auxfunc.R 34.8 KB
Newer Older
dmattek's avatar
dmattek committed
1
2
3
4
#
# Time Course Inspector: Shiny app for plotting time series data
# Author: Maciej Dobrzynski
#
dmattek's avatar
dmattek committed
5
# Auxilary functions & definitions of global constants
dmattek's avatar
dmattek committed
6
7
8
#


Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
9
10
11
12
13
library(ggplot2)
library(RColorBrewer)
library(gplots) # for heatmap.2
library(grid) # for modifying grob
library(Hmisc) # for CI calculation
dmattek's avatar
dmattek committed
14

15
16

# Global parameters ----
17

dmattek's avatar
dmattek committed
18

19
20
21
# if true, additional output printed to R console
DEB = T

22
# font sizes in pts for plots in the manuscript
Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
23
24
25
26
27
28
# PLOTFONTBASE = 8
# PLOTFONTAXISTEXT = 8
# PLOTFONTAXISTITLE = 8
# PLOTFONTFACETSTRIP = 10
# PLOTFONTLEGEND = 8

dmattek's avatar
dmattek committed
29
# font sizes in pts for screen display
Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
30
31
32
33
34
35
36
37
38
39
40
41
42
PLOTFONTBASE = 16
PLOTFONTAXISTEXT = 16
PLOTFONTAXISTITLE = 16
PLOTFONTFACETSTRIP = 20
PLOTFONTLEGEND = 16

# height (in pixels) of ribbon and single traj. plots
PLOTRIBBONHEIGHT = 500 # in pixels
PLOTTRAJHEIGHT = 500 # in pixels
PLOTPSDHEIGHT = 500 # in pixels
PLOTBOXHEIGHT = 500 # in pixels
PLOTSCATTERHEIGHT = 500 # in pixels
PLOTWIDTH = 85 # in percent
43
44

# default number of facets in plots
Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
45
PLOTNFACETDEFAULT = 3
46

dmattek's avatar
dmattek committed
47
# internal column names
dmattek's avatar
dmattek committed
48
COLRT   = 'time'
dmattek's avatar
dmattek committed
49
50
51
52
53
54
55
56
57
58
COLY    = 'y'
COLID   = 'id'
COLIDUNI = 'trackObjectsLabelUni'
COLGR   = 'group'
COLIN   = 'mid.in'
COLOBJN = 'obj.num'
COLPOSX = 'pos.x'
COLPOSY = 'pos.y'
COLIDX = 'IDX'
COLIDXDIFF = 'IDXdiff'
dmattek's avatar
dmattek committed
59
COLCL = 'cl'
dmattek's avatar
dmattek committed
60
61
62
63
64
65

# file names
FCSVOUTLIERS = 'outliers.csv'
FCSVTCCLEAN  = 'tCoursesSelected_clean.csv'
FPDFTCMEAN   = "tCoursesMeans.pdf"
FPDFTCSINGLE = "tCourses.pdf"
66
FPDFTCPSD    = 'tCoursesPsd.pdf'
dmattek's avatar
dmattek committed
67
68
69
70
FPDFBOXAUC   = 'boxplotAUC.pdf'
FPDFBOXTP    = 'boxplotTP.pdf'
FPDFSCATTER  = 'scatter.pdf'

dmattek's avatar
dmattek committed
71
# Colour definitions ----
dmattek's avatar
dmattek committed
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
rhg_cols <- c(
  "#771C19",
  "#AA3929",
  "#E25033",
  "#F27314",
  "#F8A31B",
  "#E2C59F",
  "#B6C5CC",
  "#8E9CA3",
  "#556670",
  "#000000"
)

md_cols <- c(
  "#FFFFFF",
  "#F8A31B",
  "#F27314",
  "#E25033",
  "#AA3929",
  "#FFFFCC",
  "#C2E699",
  "#78C679",
  "#238443"
)

97
# list of palettes for the heatmap
dmattek's avatar
dmattek committed
98
l.col.pal = list(
dmattek's avatar
dmattek committed
99
100
101
102
  "Spectral" = 'Spectral',
  "Red-Yellow-Green" = 'RdYlGn',
  "Red-Yellow-Blue" = 'RdYlBu',
  "Greys" = "Greys",
dmattek's avatar
dmattek committed
103
104
105
  "Reds" = "Reds",
  "Oranges" = "Oranges",
  "Greens" = "Greens",
dmattek's avatar
dmattek committed
106
  "Blues" = "Blues"
dmattek's avatar
dmattek committed
107
108
)

109
110
111
112
113
114
115
116
117
118
# list of palettes for the dendrogram
l.col.pal.dend = list(
  "Rainbow" = 'rainbow_hcl',
  "Sequential" = 'sequential_hcl',
  "Heat" = 'heat_hcl',
  "Terrain" = 'terrain_hcl',
  "Diverge HCL" = 'diverge_hcl',
  "Diverge HSV" = 'diverge_hsv'
)

dmattek's avatar
dmattek committed
119
120
121
122
123
124
125
126
127
128
129
# list of palettes for the dendrogram
l.col.pal.dend.2 = list(
  "Colorblind 10" = 'Color Blind',
  "Tableau 10" = 'Tableau 10',
  "Tableau 20" = 'Tableau 20',
  "Classic 10" = "Classic 10",
  "Classic 20" = "Classic 20",
  "Traffic 9" = 'Traffic',
  "Seattle Grays 5" = 'Seattle Grays'
)

dmattek's avatar
dmattek committed
130
131
# Clustering algorithms ----

dmattek's avatar
dmattek committed
132
133
s.cl.linkage = c("complete",
                 "average",
dmattek's avatar
dmattek committed
134
135
136
137
138
                 "single",
                 "centroid",
                 "ward.D",
                 "ward.D2",
                 "mcquitty")
dmattek's avatar
dmattek committed
139

dmattek's avatar
dmattek committed
140
141
s.cl.spar.linkage = c("complete",
                      "average",
dmattek's avatar
dmattek committed
142
143
144
                      "single",
                      "centroid")

dmattek's avatar
dmattek committed
145
146
147
148
149
150
151
152
s.cl.diss = c("euclidean", 
              "maximum", 
              "manhattan", 
              "canberra", 
              "DTW")

s.cl.spar.diss = c("squared.distance",
                   "absolute.value")
dmattek's avatar
dmattek committed
153
154
155


# Help text ----
dmattek's avatar
Added:    
dmattek committed
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
# Creates a popup with help text
# From: https://gist.github.com/jcheng5/5913297
helpPopup <- function(title, content,
                      placement=c('right', 'top', 'left', 'bottom'),
                      trigger=c('click', 'hover', 'focus', 'manual')) {
  tagList(
    singleton(
      tags$head(
        tags$script("$(function() { $(\"[data-toggle='popover']\").popover(); })")
      )
    ),
    tags$a(
      href = "#", class = "btn btn-mini", `data-toggle` = "popover",
      title = title, `data-content` = content, `data-animation` = TRUE,
      `data-placement` = match.arg(placement, several.ok=TRUE)[1],
      `data-trigger` = match.arg(trigger, several.ok=TRUE)[1],
      #tags$i(class="icon-question-sign")
      # changed based on http://stackoverflow.com/questions/30436013/info-bubble-text-in-a-shiny-interface
      icon("question")
    )
  )
}

dmattek's avatar
dmattek committed
179
helpText.server = c(
dmattek's avatar
dmattek committed
180
181
182
183
184
185
186
187
188
189
190
191
  alDataFormat =  paste0("<p>Switch between long and wide formats of input data. ",
                           "TCI accepts CSV or compressed CSV files (gz or bz2).</p>",
                           "<p><b>Long format</b> - a row is a single data point and consecutive time series are arranged vertically. ",
                           "Data file should contain at least 3 columns separated with a comma:</p>",
                           "<li>Identifier of a time series</li>",
                           "<li>Time points</li>",
                           "<li>A time-varying variable</li>",
                           "<br>",
                           "<p><b>Wide format</b> - a row is a time series with columns as time points.",
                           "At least 3 columns shuold be present:</p>",
                           "<li>First two columns in wide format should contain grouping and track IDs</li>",
                           "<li>A column with a time point. Headers of columns with time points need to be numeric</li>"), #2
dmattek's avatar
dmattek committed
192
193
194
  'Generate 60 random synthetic time series distributed evenly among 6 groups. Every time series has 60 time points.',             #3
  'Load CSV file with a column of track IDs for removal. IDs should correspond to those used for plotting.',                       #4
  'Load CSV file with 5 columns: grouping, start and end tpts of stimulation, start and end of y-position, dummy column with ID.', #5
dmattek's avatar
dmattek committed
195
  chBtrajInter = 'Interpolate missing time points and pre-existing NAs. Missing time points are rows entirely missing from the dataset. To interpolate, the interval of the time column must be provided.',                       #6
dmattek's avatar
dmattek committed
196
  'If the track ID is unique only within a group, make it unique globally by combining with grouping columns.',                    #7
majpark21's avatar
majpark21 committed
197
198
199
  'Load CSV file with 5 columns: grouping, start and end time points of stimulation, start and end of y-position, dummy column with ID.', #5
  'Interpolate missing time points indicated with NAs. In addition, add NA if a row with a time point is completely missing. The interval of the time column must be provided to know which rows are missing.',                       #6
  'If the track ID is not globally unique, try to make it unique by prepending another column to the track ID (typically the group column).',                    #7
dmattek's avatar
dmattek committed
200
201
202
  'Select columns to group data according to treatment, condition, etc.',                                                          #8
  'Select math operation to perform on a single or two columns,',                                                                  #9
  'Select range of time for further processing.',                                                                                  #10
majpark21's avatar
majpark21 committed
203
  'Divide measurements by the mean/median or calculate z-score with respect to selected time span.',                                #11
dmattek's avatar
dmattek committed
204
205
  'Fold-change or z-score with respect to selected time span.',                                                                    #12
  'Normalise with respect to this time span.',                                                                                     #13
majpark21's avatar
majpark21 committed
206
  'Calculate fold-change and z-score using the median and Median Absolute Deviation, instead of the mean and standard deviation.',                 #14
dmattek's avatar
dmattek committed
207
  'Normalise to mean/median of selected time calculated globally, per group, or for individual time series.',                      #15
dmattek's avatar
dmattek committed
208
  'Download time series after modification in this section.',                                                                      #16
dmattek's avatar
dmattek committed
209
210
211
  alertNAsPresent = "NAs present in the measurement column. Consider interpolation.",
  alertWideMissesNumericTime = "Non-numeric headers of time columns. Data in wide format should have numeric column headers corresponding to time points.",
  alertWideTooFewColumns = "Insufficient columns. Data in wide format should contain at least 3 columns: grouping, track ID, and a single time point."
212
213
)

dmattek's avatar
dmattek committed
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
# Functions for data processing ----
#' Calculate the mean and CI around time series
#'
#' @param in.dt Data table in long format
#' @param in.col.meas Name of the column with the measurement
#' @param in.col.by Column names for grouping (default NULL - no grouping). Typically, you want to use at least a column with time.
#' @param in.type Choice of normal approximation or boot-strapping
#' @param ... Other params passed to smean.cl.normal and smean.cl.boot; these include \code{conf.int} for the confidence level, \code{B} for the number of boot-strapping iterations.
#'
#' @return Datatable with columns: Mean, lower and upper CI, and grouping columns if provided.
#' @export
#' @import data.table
#' @import Hmisc
#'
#' @examples
#'
#'
#' # generate synthetic time series; 100 time points long, with 10 randomly placed NAs
#' dt.tmp = genTraj(100, 10, 6, 3, in.addna = 10)
#'
#' # calculate single stats from all time points
#' calcTrajCI(dt.tmp, 'objCyto_Intensity_MeanIntensity_imErkCor')
#'
#' # calculate the mean and CI along the time course
#' calcTrajCI(dt.tmp, 'objCyto_Intensity_MeanIntensity_imErkCor', 'Metadata_RealTime')
LOCcalcTrajCI = function(in.dt, in.col.meas, in.col.by = NULL, in.type = c('normal', 'boot'), ...) {
  in.type = match.arg(in.type)
  
  if (in.type %like% 'normal')
    loc.dt = in.dt[, as.list(smean.cl.normal(get(in.col.meas), ...)), by = in.col.by] else
      loc.dt = in.dt[, as.list(smean.cl.boot(get(in.col.meas), ...)), by = in.col.by]
    
    return(loc.dt)
}

249

250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
#' Calculate standard error of the mean
#'
#' @param x Vector
#' @param na.rm Remove NAs; default = FALSE
#'
#' @return A scalar with the result
#' @export
#'
#' @examples
LOCstderr = function(x, na.rm=FALSE) {
  if (na.rm) 
    x = na.omit(x)
  
  return(sqrt(var(x)/length(x)))
}

266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
#' Calculate the power spectrum density for time-series
#'
#' @param in.dt Data table in long format
#' @param in.col.meas Name of the column with the measurement
#' @param in.col.id Name of the column with the unique series identifier
#' @param in.col.by Column names for grouping (default NULL - no grouping). PSD of individual trajectories will be averaged within a group.
#' @param in.method Name of the method for PSD estimation, must be one of c("pgram", "ar"). Default to "pgram*.
#' @param in.return.period Wheter to return densities though periods (1/frequencies) instead of frequencies.
#' @param ... Other paramters to pass to stats::spectrum()
#'
#' @return Datatable with columns: (frequency or period), spec (the density) and grouping column
#' @export
#' @import data.table
#'
#' @examples
LOCcalcPSD <- function(in.dt,
                    in.col.meas,
                    in.col.id,
                    in.col.by,
                    in.method = "pgram",
                    in.return.period = TRUE,
287
                    in.time.btwPoints = 1,
288
289
                    ...){
  require(data.table)
290
291
292
293
294
295
296
297
298
  # Method "ar" returns $spec as matrix whereas "pgram" returns a vector, custom function to homogenze output format
  mySpectrum <- function(x, ...){
    args_spec <- list(x=x, plot=FALSE)
    inargs <- list(...)
    args_spec[names(inargs)] <- inargs
    out <- do.call(spectrum, args_spec)
    out$spec <- as.vector(out$spec)
    return(out)
  }
299
300
301
  if(!in.method %in% c("pgram", "ar")){
    stop('Method should be one of: c("pgram", "ar"')
  }
302
303
304
305
306
  dt_spec <-  in.dt[, (mySpectrum(get(in.col.meas), plot = FALSE, method = in.method)[c("freq", "spec")]), by = in.col.id]
  dt_group <- in.dt[, .SD[1, get(in.col.by)], by = in.col.id]
  setnames(dt_group, "V1", in.col.by)
  dt_spec <- merge(dt_spec, dt_group, by = in.col.id)
  dt_agg <- dt_spec[, .(spec = mean(spec)), by = c(in.col.by, "freq")]
307
  if(in.return.period){
308
    dt_agg[, period := 1/freq]
309
310
311
    dt_agg[, freq := NULL]
    # Adjust period unit to go from frame unit  to time unit
    dt_agg[, period := period * in.time.btwPoints]
312
  } else {
313
    dt_agg[, freq := freq * (1/in.time.btwPoints)]
314
    setnames(dt_agg, "freq", "frequency")
315
316
317
318
319
  }
  return(dt_agg)
}


dmattek's avatar
dmattek committed
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
#' Generate synthetic CellProfiler output with single cell time series
#'
#'
#'
#' @param in.ntpts Number of time points (default 60)
#' @param in.ntracks Number of tracks per FOV (default 10)
#' @param in.nfov Number of FOV (default 6)
#' @param in.nwells Number of wells (default 1)
#' @param in.addna Number of NAs to add randomly in the data (default NULL)
#'
#' @return Data table with the follwoing columns: Metadata_Site, Metadata_Well, Metadata_RealTime, objCyto_Intensity_MeanIntensity_imErkCor (normal distributed),
#' objNuc_Intensity_MeanIntensity_imErkCor (normal distributed), objNuc_Location_X and objNuc_Location_Y (uniform ditributed), TrackLabel
#' @export
#' @import data.table

dmattek's avatar
dmattek committed
335
LOCgenTraj <- function(in.ntpts = 60, in.ntracks = 10, in.nfov = 6, in.nwells = 1, in.addna = NULL, in.addout = NULL) {
dmattek's avatar
dmattek committed
336
337
338
339
340
341
342
343
344
345
346
  
  x.rand.1 = c(rnorm(in.ntpts * in.ntracks * in.nfov * 1/3, 0.5, 0.1), rnorm(in.ntpts * in.ntracks * in.nfov * 1/3,   1, 0.2), rnorm(in.ntpts * in.ntracks * in.nfov * 1/3,  2, 0.5))
  x.rand.2 = c(rnorm(in.ntpts * in.ntracks * in.nfov * 1/3, 0.25, 0.1), rnorm(in.ntpts * in.ntracks * in.nfov * 1/3, 0.5, 0.2),  rnorm(in.ntpts * in.ntracks * in.nfov * 1/3, 1, 0.2))
  
  # add NA's for testing
  if (!is.null(in.addna)) {
    locTabLen = length(x.rand.1)
    x.rand.1[round(runif(in.addna) * locTabLen)] = NA
    x.rand.2[round(runif(in.addna) * locTabLen)] = NA
  }
  
dmattek's avatar
dmattek committed
347
348
349
  # add outliers for testing
  if (!is.null(in.addout)) {
    locTabLen = length(x.rand.1)
Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
350
351
    x.rand.1[round(runif(in.addout) * locTabLen)] = 5
    x.rand.2[round(runif(in.addout) * locTabLen)] = 5
dmattek's avatar
dmattek committed
352
353
  }
  
dmattek's avatar
dmattek committed
354
355
  x.arg = rep(seq(1, in.ntpts), in.ntracks * in.nfov)
  
Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
356
357
358
359
360
361
362
363
  dt.nuc = data.table(well = rep(LETTERS[1:in.nwells], each = in.ntpts * in.nfov * in.ntracks / in.nwells),
                      group = rep(1:in.nfov, each = in.ntpts * in.ntracks),
                      time = x.arg,
                      y1 = x.rand.1,
                      y2  = x.rand.2,
                      posx = runif(in.ntpts * in.ntracks * in.nfov, min = 0, max = 1),
                      posy = runif(in.ntpts * in.ntracks * in.nfov, min = 0, max = 1),
                      id = rep(1:(in.ntracks*in.nfov), each = in.ntpts))
dmattek's avatar
dmattek committed
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
  
  return(dt.nuc)
}

#' Normalize Trajectory
#'
#' Returns original dt with an additional column with normalized quantity.
#' The column to be normalised is given by 'in.meas.col'.
#' The name of additional column is the same as in.meas.col but with ".norm" suffix added.
#' Normalisation is based on part of the trajectory;
#' this is defined by in.rt.min and max, and the column with time in.rt.col.#'
#'
#' @param in.dt Data table in long format
#' @param in.meas.col String with the column name to normalize
#' @param in.rt.col String with the colum name holding time
#' @param in.rt.min Lower bound for time period used for normalization
#' @param in.rt.max Upper bound for time period used for normalization
#' @param in.by.cols String vector with 'by' columns to calculate normalization per group; if NULL, no grouping is done
#' @param in.robust Whether robust measures should be used (median instead of mean, mad instead of sd); default TRUE
#' @param in.type Type of normalization: z.score or mean (i.e. fold change w.r.t. mean); default 'z-score'
#'
#' @return Returns original dt with an additional column with normalized quantity.
#' @export
#' @import data.table

LOCnormTraj = function(in.dt,
                    in.meas.col,
dmattek's avatar
dmattek committed
391
                    in.rt.col = COLRT,
dmattek's avatar
dmattek committed
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
                    in.rt.min = 10,
                    in.rt.max = 20,
                    in.by.cols = NULL,
                    in.robust = TRUE,
                    in.type = 'z.score') {
  loc.dt <-
    copy(in.dt) # copy so as not to alter original dt object w intermediate assignments
  
  if (is.null(in.by.cols)) {
    if (in.robust)
      loc.dt.pre.aggr = loc.dt[get(in.rt.col) >= in.rt.min &
                                 get(in.rt.col) <= in.rt.max, .(meas.md = median(get(in.meas.col), na.rm = TRUE),
                                                                meas.mad = mad(get(in.meas.col), na.rm = TRUE))]
    else
      loc.dt.pre.aggr = loc.dt[get(in.rt.col) >= in.rt.min &
                                 get(in.rt.col) <= in.rt.max, .(meas.md = mean(get(in.meas.col), na.rm = TRUE),
                                                                meas.mad = sd(get(in.meas.col), na.rm = TRUE))]
    
    loc.dt = cbind(loc.dt, loc.dt.pre.aggr)
  }  else {
    if (in.robust)
      loc.dt.pre.aggr = loc.dt[get(in.rt.col) >= in.rt.min &
                                 get(in.rt.col) <= in.rt.max, .(meas.md = median(get(in.meas.col), na.rm = TRUE),
                                                                meas.mad = mad(get(in.meas.col), na.rm = TRUE)), by = in.by.cols]
    else
      loc.dt.pre.aggr = loc.dt[get(in.rt.col) >= in.rt.min &
                                 get(in.rt.col) <= in.rt.max, .(meas.md = mean(get(in.meas.col), na.rm = TRUE),
                                                                meas.mad = sd(get(in.meas.col), na.rm = TRUE)), by = in.by.cols]
    
    loc.dt = merge(loc.dt, loc.dt.pre.aggr, by = in.by.cols)
  }
  
  
  if (in.type == 'z.score') {
    loc.dt[, meas.norm := (get(in.meas.col) - meas.md) / meas.mad]
  } else {
    loc.dt[, meas.norm := (get(in.meas.col) / meas.md)]
  }
  
  setnames(loc.dt, 'meas.norm', paste0(in.meas.col, '.norm'))
  
  loc.dt[, c('meas.md', 'meas.mad') := NULL]
  return(loc.dt)
}


dmattek's avatar
Added:    
dmattek committed
438

dmattek's avatar
dmattek committed
439
# Functions for clustering ----
dmattek's avatar
dmattek committed
440
441
442
443
444
445
446
447
448

# Return a dt with cell IDs and corresponding cluster assignments depending on dendrogram cut (in.k)
# This one works wth dist & hclust pair
# For sparse hierarchical clustering use getDataClSpar
# Arguments:
# in.dend  - dendrogram; usually output from as.dendrogram(hclust(distance_matrix))
# in.k - level at which dendrogram should be cut

getDataCl = function(in.dend, in.k) {
dmattek's avatar
Added:    
dmattek committed
449
450
  cat(file = stderr(), 'getDataCl \n')
  
dmattek's avatar
dmattek committed
451
  loc.clAssign = dendextend::cutree(in.dend, in.k, order_clusters_as_data = TRUE, )
dmattek's avatar
dmattek committed
452
453
454
455
  #print(loc.m)
  
  # The result of cutree containes named vector with names being cell id's
  # THIS WON'T WORK with sparse hierarchical clustering because there, the dendrogram doesn't have original id's
dmattek's avatar
dmattek committed
456
457
458
  loc.dt.clAssign = as.data.table(loc.clAssign, keep.rownames = T)
  setnames(loc.dt.clAssign, c(COLID, COLCL))
  
dmattek's avatar
dmattek committed
459
  
460
461
  #cat('===============\ndataCl:\n')
  #print(loc.dt.cl)
dmattek's avatar
dmattek committed
462
  return(loc.dt.clAssign)
dmattek's avatar
Added:    
dmattek committed
463
464
}

dmattek's avatar
dmattek committed
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483

# Return a dt with cell IDs and corresponding cluster assignments depending on dendrogram cut (in.k)
# This one works with sparse hierarchical clustering!
# Arguments:
# in.dend  - dendrogram; usually output from as.dendrogram(hclust(distance_matrix))
# in.k - level at which dendrogram should be cut
# in.id - vector of cell id's

getDataClSpar = function(in.dend, in.k, in.id) {
  cat(file = stderr(), 'getDataClSpar \n')
  
  loc.m = dendextend::cutree(in.dend, in.k, order_clusters_as_data = TRUE)
  #print(loc.m)
  
  # The result of cutree containes named vector with names being cell id's
  # THIS WON'T WORK with sparse hierarchical clustering because there, the dendrogram doesn't have original id's
  loc.dt.cl = data.table(id = in.id,
                         cl = loc.m)
  
484
485
  #cat('===============\ndataCl:\n')
  #print(loc.dt.cl)
dmattek's avatar
dmattek committed
486
487
488
489
490
  return(loc.dt.cl)
}



dmattek's avatar
Added:    
dmattek committed
491
492
493
494
495
496
497
498
499
500
501
502
503
# prepares a table with cluster numbers in 1st column and colour assignments in 2nd column
# the number of rows is determined by dendrogram cut
getClCol <- function(in.dend, in.k) {
  
  loc.col_labels <- get_leaves_branches_col(in.dend)
  loc.col_labels <- loc.col_labels[order(order.dendrogram(in.dend))]
  
  return(unique(
    data.table(cl.no = dendextend::cutree(in.dend, k = in.k, order_clusters_as_data = TRUE),
               cl.col = loc.col_labels)))
}


dmattek's avatar
dmattek committed
504
# Custom plotting functions ----
dmattek's avatar
dmattek committed
505

dmattek's avatar
dmattek committed
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544

#' Custom ggPlot theme based on theme_bw
#'
#' @param in.font.base
#' @param in.font.axis.text
#' @param in.font.axis.title
#' @param in.font.strip
#' @param in.font.legend
#'
#' @return
#' @export
#'
#' @examples
#'
LOCggplotTheme = function(in.font.base = 12,
                       in.font.axis.text = 12,
                       in.font.axis.title = 12,
                       in.font.strip = 14,
                       in.font.legend = 12) {
  loc.theme =
    theme_bw(base_size = in.font.base, base_family = "Helvetica") +
    theme(
      panel.spacing = unit(1, "lines"),
      panel.grid.minor = element_blank(),
      panel.grid.major = element_blank(),
      panel.border = element_blank(),
      axis.line = element_line(color = "black", size = 0.25),
      axis.text = element_text(size = in.font.axis.text),
      axis.title = element_text(size = in.font.axis.title),
      strip.text = element_text(size = in.font.strip, face = "bold"),
      strip.background = element_blank(),
      legend.key = element_blank(),
      legend.text = element_text(size = in.font.legend),
      legend.key.height = unit(1, "lines"),
      legend.key.width = unit(2, "lines"))
  
  return(loc.theme)
}

dmattek's avatar
dmattek committed
545
546
# Build Function to Return Element Text Object
# From: https://stackoverflow.com/a/36979201/1898713
dmattek's avatar
dmattek committed
547
LOCrotatedAxisElementText = function(angle, position='x', size = 12){
dmattek's avatar
dmattek committed
548
549
550
551
552
553
554
555
556
557
  angle     = angle[1]; 
  position  = position[1]
  positions = list(x=0, y=90, top=180, right=270)
  if(!position %in% names(positions))
    stop(sprintf("'position' must be one of [%s]",paste(names(positions),collapse=", ")), call.=FALSE)
  if(!is.numeric(angle))
    stop("'angle' must be numeric",call.=FALSE)
  rads = (-angle - positions[[ position ]])*pi/180
  hjust = round((1 - sin(rads)))/2
  vjust = round((1 + cos(rads)))/2
Maciej Dobrzynski's avatar
Maciej Dobrzynski committed
558
  element_text(size = size, angle = angle, vjust = vjust, hjust = hjust)
dmattek's avatar
dmattek committed
559
560
}

561
562
# Plot individual time series
LOCplotTraj = function(dt.arg, # input data table
dmattek's avatar
Mod:    
dmattek committed
563
564
565
566
567
568
569
570
571
572
573
                        x.arg,  # string with column name for x-axis
                        y.arg, # string with column name for y-axis
                        group.arg, # string with column name for grouping time series (typicaly cell ID)
                        facet.arg, # string with column name for facetting
                        facet.ncol.arg = 2, # default number of facet columns
                        facet.color.arg = NULL, # vector with list of colours for adding colours to facet names (currently a horizontal line on top of the facet is drawn)
                        line.col.arg = NULL, # string with column name for colouring time series (typically when individual time series are selected in UI)
                        xlab.arg = NULL, # string with x-axis label
                        ylab.arg = NULL, # string with y-axis label
                        plotlab.arg = NULL, # string with plot label
                        dt.stim.arg = NULL, # plotting additional dataset; typically to indicate stimulations (not fully implemented yet, not tested!)
574
575
                        x.stim.arg = c('tstart', 'tend'), # column names in stimulation dt with x and xend parameters
                        y.stim.arg = c('ystart', 'yend'), # column names in stimulation dt with y and yend parameters
dmattek's avatar
dmattek committed
576
577
578
579
                        tfreq.arg = 1, # unused
                        xlim.arg = NULL, # limits of x-axis; for visualisation only, not trimmimng data
                        ylim.arg = NULL, # limits of y-axis; for visualisation only, not trimmimng data
                        stim.bar.width.arg = 0.5, # width of the stimulation line; plotted under time series
dmattek's avatar
Mod:    
dmattek committed
580
                        aux.label1 = NULL, # 1st point label; used for interactive plotting; displayed in the tooltip; typically used to display values of column holding x & y coordinates
dmattek's avatar
Added:    
dmattek committed
581
                        aux.label2 = NULL,
582
                        aux.label3 = NULL,
dmattek's avatar
Added:    
dmattek committed
583
584
585
586
587
                        stat.arg = c('', 'mean', 'CI', 'SE')) {
  
  # match arguments for stat plotting
  loc.stat = match.arg(stat.arg, several.ok = TRUE)

dmattek's avatar
Added:    
dmattek committed
588
589
  
  # aux.label12 are required for plotting XY positions in the tooltip of the interactive (plotly) graph
dmattek's avatar
dmattek committed
590
591
  p.tmp = ggplot(dt.arg,
                 aes_string(x = x.arg,
dmattek's avatar
dmattek committed
592
                            y = y.arg,
dmattek's avatar
Added:    
dmattek committed
593
                            group = group.arg,
594
595
596
597
598
                            label = group.arg))
  #,
  #                          label  = aux.label1,
  #                          label2 = aux.label2,
  #                          label3 = aux.label3))
dmattek's avatar
dmattek committed
599
  
dmattek's avatar
dmattek committed
600
601
602
603
604
605
606
607
608
609
610
611
612
  if (is.null(line.col.arg)) {
    p.tmp = p.tmp +
      geom_line(alpha = 0.25, 
                              size = 0.25)
  }
  else {
    p.tmp = p.tmp + 
      geom_line(aes_string(colour = line.col.arg), 
                              alpha = 0.5, 
                              size = 0.5) +
      scale_color_manual(name = '', 
                         values =c("FALSE" = rhg_cols[7], "TRUE" = rhg_cols[3], "SELECTED" = 'green', "NOT SEL" = rhg_cols[7]))
  }
dmattek's avatar
Mod:    
dmattek committed
613
614
615
616
617
618
619
620
621
622

  # this is temporary solution for adding colour according to cluster number
  # use only when plotting traj from clustering!
  # a horizontal line is added at the top of data
  if (!is.null(facet.color.arg)) {

    loc.y.max = max(dt.arg[, c(y.arg), with = FALSE])
    loc.dt.cl = data.table(xx = 1:length(facet.color.arg), yy = loc.y.max)
    setnames(loc.dt.cl, 'xx', facet.arg)
    
dmattek's avatar
Fixed:    
dmattek committed
623
624
    # adjust facet.color.arg to plot
    
dmattek's avatar
Mod:    
dmattek committed
625
626
627
628
629
    p.tmp = p.tmp +
      geom_hline(data = loc.dt.cl, colour = facet.color.arg, yintercept = loc.y.max, size = 4) +
      scale_colour_manual(values = facet.color.arg,
                          name = '')
  }
dmattek's avatar
dmattek committed
630
  
dmattek's avatar
Added:    
dmattek committed
631
632
  if ('mean' %in% loc.stat)
    p.tmp = p.tmp + 
dmattek's avatar
dmattek committed
633
634
    stat_summary(
      aes_string(y = y.arg, group = 1),
dmattek's avatar
dmattek committed
635
636
      fun.y = mean, 
      na.rm = T,
dmattek's avatar
Added:    
dmattek committed
637
      colour = 'red',
dmattek's avatar
dmattek committed
638
639
640
641
      linetype = 'solid',
      size = 1,
      geom = "line",
      group = 1
dmattek's avatar
Added:    
dmattek committed
642
643
644
645
646
647
648
    )

  if ('CI' %in% loc.stat)
    p.tmp = p.tmp + 
    stat_summary(
      aes_string(y = y.arg, group = 1),
      fun.data = mean_cl_normal,
dmattek's avatar
dmattek committed
649
      na.rm = T,
dmattek's avatar
Added:    
dmattek committed
650
      colour = 'red',
dmattek's avatar
Mod:    
dmattek committed
651
      alpha = 0.25,
dmattek's avatar
Added:    
dmattek committed
652
653
654
655
656
657
658
659
660
      geom = "ribbon",
      group = 1
    )
  
  if ('SE' %in% loc.stat)
    p.tmp = p.tmp + 
    stat_summary(
      aes_string(y = y.arg, group = 1),
      fun.data = mean_se,
dmattek's avatar
dmattek committed
661
      na.rm = T,
dmattek's avatar
Added:    
dmattek committed
662
      colour = 'red',
dmattek's avatar
Mod:    
dmattek committed
663
      alpha = 0.25,
dmattek's avatar
Added:    
dmattek committed
664
665
666
667
668
669
670
      geom = "ribbon",
      group = 1
    )
  
  
  
  p.tmp = p.tmp + 
dmattek's avatar
dmattek committed
671
672
673
    facet_wrap(as.formula(paste("~", facet.arg)),
               ncol = facet.ncol.arg,
               scales = "free_x")
674
675
676
677

  # plot stimulation bars underneath time series
  # dt.stim.arg is read separately and should contain 4 columns with
  # xy positions of beginning and end of the bar
dmattek's avatar
dmattek committed
678
679
  if(!is.null(dt.stim.arg)) {
    p.tmp = p.tmp + geom_segment(data = dt.stim.arg,
680
681
682
683
684
                                 aes_string(x = x.stim.arg[1],
                                            xend = x.stim.arg[2],
                                            y = y.stim.arg[1],
                                            yend = y.stim.arg[2],
                                            group = 'group'),
dmattek's avatar
dmattek committed
685
                                 colour = rhg_cols[[3]],
686
                                 size = stim.bar.width.arg) 
dmattek's avatar
dmattek committed
687
688
  }
  
dmattek's avatar
dmattek committed
689
  p.tmp = p.tmp + coord_cartesian(xlim = xlim.arg, ylim = ylim.arg)
dmattek's avatar
dmattek committed
690
  
dmattek's avatar
dmattek committed
691
692
693
694
  p.tmp = p.tmp + 
    xlab(paste0(xlab.arg, "\n")) +
    ylab(paste0("\n", ylab.arg)) +
    ggtitle(plotlab.arg) +
695
696
697
698
699
    LOCggplotTheme(in.font.base = PLOTFONTBASE, 
                   in.font.axis.text = PLOTFONTAXISTEXT, 
                   in.font.axis.title = PLOTFONTAXISTITLE, 
                   in.font.strip = PLOTFONTFACETSTRIP, 
                   in.font.legend = PLOTFONTLEGEND) + 
700
    theme(legend.position = "top")
dmattek's avatar
dmattek committed
701
  
dmattek's avatar
Mod:    
dmattek committed
702
  return(p.tmp)
dmattek's avatar
dmattek committed
703
704
}

705
706
707
708
709
710
711
712
713
714
# Plot average time series with CI together in one facet
LOCplotTrajRibbon = function(dt.arg, # input data table
                          x.arg, # string with column name for x-axis
                          y.arg, # string with column name for y-axis
                          group.arg = NULL, # string with column name for grouping time series (here, it's a column corresponding to grouping by condition)
                          col.arg = NULL, # colour pallette for individual time series
                          dt.stim.arg = NULL, # data table with stimulation pattern
                          x.stim.arg = c('tstart', 'tend'), # column names in stimulation dt with x and xend parameters
                          y.stim.arg = c('ystart', 'yend'), # column names in stimulation dt with y and yend parameters
                          stim.bar.width.arg = 0.5,
dmattek's avatar
dmattek committed
715
716
                          xlim.arg = NULL, # limits of x-axis; for visualisation only, not trimmimng data
                          ylim.arg = NULL, # limits of y-axis; for visualisation only, not trimmimng data
717
                          ribbon.lohi.arg = c('Lower', 'Upper'), # column names containing lower and upper bound for plotting the ribbon, e.g. for CI; set to NULL to avoid plotting the ribbon
718
719
720
721
722
723
                          ribbon.fill.arg = 'grey50',
                          ribbon.alpha.arg = 0.5,
                          xlab.arg = NULL,
                          ylab.arg = NULL,
                          plotlab.arg = NULL) {
  
724
725
726
727
728
  p.tmp = ggplot(dt.arg, aes_string(x = x.arg, group = group.arg))
  
  if (!is.null(ribbon.lohi.arg))
    p.tmp = p.tmp + 
      geom_ribbon(aes_string(ymin = ribbon.lohi.arg[1], ymax = ribbon.lohi.arg[2]),
729
                fill = ribbon.fill.arg,
730
731
732
                alpha = ribbon.alpha.arg)
  
  p.tmp = p.tmp + geom_line(aes_string(y = y.arg, colour = group.arg))
733
  
dmattek's avatar
dmattek committed
734

735
736
737
738
739
740
741
742
743
744
745
746
747
748
  # plot stimulation bars underneath time series
  # dt.stim.arg is read separately and should contain 4 columns with
  # xy positions of beginning and end of the bar
  if(!is.null(dt.stim.arg)) {
    p.tmp = p.tmp + geom_segment(data = dt.stim.arg,
                                 aes_string(x = x.stim.arg[1],
                                     xend = x.stim.arg[2],
                                     y = y.stim.arg[1],
                                     yend = y.stim.arg[2]),
                                 colour = rhg_cols[[3]],
                                 size = stim.bar.width.arg,
                                 group = 1) 
  }

dmattek's avatar
dmattek committed
749
  p.tmp = p.tmp + coord_cartesian(xlim = xlim.arg, ylim = ylim.arg)
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
  
  if (is.null(col.arg)) {
    p.tmp = p.tmp +
      scale_color_discrete(name = '')
  } else {
    p.tmp = p.tmp +
      scale_colour_manual(values = col.arg, name = '')
  }
  
  if (!is.null(plotlab.arg))
    p.tmp = p.tmp + ggtitle(plotlab.arg)
  
  p.tmp = p.tmp +
    xlab(xlab.arg) +
    ylab(ylab.arg)
  
  return(p.tmp)
767
768
}

769
# Plot average power spectrum density per facet
majpark21's avatar
majpark21 committed
770
771
772
773
774
LOCplotPSD <- function(dt.arg, # input data table
                    x.arg, # string with column name for x-axis
                    y.arg, # string with column name for y-axis
                    group.arg=NULL, # string with column name for grouping time series (here, it's a column corresponding to grouping by condition)
                    xlab.arg = x.arg,
775
                    ylab.arg = y.arg,
776
                    facet.color.arg = NULL){
majpark21's avatar
majpark21 committed
777
778
779
780
781
782
  require(ggplot2)
  if(length(setdiff(c(x.arg, y.arg, group.arg), colnames(dt.arg))) > 0){
    stop(paste("Missing columns in dt.arg: ", setdiff(c(x.arg, y.arg, group.arg), colnames(dt.arg))))
  }
  p.tmp <- ggplot(dt.arg, aes_string(x=x.arg, y=y.arg)) +
    geom_line() +
783
    geom_rug(sides="b", alpha = 1, color = "lightblue") +
majpark21's avatar
majpark21 committed
784
785
    facet_wrap(group.arg) +
    labs(x = xlab.arg, y = ylab.arg)
786
  
787
788
789
790
791
792
793
794
  if (!is.null(facet.color.arg)) {
    
    loc.y.max = max(dt.arg[, c(y.arg), with = FALSE])
    loc.dt.cl = data.table(xx = 1:length(facet.color.arg), yy = loc.y.max)
    setnames(loc.dt.cl, 'xx', group.arg)
    
    # adjust facet.color.arg to plot
    
795
    p.tmp = p.tmp +
796
797
798
      geom_hline(data = loc.dt.cl, colour = facet.color.arg, yintercept = loc.y.max, size = 4) +
      scale_colour_manual(values = facet.color.arg,
                          name = '')
799
800
  }
  
majpark21's avatar
majpark21 committed
801
802
  return(p.tmp)
}
803

dmattek's avatar
dmattek committed
804
805
806
807
808
809
810
811
812
813
814
815
816
#' Plot a scatter plot with an optional linear regression
#'
#' @param dt.arg input of data.table with 2 columns with x and y coordinates
#' @param facet.arg 
#' @param facet.ncol.arg 
#' @param xlab.arg 
#' @param ylab.arg 
#' @param plotlab.arg 
#' @param alpha.arg 
#' @param trend.arg 
#' @param ci.arg 

LOCggplotScat = function(dt.arg, 
dmattek's avatar
dmattek committed
817
818
819
820
821
822
                        facet.arg = NULL,
                        facet.ncol.arg = 2,
                        xlab.arg = NULL,
                        ylab.arg = NULL,
                        plotlab.arg = NULL,
                        alpha.arg = 1,
dmattek's avatar
dmattek committed
823
824
                        trend.arg = T,
                        ci.arg = 0.95) {
dmattek's avatar
dmattek committed
825
  
dmattek's avatar
dmattek committed
826
  p.tmp = ggplot(dt.arg, aes(x = x, y = y, label = id)) +
dmattek's avatar
dmattek committed
827
828
829
    geom_point(alpha = alpha.arg)

  if (trend.arg) {
dmattek's avatar
dmattek committed
830
831
    p.tmp = p.tmp +
      stat_smooth(
dmattek's avatar
dmattek committed
832
        method = "lm",
dmattek's avatar
dmattek committed
833
        fullrange = FALSE,
dmattek's avatar
dmattek committed
834
        level = ci.arg,
dmattek's avatar
dmattek committed
835
836
837
        colour = 'blue'
      )
  }
dmattek's avatar
dmattek committed
838

dmattek's avatar
dmattek committed
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
  if (!is.null(facet.arg)) {
    p.tmp = p.tmp +
      facet_wrap(as.formula(paste("~", facet.arg)),
                 ncol = facet.ncol.arg)
    
  }
  
  if (!is.null(xlab.arg))
    p.tmp = p.tmp +
      xlab(paste0(xlab.arg, "\n"))
  
  if (!is.null(ylab.arg))
    p.tmp = p.tmp +
      ylab(paste0("\n", ylab.arg))
  
  if (!is.null(plotlab.arg))
    p.tmp = p.tmp +
      ggtitle(paste0(plotlab.arg, "\n"))
  
  p.tmp = p.tmp +
859
860
861
862
863
    LOCggplotTheme(in.font.base = PLOTFONTBASE, 
                   in.font.axis.text = PLOTFONTAXISTEXT, 
                   in.font.axis.title = PLOTFONTAXISTITLE, 
                   in.font.strip = PLOTFONTFACETSTRIP, 
                   in.font.legend = PLOTFONTLEGEND) + 
864
865
    theme(legend.position = "none")

dmattek's avatar
dmattek committed
866
867
  return(p.tmp)
}
dmattek's avatar
dmattek committed
868

869

dmattek's avatar
dmattek committed
870
LOCplotHeatmap <- function(data.arg,
dmattek's avatar
Mod:    
dmattek committed
871
872
873
874
875
876
877
878
879
880
881
882
                          dend.arg,
                          palette.arg,
                          palette.rev.arg = TRUE,
                          dend.show.arg = TRUE,
                          key.show.arg = TRUE,
                          margin.x.arg = 5,
                          margin.y.arg = 20,
                          nacol.arg = 0.5,
                          colCol.arg = NULL,
                          labCol.arg = NULL,
                          font.row.arg = 1,
                          font.col.arg = 1,
883
                          breaks.arg = NULL,
dmattek's avatar
Mod:    
dmattek committed
884
885
                          title.arg = 'Clustering') {
  
886
887
  loc.n.colbreaks = 99
  
dmattek's avatar
Mod:    
dmattek committed
888
889
  if (palette.rev.arg)
    my_palette <-
890
    rev(colorRampPalette(brewer.pal(9, palette.arg))(n = loc.n.colbreaks))
dmattek's avatar
Mod:    
dmattek committed
891
892
  else
    my_palette <-
893
    colorRampPalette(brewer.pal(9, palette.arg))(n = loc.n.colbreaks)
dmattek's avatar
Mod:    
dmattek committed
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
  
  
  col_labels <- get_leaves_branches_col(dend.arg)
  col_labels <- col_labels[order(order.dendrogram(dend.arg))]
  
  if (dend.show.arg) {
    assign("var.tmp.1", dend.arg)
    var.tmp.2 = "row"
  } else {
    assign("var.tmp.1", FALSE)
    var.tmp.2 = "none"
  }
  
  loc.p = heatmap.2(
    data.arg,
    Colv = "NA",
    Rowv = var.tmp.1,
    srtCol = 90,
    dendrogram = var.tmp.2,
    trace = "none",
    key = key.show.arg,
    margins = c(margin.x.arg, margin.y.arg),
    col = my_palette,
    na.col = grey(nacol.arg),
    denscol = "black",
    density.info = "density",
    RowSideColors = col_labels,
    colRow = col_labels,
    colCol = colCol.arg,
    labCol = labCol.arg,
    #      sepcolor = grey(input$inPlotHierGridColor),
    #      colsep = 1:ncol(loc.dm),
    #      rowsep = 1:nrow(loc.dm),
    cexRow = font.row.arg,
    cexCol = font.col.arg,
dmattek's avatar
dmattek committed
929
930
    main = title.arg,
    symbreaks = FALSE,
931
932
    symkey = FALSE,
    breaks = if (is.null(breaks.arg)) NULL else seq(breaks.arg[1], breaks.arg[2], length.out = loc.n.colbreaks+1)
dmattek's avatar
Mod:    
dmattek committed
933
934
935
936
  )
  
  return(loc.p)
}