We read in input.scone.csv, which is our file modified (and renamed) from the get.marker.names() function. The K-nearest neighbor generation is derived from the Fast Nearest Neighbors (FNN) R package, within our function Fnn(), which takes as input the “input markers” to be used, along with the concatenated data previously generated, and the desired k. We advise the default selection to the total number of cells in the dataset divided by 100, as has been optimized on existing mass cytometry datasets. The output of this function is a matrix of each cell and the identity of its k-nearest neighbors, in terms of its row number in the dataset used here as input.
library(Sconify)
# Markers from the user-generated excel file
marker.file <- system.file('extdata', 'markers.csv', package = "Sconify")
markers <- ParseMarkers(marker.file)
# How to convert your excel sheet into vector of static and functional markers
markers
## $input
## [1] "CD3(Cd110)Di" "CD3(Cd111)Di"
## [3] "CD3(Cd112)Di" "CD235-61-7-15(In113)Di"
## [5] "CD3(Cd114)Di" "CD45(In115)Di"
## [7] "CD19(Nd142)Di" "CD22(Nd143)Di"
## [9] "IgD(Nd145)Di" "CD79b(Nd146)Di"
## [11] "CD20(Sm147)Di" "CD34(Nd148)Di"
## [13] "CD179a(Sm149)Di" "CD72(Eu151)Di"
## [15] "IgM(Eu153)Di" "Kappa(Sm154)Di"
## [17] "CD10(Gd156)Di" "Lambda(Gd157)Di"
## [19] "CD24(Dy161)Di" "TdT(Dy163)Di"
## [21] "Rag1(Dy164)Di" "PreBCR(Ho165)Di"
## [23] "CD43(Er167)Di" "CD38(Er168)Di"
## [25] "CD40(Er170)Di" "CD33(Yb173)Di"
## [27] "HLA-DR(Yb174)Di"
##
## $functional
## [1] "pCrkL(Lu175)Di" "pCREB(Yb176)Di" "pBTK(Yb171)Di"
## [4] "pS6(Yb172)Di" "cPARP(La139)Di" "pPLCg2(Pr141)Di"
## [7] "pSrc(Nd144)Di" "Ki67(Sm152)Di" "pErk12(Gd155)Di"
## [10] "pSTAT3(Gd158)Di" "pAKT(Tb159)Di" "pBLNK(Gd160)Di"
## [13] "pP38(Tm169)Di" "pSTAT5(Nd150)Di" "pSyk(Dy162)Di"
## [16] "tIkBa(Er166)Di"
# Get the particular markers to be used as knn and knn statistics input
input.markers <- markers[[1]]
funct.markers <- markers[[2]]
# Selection of the k. See "Finding Ideal K" vignette
k <- 30
# The built-in scone functions
wand.nn <- Fnn(cell.df = wand.combined, input.markers = input.markers, k = k)
# Cell identity is in rows, k-nearest neighbors are columns
# List of 2 includes the cell identity of each nn,
# and the euclidean distance between
# itself and the cell of interest
# Indices
str(wand.nn[[1]])
## int [1:1000, 1:30] 871 628 147 351 895 994 478 2 635 625 ...
wand.nn[[1]][1:20, 1:10]
## [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
## [1,] 871 628 773 436 735 155 804 798 910 235
## [2,] 628 8 922 65 842 153 839 88 310 156
## [3,] 147 315 661 655 102 285 609 905 922 888
## [4,] 351 231 740 671 983 820 468 341 993 875
## [5,] 895 528 232 579 981 930 700 167 970 743
## [6,] 994 943 262 947 831 909 952 789 979 318
## [7,] 478 666 274 914 928 772 30 925 577 26
## [8,] 2 310 409 922 88 284 628 153 507 842
## [9,] 635 485 36 173 920 412 539 398 265 694
## [10,] 625 202 770 329 264 396 609 123 253 580
## [11,] 832 533 388 339 88 730 555 295 463 550
## [12,] 971 89 946 811 947 179 292 161 589 104
## [13,] 409 510 622 285 533 460 449 288 467 628
## [14,] 880 40 796 772 256 797 363 94 767 928
## [15,] 163 504 337 772 464 51 675 548 706 261
## [16,] 792 970 790 385 743 499 547 579 180 895
## [17,] 896 774 607 332 112 499 970 895 260 475
## [18,] 218 774 272 499 701 743 607 242 492 358
## [19,] 389 399 612 516 407 587 45 586 119 676
## [20,] 996 781 148 339 153 832 735 467 773 529
# Distance
str(wand.nn[[2]])
## num [1:1000, 1:30] 2.76 2.41 3.17 2.86 3.51 ...
wand.nn[[2]][1:20, 1:10]
## [,1] [,2] [,3] [,4] [,5] [,6] [,7]
## [1,] 2.755977 2.766653 2.926942 2.964926 2.979968 2.985558 3.106700
## [2,] 2.407407 2.775803 2.815737 2.919770 3.008812 3.015529 3.068766
## [3,] 3.169344 3.289579 3.307120 3.394217 3.465452 3.473550 3.498359
## [4,] 2.864089 2.953933 2.978737 2.993218 2.997617 3.205448 3.217035
## [5,] 3.511983 3.600126 3.710470 3.795016 3.807232 3.881188 3.903024
## [6,] 1.991096 2.139081 2.595782 2.607109 2.774993 2.812816 2.823113
## [7,] 4.074958 4.114655 4.159029 4.167930 4.238503 4.339856 4.674532
## [8,] 2.775803 2.900045 3.019769 3.145088 3.195962 3.221582 3.267354
## [9,] 3.392699 3.467530 3.586636 3.682744 3.800792 3.823090 3.888237
## [10,] 3.052123 3.776271 3.863842 4.128827 4.157861 4.233044 4.238328
## [11,] 2.435656 2.630898 2.637653 2.701264 2.750633 2.854395 2.907796
## [12,] 2.343261 2.696411 2.809258 2.967441 3.159903 3.174668 3.275361
## [13,] 3.584366 3.735117 3.786670 3.864378 3.879198 3.916573 3.931665
## [14,] 2.948288 3.363588 3.490019 3.694749 3.748988 3.773205 3.778793
## [15,] 3.650100 3.678814 3.708406 3.713332 3.717871 3.740663 3.782321
## [16,] 3.708601 3.783014 3.896379 3.960447 4.028409 4.052049 4.073326
## [17,] 3.943752 4.112826 4.170143 4.332826 4.336712 4.391361 4.393850
## [18,] 4.304730 4.468533 4.497335 4.541565 4.577728 4.847785 4.857236
## [19,] 3.810662 4.029647 4.324104 4.352560 4.473270 4.751188 4.770297
## [20,] 2.990672 3.125352 3.158027 3.190994 3.211358 3.323551 3.338076
## [,8] [,9] [,10]
## [1,] 3.128773 3.145789 3.310189
## [2,] 3.071325 3.072693 3.195735
## [3,] 3.515097 3.608535 3.609777
## [4,] 3.223906 3.291146 3.331427
## [5,] 3.994113 4.012849 4.154509
## [6,] 2.841366 2.908637 2.978875
## [7,] 4.694901 4.753449 4.758575
## [8,] 3.354773 3.411198 3.460021
## [9,] 4.024737 4.049032 4.106147
## [10,] 4.247532 4.266171 4.270631
## [11,] 2.914027 2.919281 2.930220
## [12,] 3.368900 3.462550 3.568787
## [13,] 3.935371 3.963441 3.963589
## [14,] 3.790889 3.883496 3.891902
## [15,] 3.816347 3.848661 3.906551
## [16,] 4.075601 4.156030 4.205589
## [17,] 4.493268 4.514198 4.523136
## [18,] 4.895267 4.932999 4.938203
## [19,] 4.954891 5.079034 5.148119
## [20,] 3.375435 3.388791 3.397870
This function iterates through each KNN, and performs a series of calculations. The first is fold change values for each maker per KNN, where the user chooses whether this will be based on medians or means. The second is a statistical test, where the user chooses t test or Mann-Whitney U test. I prefer the latter, because it does not assume any properties of the distributions. Of note, the p values are adjusted for false discovery rate, and therefore are called q values in the output of this function. The user also inputs a threshold parameter (default 0.05), where the fold change values will only be shown if the corresponding statistical test returns a q value below said threshold. Finally, the “multiple.donor.compare” option, if set to TRUE will perform a t test based on the mean per-marker values of each donor. This is to allow the user to make comparisons across replicates or multiple donors if that is relevant to the user’s biological questions. This function returns a matrix of cells by computed values (change and statistical test results, labeled either marker.change or marker.qvalue). This matrix is intermediate, as it gets concatenated with the original input matrix in the post-processing step (see the relevant vignette). We show the code and the output below. See the post-processing vignette, where we show how this gets combined with the input data, and additional analysis is performed.
wand.scone <- SconeValues(nn.matrix = wand.nn,
cell.data = wand.combined,
scone.markers = funct.markers,
unstim = "basal")
wand.scone
## # A tibble: 1,000 x 34
## `pCrkL(Lu175)Di~ `pCREB(Yb176)Di~ `pBTK(Yb171)Di.~ `pS6(Yb172)Di.I~
## <dbl> <dbl> <dbl> <dbl>
## 1 1 1 0.991 0.907
## 2 1 1 0.991 0.951
## 3 1 1 0.991 0.948
## 4 1 1 0.991 0.977
## 5 1 1 0.991 0.957
## 6 1 1 0.991 0.948
## 7 1 1 0.991 0.857
## 8 1 1 0.991 0.907
## 9 1 1 0.991 0.852
## 10 1 1 0.554 0.978
## # ... with 990 more rows, and 30 more variables:
## # `cPARP(La139)Di.IL7.qvalue` <dbl>, `pPLCg2(Pr141)Di.IL7.qvalue` <dbl>,
## # `pSrc(Nd144)Di.IL7.qvalue` <dbl>, `Ki67(Sm152)Di.IL7.qvalue` <dbl>,
## # `pErk12(Gd155)Di.IL7.qvalue` <dbl>,
## # `pSTAT3(Gd158)Di.IL7.qvalue` <dbl>, `pAKT(Tb159)Di.IL7.qvalue` <dbl>,
## # `pBLNK(Gd160)Di.IL7.qvalue` <dbl>, `pP38(Tm169)Di.IL7.qvalue` <dbl>,
## # `pSTAT5(Nd150)Di.IL7.qvalue` <dbl>, `pSyk(Dy162)Di.IL7.qvalue` <dbl>,
## # `tIkBa(Er166)Di.IL7.qvalue` <dbl>, `pCrkL(Lu175)Di.IL7.change` <dbl>,
## # `pCREB(Yb176)Di.IL7.change` <dbl>, `pBTK(Yb171)Di.IL7.change` <dbl>,
## # `pS6(Yb172)Di.IL7.change` <dbl>, `cPARP(La139)Di.IL7.change` <dbl>,
## # `pPLCg2(Pr141)Di.IL7.change` <dbl>, `pSrc(Nd144)Di.IL7.change` <dbl>,
## # `Ki67(Sm152)Di.IL7.change` <dbl>, `pErk12(Gd155)Di.IL7.change` <dbl>,
## # `pSTAT3(Gd158)Di.IL7.change` <dbl>, `pAKT(Tb159)Di.IL7.change` <dbl>,
## # `pBLNK(Gd160)Di.IL7.change` <dbl>, `pP38(Tm169)Di.IL7.change` <dbl>,
## # `pSTAT5(Nd150)Di.IL7.change` <dbl>, `pSyk(Dy162)Di.IL7.change` <dbl>,
## # `tIkBa(Er166)Di.IL7.change` <dbl>, IL7.fraction.cond.2 <dbl>,
## # density <dbl>
If one wants to export KNN data to perform other statistics not available in this package, then I provide a function that produces a list of each cell identity in the original input data matrix, and a matrix of all cells x features of its KNN.
I also provide a function to find the KNN density estimation independently of the rest of the “scone.values” analysis, to save time if density is all the user wants. With this density estimation, one can perform interesting analysis, ranging from understanding phenotypic density changes along a developmental progression (see post-processing vignette for an example), to trying out density-based binning methods (eg. X-shift). Of note, this density is specifically one divided by the aveage distance to k-nearest neighbors. This specific measure is related to the Shannon Entropy estimate of that point on the manifold (https://hal.archives-ouvertes.fr/hal-01068081/document).
I use this metric to avoid the unusual properties of the volume of a sphere as it increases in dimensions (https://en.wikipedia.org/wiki/Volume_of_an_n-ball). This being said, one can modify this vector to be such a density estimation (example http://www.cs.haifa.ac.il/~rita/ml_course/lectures_old/KNN.pdf), by treating the distance to knn as the radius of a n-dimensional sphere and incoroprating said volume accordingly.
An individual with basic programming skills can iterate through these elements to perform the statistics of one’s choosing. Examples would include per-KNN regression and classification, or feature imputation. The additional functionality is shown below, with the example knn.list in the package being the first ten instances:
# Constructs KNN list, computes KNN density estimation
wand.knn.list <- MakeKnnList(cell.data = wand.combined, nn.matrix = wand.nn)
wand.knn.list[[8]]
## # A tibble: 30 x 51
## `CD3(Cd110)Di` `CD3(Cd111)Di` `CD3(Cd112)Di` `CD235-61-7-15(~
## <dbl> <dbl> <dbl> <dbl>
## 1 -0.362 -0.423 -1.01 -0.787
## 2 -0.0595 -0.264 -0.454 -1.16
## 3 -0.106 0.388 -0.143 0.665
## 4 -0.0160 -0.136 -0.204 -0.392
## 5 -0.180 -0.255 -0.0147 -0.455
## 6 0.622 1.01 -0.262 -0.427
## 7 -0.0179 -0.0305 -0.663 -0.897
## 8 -0.352 -0.137 -0.458 -0.859
## 9 -0.659 1.29 -0.955 -0.209
## 10 -0.388 -0.839 -0.392 0.560
## # ... with 20 more rows, and 47 more variables: `CD3(Cd114)Di` <dbl>,
## # `CD45(In115)Di` <dbl>, `CD19(Nd142)Di` <dbl>, `CD22(Nd143)Di` <dbl>,
## # `IgD(Nd145)Di` <dbl>, `CD79b(Nd146)Di` <dbl>, `CD20(Sm147)Di` <dbl>,
## # `CD34(Nd148)Di` <dbl>, `CD179a(Sm149)Di` <dbl>, `CD72(Eu151)Di` <dbl>,
## # `IgM(Eu153)Di` <dbl>, `Kappa(Sm154)Di` <dbl>, `CD10(Gd156)Di` <dbl>,
## # `Lambda(Gd157)Di` <dbl>, `CD24(Dy161)Di` <dbl>, `TdT(Dy163)Di` <dbl>,
## # `Rag1(Dy164)Di` <dbl>, `PreBCR(Ho165)Di` <dbl>, `CD43(Er167)Di` <dbl>,
## # `CD38(Er168)Di` <dbl>, `CD40(Er170)Di` <dbl>, `CD33(Yb173)Di` <dbl>,
## # `HLA-DR(Yb174)Di` <dbl>, Time <dbl>, Cell_length <dbl>,
## # `cPARP(La139)Di` <dbl>, `pPLCg2(Pr141)Di` <dbl>,
## # `pSrc(Nd144)Di` <dbl>, `pSTAT5(Nd150)Di` <dbl>, `Ki67(Sm152)Di` <dbl>,
## # `pErk12(Gd155)Di` <dbl>, `pSTAT3(Gd158)Di` <dbl>,
## # `pAKT(Tb159)Di` <dbl>, `pBLNK(Gd160)Di` <dbl>, `pSyk(Dy162)Di` <dbl>,
## # `tIkBa(Er166)Di` <dbl>, `pP38(Tm169)Di` <dbl>, `pBTK(Yb171)Di` <dbl>,
## # `pS6(Yb172)Di` <dbl>, `pCrkL(Lu175)Di` <dbl>, `pCREB(Yb176)Di` <dbl>,
## # `DNA1(Ir191)Di` <dbl>, `DNA2(Ir193)Di` <dbl>,
## # `Viability1(Pt195)Di` <dbl>, `Viability2(Pt196)Di` <dbl>,
## # wanderlust <dbl>, condition <chr>
# Finds the KNN density estimation for each cell, ordered by column, in the
# original data matrix
wand.knn.density <- GetKnnDe(nn.matrix = wand.nn)
str(wand.knn.density)
## num [1:1000] 0.302 0.31 0.266 0.295 0.235 ...