 0.5, it is showing similar accuracy which is as expected
270 | because the initial label contains equal amount of true and false information
271 | and thus making it difficult for the algorithm to learn the true label. This
272 | shows that scReClassify is also robust to noisy cell type labels.
273 |
274 |
275 | # Part B. scReClassify (mislabeled cell type correction)
276 |
277 | scReClassify has shown promising result with the synthetic noise we have
278 | created. Here we will use scReClassify on the actual cell type label from public
279 | repository. The data we will use is a mouse fetal liver dataset from GEO with an
280 | accession ID GSE87795.
281 |
282 |
283 | ```{r}
284 | # PCA procedure
285 | reducedDim(dat, "matPCs") <- matPCs(dat, assay = "logNorm", 0.7)
286 |
287 |
288 | # run scReClassify
289 | set.seed(1)
290 | cellTypes.reclassify <- multiAdaSampling(dat, cellTypes,
291 | reducedDimName = "matPCs",
292 | classifier = "svm", percent = 1, L = 10)
293 |
294 | # Verification by marker genes
295 | End <- c("ITGA2B", "ITGB3")
296 | ```
297 |
298 | Below is a table of cell type labels classified to a different cell types after
299 | scReClassify.
300 |
301 | ```{r}
302 | # check examples
303 | idx <- which(cellTypes.reclassify$final != cellTypes)
304 |
305 | cbind(original=cellTypes[idx], reclassify=cellTypes.reclassify$final[idx]) %>%
306 | DT::datatable()
307 | ```
308 |
309 |
310 | Here, we visualise the expression level of the a cells that is reclassified for
311 | demontration purpose. The box plots are the marker gene expression levels
312 | grouped by cells types. The expression level of the reclassified cell (Cell ID:
313 | E13.5_C14) are highlighted as red marker.
314 |
315 | ```{r}
316 | mat <- assay(dat, "logNorm")
317 |
318 | c1 <- mat[, which(cellTypes=="Endothelial Cell")]
319 | c2 <- mat[, which(cellTypes=="Erythrocyte")]
320 | c3 <- mat[, which(cellTypes=="Hepatoblast")]
321 | c4 <- mat[, which(cellTypes=="Macrophage")]
322 | c5 <- mat[, which(cellTypes=="Megakaryocyte")]
323 | c6 <- mat[, which(cellTypes=="Mesenchymal Cell")]
324 | cs <- rainbow(length(table(cellTypes)))
325 |
326 |
327 | # (example 1 E13.5_C14)
328 | #####
329 | par(mfrow=c(1,2))
330 | marker <- End[1]
331 | boxplot(c1[marker,], c2[marker,], c3[marker,],
332 | c4[marker,], c5[marker,], c6[marker,],
333 | col=cs, main=marker,
334 | names=c("Others", "Others", "Others", "Orignal",
335 | "Reclassified", "Others"), las=2, xlab = "Labels",
336 | ylab = "log2FPKM")
337 | points(5, mat[marker, which(colnames(mat) %in% "E13.5_C14")],
338 | pch=16, col="red", cex=2)
339 |
340 | marker <- End[2]
341 | boxplot(c1[marker,], c2[marker,], c3[marker,],
342 | c4[marker,], c5[marker,], c6[marker,],
343 | col=cs, main=marker,
344 | names=c("Others", "Others", "Others", "Orignal",
345 | "Reclassified", "Others"), las=2, xlab = "Labels",
346 | ylab = "log2FPKM")
347 | points(5, mat[marker, which(colnames(mat) %in% "E13.5_C14")],
348 | pch=16, col="red", cex=2)
349 | ```
350 | As shown in the boxplots above, the expression level of the reclassified cell
351 | (red dot) is similar to the expression levels of the reclassified cell types in
352 | a marker gene. This highlights that the E13.5_C14 cell has a similar expression
353 | profiles to the reclassified cell types rather than its originally labeled cell
354 | type. Thus, we were able to identify that E13.5_C14 potentially belongs to the
355 | reclassified cell type with scReClassify.
356 |
357 |
358 | # SessionInfo
359 |
360 | ```{r}
361 | sessionInfo()
362 | ```
363 |
364 |
365 |
--------------------------------------------------------------------------------