server.R 24.3 KB
Newer Older
dmattek's avatar
dmattek committed
1

2

dmattek's avatar
dmattek committed
3

dmattek's avatar
dmattek committed
4
5
6
7
8
9
10
11
12
13
# This is the server logic for a Shiny web application.
# You can find out more about building applications with Shiny here:
#
# http://shiny.rstudio.com
#

library(shiny)
library(shinyjs) #http://deanattali.com/shinyjs/
library(data.table)
library(ggplot2)
dmattek's avatar
dmattek committed
14
library(gplots) # for heatmap.2
dmattek's avatar
dmattek committed
15
library(plotly)
dmattek's avatar
dmattek committed
16
17
library(d3heatmap) # for interactive heatmap
library(dendextend) # for color_branches
18
library(colorspace) # for palettes (used to colour dendrogram)
dmattek's avatar
dmattek committed
19
20
21
library(RColorBrewer)
library(sparcl) # sparse hierarchical and k-means
library(scales) # for percentages on y scale
dmattek's avatar
Added:    
dmattek committed
22
23
library(dtw) # for dynamic time warping
library(imputeTS) # for interpolating NAs
24
library(tca) # for time series manipulatiom, e.g. normTraj, genTraj, plotTrajRibbon
dmattek's avatar
dmattek committed
25

26
# increase file upload limit
dmattek's avatar
Added:    
dmattek committed
27
options(shiny.maxRequestSize = 200 * 1024 ^ 2)
dmattek's avatar
dmattek committed
28

dmattek's avatar
dmattek committed
29
shinyServer(function(input, output, session) {
30
  useShinyjs()
dmattek's avatar
dmattek committed
31
  
32
33
34
35
36
37
  # This is only set at session start
  # we use this as a way to determine which input was
  # clicked in the dataInBoth reactive
  counter <- reactiveValues(
    # The value of inDataGen1,2 actionButton is the number of times they were pressed
    dataGen1     = isolate(input$inDataGen1),
dmattek's avatar
Added:    
dmattek committed
38
39
    dataLoadNuc  = isolate(input$inButLoadNuc),
    dataLoadTrajRem = isolate(input$inButLoadTrajRem)
40
    #dataLoadStim = isolate(input$inButLoadStim)
dmattek's avatar
dmattek committed
41
42
  )
  
dmattek's avatar
dmattek committed
43
44
45
  ####
  ## UI for side panel
  
dmattek's avatar
dmattek committed
46
  # FILE LOAD
47
48
49
50
51
52
53
54
55
56
  # This button will reset the inFileLoad
  observeEvent(input$inButReset, {
    reset("inFileLoadNuc")  # reset is a shinyjs function
    #reset("inButLoadStim")  # reset is a shinyjs function
  })
  
  # generate random dataset 1
  dataGen1 <- eventReactive(input$inDataGen1, {
    cat("dataGen1\n")
    
57
    return(tca::genTraj(in.nwells = 3))
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
  })
  
  # load main data file
  dataLoadNuc <- eventReactive(input$inButLoadNuc, {
    cat("dataLoadNuc\n")
    locFilePath = input$inFileLoadNuc$datapath
    
    counter$dataLoadNuc <- input$inButLoadNuc - 1
    
    if (is.null(locFilePath) || locFilePath == '')
      return(NULL)
    else {
      return(fread(locFilePath))
    }
  })
  
dmattek's avatar
dmattek committed
74
75
76
77
78
79
80
  # This button will reset the inFileLoad
  observeEvent(input$butReset, {
    reset("inFileLoadNuc")  # reset is a shinyjs function
    #    reset("inFileStimLoad")  # reset is a shinyjs function
    
  })
  
dmattek's avatar
Added:    
dmattek committed
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
  # UI for loading csv with cell IDs for trajectory removal
  output$uiFileLoadTrajRem = renderUI({
    cat(file = stderr(), 'UI uiFileLoadTrajRem\n')
    
    if(input$chBtrajRem) 
      fileInput(
        'inFileLoadTrajRem',
        'Select data file (e.g. badTraj.csv) and press "Load Data"',
        accept = c('text/csv', 'text/comma-separated-values,text/plain')
      )
  })
  
  output$uiButLoadTrajRem = renderUI({
    cat(file = stderr(), 'UI uiButLoadTrajRem\n')
    
    if(input$chBtrajRem)
      actionButton("inButLoadTrajRem", "Load Data")
  })

  # load main data file
  dataLoadTrajRem <- eventReactive(input$inButLoadTrajRem, {
    cat(file = stderr(), "dataLoadTrajRem\n")
    locFilePath = input$inFileLoadTrajRem$datapath
    
    counter$dataLoadTrajRem <- input$inButLoadTrajRem - 1
    
    if (is.null(locFilePath) || locFilePath == '')
      return(NULL)
    else {
      return(fread(locFilePath))
    }
  })
  
dmattek's avatar
dmattek committed
114
115
  
  # COLUMN SELECTION
dmattek's avatar
dmattek committed
116
117
118
  output$varSelTrackLabel = renderUI({
    cat(file = stderr(), 'UI varSelTrackLabel\n')
    locCols = getDataNucCols()
119
    locColSel = locCols[grep('(T|t)rack|ID|id', locCols)[1]] # index 1 at the end in case more matches; select 1st; matches TrackLabel, tracklabel, Track Label etc
dmattek's avatar
dmattek committed
120
121
122
    
    selectInput(
      'inSelTrackLabel',
dmattek's avatar
dmattek committed
123
      'Select Track Label (e.g. objNuc_TrackObjects_Label):',
dmattek's avatar
dmattek committed
124
125
126
127
128
129
130
131
132
      locCols,
      width = '100%',
      selected = locColSel
    )
  })
  
  output$varSelTime = renderUI({
    cat(file = stderr(), 'UI varSelTime\n')
    locCols = getDataNucCols()
133
    locColSel = locCols[grep('(T|t)ime|Metadata_T', locCols)[1]] # index 1 at the end in case more matches; select 1st; matches RealTime, realtime, real time, etc.
dmattek's avatar
dmattek committed
134
135
136
    
    selectInput(
      'inSelTime',
dmattek's avatar
dmattek committed
137
      'Select time column (e.g. Metadata_T, RealTime):',
dmattek's avatar
dmattek committed
138
139
140
141
142
      locCols,
      width = '100%',
      selected = locColSel
    )
  })
143
144
145
146

  output$varSelTimeFreq = renderUI({
    cat(file = stderr(), 'UI varSelTimeFreq\n')
    
147
148
149
150
151
152
153
154
155
156
    if (input$chBtrajInter) {
      numericInput(
        'inSelTimeFreq',
        'Provide time frequency:',
        min = 1,
        step = 1,
        width = '100%',
        value = 1
      )
    }
157
  })
dmattek's avatar
dmattek committed
158
159
160
161
162
163
164
165
  
  # This is main field to select plot facet grouping
  # It's typically a column with the entire experimental description,
  # e.g. in Yannick's case it's Stim_All_Ch or Stim_All_S.
  # In Coralie's case it's a combination of 3 columns called Stimulation_...
  output$varSelGroup = renderUI({
    cat(file = stderr(), 'UI varSelGroup\n')
    
dmattek's avatar
dmattek committed
166
167
168
169
170
    if (input$chBgroup) {
      
      locCols = getDataNucCols()
      
      if (!is.null(locCols)) {
171
172
173
        locColSel = locCols[grep('(G|g)roup|(S|s)tim_All|(S|s)timulation|(S|s)ite', locCols)[1]]

        #cat('UI varSelGroup::locColSel ', locColSel, '\n')
dmattek's avatar
dmattek committed
174
175
176
177
178
179
180
181
        selectInput(
          'inSelGroup',
          'Select one or more facet groupings (e.g. Site, Well, Channel):',
          locCols,
          width = '100%',
          selected = locColSel,
          multiple = TRUE
        )
dmattek's avatar
dmattek committed
182
183
184
185
186
187
188
      }
    }
  })
  
  output$varSelSite = renderUI({
    cat(file = stderr(), 'UI varSelSite\n')
    
189
    if (input$chBtrackUni) {
dmattek's avatar
Added:    
dmattek committed
190
      locCols = getDataNucCols()
191
      locColSel = locCols[grep('(S|s)ite|(S|s)eries', locCols)[1]] # index 1 at the end in case more matches; select 1st
dmattek's avatar
Added:    
dmattek committed
192
193
194
195
196
197
198
199
200
      
      selectInput(
        'inSelSite',
        'Select FOV (e.g. Metadata_Site or Metadata_Series):',
        locCols,
        width = '100%',
        selected = locColSel
      )
    }
dmattek's avatar
dmattek committed
201
202
203
204
205
206
207
208
209
210
  })
  
  
  
  
  output$varSelMeas1 = renderUI({
    cat(file = stderr(), 'UI varSelMeas1\n')
    locCols = getDataNucCols()
    
    if (!is.null(locCols)) {
211
      locColSel = locCols[grep('objCyto_Intensity_MeanIntensity_imErkCor|(R|r)atio|(I|i)ntensity|y', locCols)[1]] # index 1 at the end in case more matches; select 1st
dmattek's avatar
dmattek committed
212

dmattek's avatar
dmattek committed
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
      selectInput(
        'inSelMeas1',
        'Select 1st measurement:',
        locCols,
        width = '100%',
        selected = locColSel
      )
    }
  })
  
  
  output$varSelMeas2 = renderUI({
    cat(file = stderr(), 'UI varSelMeas2\n')
    locCols = getDataNucCols()
    
    if (!is.null(locCols) &&
        !(input$inSelMath %in% c('', '1 / '))) {
230
      locColSel = locCols[grep('objNuc_Intensity_MeanIntensity_imErkCor', locCols)[1]] # index 1 at the end in case more matches; select 1st
dmattek's avatar
dmattek committed
231

dmattek's avatar
dmattek committed
232
233
234
235
236
237
238
239
240
241
      selectInput(
        'inSelMeas2',
        'Select 2nd measurement',
        locCols,
        width = '100%',
        selected = locColSel
      )
    }
  })
  
dmattek's avatar
dmattek committed
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
  # UI for trimming x-axis (time)
  output$uiSlTimeTrim = renderUI({
    cat(file = stderr(), 'UI uiSlTimeTrim\n')
    
    if (input$chBtimeTrim) {
      locTpts  = getDataTpts()
      
      if(is.null(locTpts))
        return(NULL)
      
      locRTmin = min(locTpts)
      locRTmax = max(locTpts)
      
      sliderInput(
        'slTimeTrim',
        label = 'Time range to include',
        min = locRTmin,
        max = locRTmax,
        value = c(locRTmin, locRTmax),
        step = 1
      )
      
    }
  })
dmattek's avatar
dmattek committed
266
  
dmattek's avatar
dmattek committed
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
  # UI for normalization
  
  output$uiChBnorm = renderUI({
    cat(file = stderr(), 'UI uiChBnorm\n')
    
    if (input$chBnorm) {
      radioButtons(
        'rBnormMeth',
        label = 'Select method',
        choices = list('fold-change' = 'mean', 'z-score' = 'z.score')
      )
    }
  })
  
  output$uiSlNorm = renderUI({
    cat(file = stderr(), 'UI uiSlNorm\n')
    
    if (input$chBnorm) {
      locTpts  = getDataTpts()
      
      if(is.null(locTpts))
        return(NULL)
      
      locRTmin = min(locTpts)
      locRTmax = max(locTpts)
      
      sliderInput(
        'slNormRtMinMax',
        label = 'Time range for norm.',
        min = locRTmin,
        max = locRTmax,
dmattek's avatar
dmattek committed
298
299
        value = c(locRTmin, 0.1 * locRTmax), 
        step = 1
dmattek's avatar
dmattek committed
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
      )
    }
  })
  
  output$uiChBnormRobust = renderUI({
    cat(file = stderr(), 'UI uiChBnormRobust\n')
    
    if (input$chBnorm) {
      checkboxInput('chBnormRobust',
                    label = 'Robust stats',
                    FALSE)
    }
  })
  
  output$uiChBnormGroup = renderUI({
    cat(file = stderr(), 'UI uiChBnormGroup\n')
    
    if (input$chBnorm) {
      radioButtons('chBnormGroup',
dmattek's avatar
Mod:    
dmattek committed
319
                   label = 'Normalisation grouping',
320
                   choices = list('Entire dataset' = 'none', 'Per facet' = 'group', 'Per trajectory' = 'id'))
dmattek's avatar
dmattek committed
321
322
323
324
    }
  })
  
  
dmattek's avatar
dmattek committed
325
326
327
328
329
  # UI for removing outliers
  output$uiSlOutliers = renderUI({
    cat(file = stderr(), 'UI uiSlOutliers\n')
    
    if (input$chBoutliers) {
dmattek's avatar
Mod:    
dmattek committed
330
      
dmattek's avatar
dmattek committed
331
332
333
334
335
      sliderInput(
        'slOutliersPerc',
        label = 'Percentage of middle data',
        min = 90,
        max = 100,
dmattek's avatar
Fixed:    
dmattek committed
336
        value = 99.5, 
dmattek's avatar
dmattek committed
337
338
        step = 0.1
      )
dmattek's avatar
dmattek committed
339
      
dmattek's avatar
Mod:    
dmattek committed
340
      
dmattek's avatar
dmattek committed
341
342
343
    }
  })
  
dmattek's avatar
dmattek committed
344
345
346
347
348
349
350
351
352
  output$uiTxtOutliers = renderUI({
    if (input$chBoutliers) {
      
      p("Total tracks")
      
    }
    
  })
  
dmattek's avatar
dmattek committed
353
  
dmattek's avatar
dmattek committed
354
355
356
  ####
  ## data processing
  
357
358
359
360
361
362
363
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
391
392
393
394
395
396
397
398
399
400
401
402
403
404
  dataInBoth <- reactive({
    # Without direct references to inDataGen1,2 and inFileLoad, inDataGen2
    #    does not trigger running this reactive once inDataGen1 is used.
    # This is one of the more nuanced areas of reactive programming in shiny
    #    due to the if else logic, it isn't fetched once inDataGen1 is available
    # The morale is use direct retrieval of inputs to guarantee they are available
    #    for if else logic checks!
    
    locInGen1 = input$inDataGen1
    locInLoadNuc = input$inButLoadNuc
    #locInLoadStim = input$inButLoadStim
    
    cat(
      "dataInBoth\ninGen1: ",
      locInGen1,
      "   prev=",
      isolate(counter$dataGen1),
      "\ninDataNuc: ",
      locInLoadNuc,
      "   prev=",
      isolate(counter$dataLoadNuc),
      # "\ninDataStim: ",
      # locInLoadStim,
      # "   prev=",
      # isolate(counter$dataLoadStim),
      "\n"
    )
    
    # isolate the checks of counter reactiveValues
    # as we set the values in this same reactive
    if (locInGen1 != isolate(counter$dataGen1)) {
      cat("dataInBoth if inDataGen1\n")
      dm = dataGen1()
      # no need to isolate updating the counter reactive values!
      counter$dataGen1 <- locInGen1
    } else if (locInLoadNuc != isolate(counter$dataLoadNuc)) {
      cat("dataInBoth if inDataLoadNuc\n")
      dm = dataLoadNuc()
      # no need to isolate updating the counter reactive values!
      counter$dataLoadNuc <- locInLoadNuc
    } else {
      cat("dataInBoth else\n")
      dm = NULL
    }
    return(dm)
  })
  
  # return column names of the main dt
dmattek's avatar
dmattek committed
405
  getDataNucCols <- reactive({
406
407
408
409
410
411
412
413
414
415
416
    cat(file = stderr(), 'getDataNucCols: in\n')
    loc.dt = dataInBoth()
    
    if (is.null(loc.dt))
      return(NULL)
    else
      return(colnames(loc.dt))
  })
  
  # return dt with an added column with unique track object label
  dataMod <- reactive({
dmattek's avatar
dmattek committed
417
    cat(file = stderr(), 'dataMod\n')
418
419
    loc.dt = dataInBoth()
    
dmattek's avatar
dmattek committed
420
    if (is.null(loc.dt))
421
422
      return(NULL)
    
423
    if (input$chBtrackUni) {
dmattek's avatar
Added:    
dmattek committed
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
      loc.types = lapply(loc.dt, class)
      if(loc.types[[input$inSelTrackLabel]] %in% c('numeric', 'integer') & loc.types[[input$inSelSite]] %in% c('numeric', 'integer'))
      {
        loc.dt[, trackObjectsLabelUni := paste(sprintf("%03d", get(input$inSelSite)),
                                               sprintf("%04d", get(input$inSelTrackLabel)),
                                               sep = "_")]
      } else if(loc.types[[input$inSelTrackLabel]] %in% c('numeric', 'integer')) {
        loc.dt[, trackObjectsLabelUni := paste(sprintf("%s", get(input$inSelSite)),
                                               sprintf("%04d", get(input$inSelTrackLabel)),
                                               sep = "_")]
      } else if(loc.types[[input$inSelSite]] %in% c('numeric', 'integer')) {
        loc.dt[, trackObjectsLabelUni := paste(sprintf("%03d", get(input$inSelSite)),
                                               sprintf("%s", get(input$inSelTrackLabel)),
                                               sep = "_")]
      } else {
        loc.dt[, trackObjectsLabelUni := paste(sprintf("%s", get(input$inSelSite)),
                                               sprintf("%s", get(input$inSelTrackLabel)),
                                               sep = "_")]
      }
dmattek's avatar
Added:    
dmattek committed
443
    } else {
dmattek's avatar
Added:    
dmattek committed
444
      loc.dt[, trackObjectsLabelUni := get(input$inSelTrackLabel)]
dmattek's avatar
Added:    
dmattek committed
445
446
    }
    
dmattek's avatar
dmattek committed
447
    
dmattek's avatar
Added:    
dmattek committed
448
449
450
451
452
453
    # remove trajectories based on uploaded csv

    if (input$chBtrajRem) {
      cat(file = stderr(), 'dataMod: trajRem not NULL\n')
      
      loc.dt.rem = dataLoadTrajRem()
dmattek's avatar
dmattek committed
454
      loc.dt = loc.dt[!(trackObjectsLabelUni %in% loc.dt.rem[[1]])]
dmattek's avatar
Added:    
dmattek committed
455
456
    }
    
457
458
459
    return(loc.dt)
  })
  
dmattek's avatar
dmattek committed
460
461
462
463
464
  # return all unique track object labels (created in dataMod)
  # This will be used to display in UI for trajectory highlighting
  getDataTrackObjLabUni <- reactive({
    cat(file = stderr(), 'getDataTrackObjLabUni\n')
    loc.dt = dataMod()
465
    
dmattek's avatar
dmattek committed
466
467
468
469
    if (is.null(loc.dt))
      return(NULL)
    else
      return(unique(loc.dt$trackObjectsLabelUni))
470
471
  })
  
dmattek's avatar
Mod:    
dmattek committed
472
  
dmattek's avatar
dmattek committed
473
474
475
  # return all unique time points (real time)
  # This will be used to display in UI for box-plot
  # These timepoints are from the original dt and aren't affected by trimming of x-axis
dmattek's avatar
dmattek committed
476
477
478
  getDataTpts <- reactive({
    cat(file = stderr(), 'getDataTpts\n')
    loc.dt = dataMod()
479
    
dmattek's avatar
dmattek committed
480
481
482
483
    if (is.null(loc.dt))
      return(NULL)
    else
      return(unique(loc.dt[[input$inSelTime]]))
484
485
  })
  
dmattek's avatar
dmattek committed
486
  
487
488
489
  
  # prepare data for plotting time courses
  # returns dt with these columns:
dmattek's avatar
dmattek committed
490
  #    realtime - selected from input
dmattek's avatar
dmattek committed
491
  #    y        - measurement selected from input
dmattek's avatar
dmattek committed
492
  #               (can be a single column or result of an operation on two cols)
493
494
  #    id       - trackObjectsLabelUni; created in dataMod based on TrackObjects_Label
  #               and FOV column such as Series or Site (if TrackObjects_Label not unique across entire dataset)
dmattek's avatar
dmattek committed
495
496
  #    group    - grouping variable for facetting from input
  #    mid.in   - column with trajectory selection status from the input file or
497
498
499
500
  #               highlight status from UI 
  #               (column created if mid.in present in uploaded data or tracks are selected in the UI)
  #    obj.num  - created if ObjectNumber column present in the input data 
  #    pos.x,y  - created if columns with x and y positions present in the input data
501
  data4trajPlot <- reactive({
dmattek's avatar
dmattek committed
502
    cat(file = stderr(), 'data4trajPlot\n')
503
504
    
    loc.dt = dataMod()
dmattek's avatar
dmattek committed
505
    if (is.null(loc.dt))
506
507
      return(NULL)
    
508
    # create expression for 'y' column based on measurements and math operations selected in UI
dmattek's avatar
dmattek committed
509
    if (input$inSelMath == '')
510
511
512
513
514
515
      loc.s.y = input$inSelMeas1
    else if (input$inSelMath == '1 / ')
      loc.s.y = paste0(input$inSelMath, input$inSelMeas1)
    else
      loc.s.y = paste0(input$inSelMeas1, input$inSelMath, input$inSelMeas2)
    
516
    # create expression for 'group' column
517
518
    # creates a merged column based on other columns from input
    # used for grouping of plot facets
dmattek's avatar
dmattek committed
519
520
521
522
523
524
525
526
527
528
529
    if (input$chBgroup) {
      if(length(input$inSelGroup) == 0)
        return(NULL)
      
      loc.s.gr = sprintf("paste(%s, sep=';')",
                         paste(input$inSelGroup, sep = '', collapse = ','))
    } else {
      # if no grouping required, fill 'group' column with 0
      # because all the plotting relies on the presence of the group column
      loc.s.gr = "paste('0')"
    }
530
    
dmattek's avatar
dmattek committed
531
532

    # column name with time
533
534
    loc.s.rt = input$inSelTime
    
dmattek's avatar
dmattek committed
535
536
    # Assign tracks selected for highlighting in UI
    loc.tracks.highlight = input$inSelHighlight
537
    locButHighlight = input$chBhighlightTraj
dmattek's avatar
dmattek committed
538
    
dmattek's avatar
Added:    
dmattek committed
539
540
    
    # Find column names with position
541
    loc.s.pos.x = names(loc.dt)[grep('(L|l)ocation.*X|(P|p)os.x|(P|p)osx', names(loc.dt))[1]]
542
    loc.s.pos.y = names(loc.dt)[grep('(L|l)ocation.*Y|(P|p)os.y|(P|p)osy', names(loc.dt))[1]]
dmattek's avatar
Added:    
dmattek committed
543
    
544
    cat('Position columns: ', loc.s.pos.x, loc.s.pos.y, '\n')
545
546
    
    if (!is.na(loc.s.pos.x) & !is.na(loc.s.pos.y))
dmattek's avatar
Added:    
dmattek committed
547
548
549
550
      locPos = TRUE
    else
      locPos = FALSE
    
551
552
553
554
    
    # Find column names with ObjectNumber
    # This is different from TrackObject_Label and is handy to keep
    # because labels on segmented images are typically ObjectNumber
555
556
557
558
559
560
    loc.s.objnum = names(loc.dt)[grep('(O|o)bject(N|n)umber', names(loc.dt))[1]]
    #cat('data4trajPlot::loc.s.objnum ', loc.s.objnum, '\n')
    if (is.na(loc.s.objnum)) {
      locObjNum = FALSE
    }
    else {
dmattek's avatar
dmattek committed
561
      loc.s.objnum = loc.s.objnum[1]
562
      locObjNum = TRUE
dmattek's avatar
dmattek committed
563
    }
564
565
    
    
566
567
    # if dataset contains column mid.in with trajectory filtering status,
    # then, include it in plotting
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
    if (sum(names(loc.dt) %in% 'mid.in') > 0)
      locMidIn = TRUE
    else
      locMidIn = FALSE
    
    ## Build expression for selecting columns from loc.dt
    # Core columns
    s.colexpr = paste0('.(y = ', loc.s.y,
                       ', id = trackObjectsLabelUni', 
                       ', group = ', loc.s.gr,
                       ', realtime = ', loc.s.rt)
    
    # account for the presence of 'mid.in' column in uploaded data
    if(locMidIn)
      s.colexpr = paste0(s.colexpr, 
                         ', mid.in = mid.in')
    
    # include position x,y columns in uploaded data
    if(locPos)
      s.colexpr = paste0(s.colexpr, 
                         ', pos.x = ', loc.s.pos.x,
                         ', pos.y = ', loc.s.pos.y)
    
    # include ObjectNumber column
    if(locObjNum)
      s.colexpr = paste0(s.colexpr, 
                         ', obj.num = ', loc.s.objnum)
    
    # close bracket, finish the expression
    s.colexpr = paste0(s.colexpr, ')')
    
    # create final dt for output based on columns selected above
    loc.out = loc.dt[, eval(parse(text = s.colexpr))]
    
    
    # if track selection ON
    if (locButHighlight){
      # add a 3rd level with status of track selection
      # to a column with trajectory filtering status in the uploaded file
      if(locMidIn)
        loc.out[, mid.in := ifelse(id %in% loc.tracks.highlight, 'SELECTED', mid.in)]
      else
dmattek's avatar
Mod:    
dmattek committed
610
        # add a column with status of track selection
611
        loc.out[, mid.in := ifelse(id %in% loc.tracks.highlight, 'SELECTED', 'NOT SEL')]
612
    }
613
      
dmattek's avatar
dmattek committed
614

615
    ## Interpolate missing data and NA data points
616
    # From: https://stackoverflow.com/questions/28073752/r-how-to-add-rows-for-missing-values-for-unique-group-sequences
617
618
619
    # Tracks are interpolated only within first and last time points of every cell id
    # Datasets can have different realtime frequency (e.g. every 1', 2', etc),
    # or the frame number metadata can be missing, as is the case for tCourseSelected files that already have realtime column.
620
    # Therefore, we cannot rely on that info to get time frequency; user provides this number!
621
    
622
623
    setkey(loc.out, group, id, realtime)

624
625
    if (input$chBtrajInter) {
      # here we fill missing data with NA's
626
      loc.out = loc.out[setkey(loc.out[, .(seq(min(realtime, na.rm = T), max(realtime, na.rm = T), input$inSelTimeFreq)), by = .(group, id)], group, id, V1)]
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
      
      # x-check: print all rows with NA's
      print('Rows with NAs:')
      print(loc.out[rowSums(is.na(loc.out)) > 0, ])
      
      # NA's may be already present in the dataset'.
      # Interpolate (linear) them with na.interpolate as well
      if(locPos)
        s.cols = c('y', 'pos.x', 'pos.y')
      else
        s.cols = c('y')
      
      loc.out[, (s.cols) := lapply(.SD, na.interpolation), by = id, .SDcols = s.cols]
      
      
      # !!! Current issue with interpolation:
      # The column mid.in is not taken into account.
      # If a trajectory is selected in the UI,
      # the mid.in column is added (if it doesn't already exist in the dataset),
      # and for the interpolated point, it will still be NA. Not really an issue.
      #
      # Also, think about the current option of having mid.in column in the uploaded dataset.
      # Keep it? Expand it?
      # Create a UI filed for selecting the column with mid.in data.
      # What to do with that column during interpolation (see above)
      
    }    
dmattek's avatar
Mod:    
dmattek committed
654
    
655
    ## Trim x-axis (time)
dmattek's avatar
dmattek committed
656
657
658
    if(input$chBtimeTrim) {
      loc.out = loc.out[realtime >= input$slTimeTrim[[1]] & realtime <= input$slTimeTrim[[2]] ]
    }
dmattek's avatar
dmattek committed
659
    
660
    ## Normalization
dmattek's avatar
dmattek committed
661
    # F-n myNorm adds additional column with .norm suffix
dmattek's avatar
dmattek committed
662
    if (input$chBnorm) {
663
      loc.out = tca::normTraj(
dmattek's avatar
dmattek committed
664
665
666
667
668
669
670
671
672
673
        in.dt = loc.out,
        in.meas.col = 'y',
        in.rt.col = 'realtime',
        in.rt.min = input$slNormRtMinMax[1],
        in.rt.max = input$slNormRtMinMax[2],
        in.type = input$rBnormMeth,
        in.robust = input$chBnormRobust,
        in.by.cols = if(input$chBnormGroup %in% 'none') NULL else input$chBnormGroup
      )
      
dmattek's avatar
dmattek committed
674
675
      # Column with normalized data is renamed to the original name
      # Further code assumes column name y produced by data4trajPlot
dmattek's avatar
dmattek committed
676
677
678
679
      loc.out[, y := NULL]
      setnames(loc.out, 'y.norm', 'y')
    }
    
dmattek's avatar
dmattek committed
680
681
682
683
684
685
    ##### MOD HERE
    ## display number of filtered tracks in textUI: uiTxtOutliers
    ## How? 
    ## 1. through reactive values?
    ## 2. through additional comumn to tag outliers?
    
dmattek's avatar
dmattek committed
686
687
688
689
690
691
692
693
694
695
    # Remove outliers
    # 1. Scale all points (independently per track)
    # 2. Pick time points that exceed the bounds
    # 3. Identify IDs of outliers
    # 4. Select cells that don't have these IDs
    
    cat('Ncells orig = ', length(unique(loc.out$id)), '\n')
    
    if (input$chBoutliers) {
      loc.out[, y.sc := scale(y)]  
696
697
      loc.tmp = loc.out[ y.sc < quantile(y.sc, (1 - input$slOutliersPerc * 0.01)*0.5, na.rm = T) | 
                           y.sc > quantile(y.sc, 1 - (1 - input$slOutliersPerc * 0.01)*0.5, na.rm = T)]
dmattek's avatar
dmattek committed
698
699
700
701
702
      loc.out = loc.out[!(id %in% unique(loc.tmp$id))]
      loc.out[, y.sc := NULL]
    }
    
    cat('Ncells trim = ', length(unique(loc.out$id)), '\n')
dmattek's avatar
Mod:    
dmattek committed
703
    
dmattek's avatar
dmattek committed
704
    return(loc.out)
dmattek's avatar
dmattek committed
705
706
  })
  
dmattek's avatar
dmattek committed
707
708
709
710
711
712
713
714
715
716
717
718
719
  
  
  # prepare data for clustering
  # return a matrix with:
  # cells as columns
  # time points as rows
  data4clust <- reactive({
    cat(file = stderr(), 'data4clust\n')
    
    loc.dt = data4trajPlot()
    if (is.null(loc.dt))
      return(NULL)
    
dmattek's avatar
Added:    
dmattek committed
720
    #print(loc.dt)
dmattek's avatar
dmattek committed
721
    loc.out = dcast(loc.dt, id ~ realtime, value.var = 'y')
dmattek's avatar
Added:    
dmattek committed
722
    #print(loc.out)
dmattek's avatar
dmattek committed
723
724
    loc.rownames = loc.out$id
    
dmattek's avatar
Mod:    
dmattek committed
725
    
dmattek's avatar
dmattek committed
726
727
    loc.out = as.matrix(loc.out[, -1])
    rownames(loc.out) = loc.rownames
dmattek's avatar
Added:    
dmattek committed
728
    
729
730
    # This might be removed entirely because all NA treatment happens in data4trajPlot
    # Clustering should work with NAs present. These might result from data itself or from missing time point rows that were turned into NAs when dcast-ing from long format.
dmattek's avatar
Added:    
dmattek committed
731
732
733
734
    # Remove NA's
    # na.interpolation from package imputeTS works with multidimensional data
    # but imputation is performed for each column independently
    # The matrix for clustering contains time series in rows, hence transposing it twice
735
    # loc.out = t(na.interpolation(t(loc.out)))
dmattek's avatar
Added:    
dmattek committed
736
    
dmattek's avatar
dmattek committed
737
    return(loc.out)
dmattek's avatar
Mod:    
dmattek committed
738
  }) 
dmattek's avatar
dmattek committed
739
  
dmattek's avatar
dmattek committed
740
  
dmattek's avatar
Added:    
dmattek committed
741
742
743
744
745
746
747
748
749
  # download data as prepared for plotting
  # after all modification
  output$downloadDataClean <- downloadHandler(
    filename = 'tCoursesSelected_clean.csv',
    content = function(file) {
      write.csv(data4trajPlot(), file, row.names = FALSE)
    }
  )
  
750
751
  ###### Trajectory plotting
  callModule(modTrajRibbonPlot, 'modTrajRibbon', 
752
753
             in.data = data4trajPlot,
             in.fname = function() return( "tCoursesMeans.pdf"))
dmattek's avatar
dmattek committed
754
  
755
  ###### Trajectory plotting
756
757
758
  callModule(modTrajPlot, 'modTrajPlot', 
             in.data = data4trajPlot, 
             in.fname = function() {return( "tCourses.pdf")})
759
760
761
  
  ## UI for selecting trajectories
  # The output data table of data4trajPlot is modified based on inSelHighlight field
dmattek's avatar
dmattek committed
762
763
  output$varSelHighlight = renderUI({
    cat(file = stderr(), 'UI varSelHighlight\n')
dmattek's avatar
dmattek committed
764
    
dmattek's avatar
dmattek committed
765
766
767
    locBut = input$chBhighlightTraj
    if (!locBut)
      return(NULL)
dmattek's avatar
dmattek committed
768
    
dmattek's avatar
dmattek committed
769
    loc.v = getDataTrackObjLabUni()
dmattek's avatar
dmattek committed
770
    if (!is.null(loc.v)) {
771
      selectInput(
dmattek's avatar
dmattek committed
772
773
774
        'inSelHighlight',
        'Select one or more rajectories:',
        loc.v,
775
        width = '100%',
dmattek's avatar
dmattek committed
776
        multiple = TRUE
777
      )
dmattek's avatar
dmattek committed
778
779
780
    }
  })
  
781
  ###### AUC calculation and plotting
782
  callModule(modAUCplot, 'tabAUC', data4trajPlot, in.fname = function() return('boxplotAUC.pdf'))
dmattek's avatar
Added:    
dmattek committed
783
  
dmattek's avatar
Added:    
dmattek committed
784
  ###### Box-plot
785
  callModule(tabBoxPlot, 'tabBoxPlot', data4trajPlot, in.fname = function() return('boxplotTP.pdf'))
dmattek's avatar
dmattek committed
786
  
dmattek's avatar
dmattek committed
787
  ###### Scatter plot
788
  callModule(tabScatterPlot, 'tabScatter', data4trajPlot, in.fname = function() return('scatter.pdf'))
dmattek's avatar
dmattek committed
789
  
dmattek's avatar
dmattek committed
790
  ##### Hierarchical clustering
dmattek's avatar
Added:    
dmattek committed
791
  callModule(clustHier, 'tabClHier', data4clust, data4trajPlot)
dmattek's avatar
dmattek committed
792
793
  
  ##### Sparse hierarchical clustering using sparcl
dmattek's avatar
Added:    
dmattek committed
794
  callModule(clustHierSpar, 'tabClHierSpar', data4clust, data4trajPlot)
dmattek's avatar
dmattek committed
795

dmattek's avatar
Mod:    
dmattek committed
796
  
dmattek's avatar
dmattek committed
797
})