Analysis of the wave2 Beat-AML venetoclax data using pamr
Shanshan Pei
3/1/2018
- Load data
- Optional pre-processing
- Non-interactive step-by-step analysis
- Train the classifeir and cross-validate
- Plot the cross-validated error curves
- Compute the confusion matrix for a particular model
- Plot the cross-validated class probabilities by class
- Plot the class centroids
- Make a gene plot of the most significant genes
- List the significant genes
- Try heterogeneity analysis, with class “BL” taken to be the normal group
- Identify better threshold scalings
- Interactive analysis
- References
- Session Info
Load data
## Loading required package: cluster## Loading required package: survival##
## Read in 2308 genes
## Read in 63 samples
## Read in 63 sample labels
##
## Make sure these figures are correct!!Optional pre-processing
Impute missing expression values (optional)
Bacth adjust (optional)
Output adjusted data (optional)
Non-interactive step-by-step analysis
Train the classifeir and cross-validate
## 123456789101112131415161718192021222324252627282930## Call:
## pamr.train(data = khan.data)
## threshold nonzero errors
## 1 0.000 2308 2
## 2 0.262 2289 1
## 3 0.524 2145 1
## 4 0.786 1878 0
## 5 1.048 1494 0
## 6 1.309 1137 0
## 7 1.571 853 0
## 8 1.833 609 0
## 9 2.095 436 0
## 10 2.357 330 0
## 11 2.619 244 0
## 12 2.881 193 0
## 13 3.143 151 0
## 14 3.404 107 0
## 15 3.666 87 0
## 16 3.928 68 0
## 17 4.190 52 0
## 18 4.452 39 0
## 19 4.714 32 1
## 20 4.976 23 4
## 21 5.238 21 11
## 22 5.499 16 14
## 23 5.761 11 16
## 24 6.023 10 18
## 25 6.285 9 21
## 26 6.547 7 21
## 27 6.809 5 23
## 28 7.071 4 39
## 29 7.333 1 40
## 30 7.595 0 40## 1234Fold 1 :123456789101112131415161718192021222324252627282930
## Fold 2 :123456789101112131415161718192021222324252627282930
## Fold 3 :123456789101112131415161718192021222324252627282930
## Fold 4 :123456789101112131415161718192021222324252627282930
## Fold 5 :123456789101112131415161718192021222324252627282930
## Fold 6 :123456789101112131415161718192021222324252627282930
## Fold 7 :123456789101112131415161718192021222324252627282930
## Fold 8 :123456789101112131415161718192021222324252627282930## Call:
## pamr.cv(fit = khan.train, data = khan.data)
## threshold nonzero errors
## 1 0.000 2308 2
## 2 0.262 2289 2
## 3 0.524 2145 2
## 4 0.786 1878 2
## 5 1.048 1494 2
## 6 1.309 1137 2
## 7 1.571 853 1
## 8 1.833 609 1
## 9 2.095 436 1
## 10 2.357 330 0
## 11 2.619 244 0
## 12 2.881 193 0
## 13 3.143 151 0
## 14 3.404 107 0
## 15 3.666 87 0
## 16 3.928 68 0
## 17 4.190 52 0
## 18 4.452 39 1
## 19 4.714 32 1
## 20 4.976 23 4
## 21 5.238 21 12
## 22 5.499 16 14
## 23 5.761 11 15
## 24 6.023 10 18
## 25 6.285 9 21
## 26 6.547 7 21
## 27 6.809 5 25
## 28 7.071 4 32
## 29 7.333 1 37
## 30 7.595 0 37Plot the cross-validated error curves
Compute the confusion matrix for a particular model
## BL EWS NB RMS Class Error rate
## BL 8 0 0 0 0
## EWS 0 23 0 0 0
## NB 0 0 12 0 0
## RMS 0 0 0 20 0
## Overall error rate= 0Plot the cross-validated class probabilities by class
Plot the class centroids
Make a gene plot of the most significant genes
List the significant genes
## id BL-score EWS-score NB-score RMS-score
## [1,] GENE1389 -0.0629 0.5972 0 0
## [2,] GENE1955 0 0 0 0.5729
## [3,] GENE187 0 -0.0576 0 0.5631
## [4,] GENE2050 0 -0.5301 0 0
## [5,] GENE246 0 0.5219 0 0
## [6,] GENE2198 -0.5083 0 0 0
## [7,] GENE509 0 0 0 0.4803
## [8,] GENE2046 0 0 0 0.4688
## [9,] GENE2022 -0.4635 0 0 0
## [10,] GENE851 -0.4424 0 0 0
## [11,] GENE1319 0 0.426 0 0
## [12,] GENE1003 0 0 0 0.4136
## [13,] GENE1954 0 0.3966 0 0
## [14,] GENE1 -0.3915 0 0 0
## [15,] GENE842 0 0 -0.3641 0
## [16,] GENE1708 0 0.3226 0 0
## [17,] GENE129 0 0 0 0.3107
## [18,] GENE1427 -0.3075 0 0 0
## [19,] GENE566 0 0.2897 0 0
## [20,] GENE545 0 0.2747 0 0
## [21,] GENE836 0.2693 0 0 0
## [22,] GENE1645 0 0.2659 0 0
## [23,] GENE107 -0.2552 0 -0.0238 0
## [24,] GENE2162 -0.2552 0 0 0
## [25,] GENE255 0 0 0.2441 0
## [26,] GENE846 0.2402 0 0 0
## [27,] GENE1055 0 0 0 0.2326
## [28,] GENE819 0 0 -0.2296 0
## [29,] GENE554 0 0 0 0.2292
## [30,] GENE742 0 0 0.2248 0
## [31,] GENE1066 -0.1943 0 0 0
## [32,] GENE1886 -0.1932 0 0 0
## [33,] GENE174 0 0 0 0.1917
## [34,] GENE1911 0 0 0 0.1455
## [35,] GENE1764 0 0 0.1424 0
## [36,] GENE1194 0 0 0 0.1296
## [37,] GENE1916 0.1192 0 0 0
## [38,] GENE1750 -0.1166 0 0 0
## [39,] GENE368 0 0.1122 0 0
## [40,] GENE783 0.0981 0 0 0
## [41,] GENE603 0 0 0 0.0896
## [42,] GENE1723 0 0 0 0.0836
## [43,] GENE544 -0.0818 0 0 0
## [44,] GENE1896 0 0 0 0.0745
## [45,] GENE2 0 0 0 0.0667
## [46,] GENE248 -0.0665 0 0 0
## [47,] GENE1601 0 0 0.0596 0
## [48,] GENE338 0 0 0 0.0538
## [49,] GENE1799 0 -0.0469 0 0
## [50,] GENE433 0 0 0 0.0439
## [51,] GENE1980 -0.0265 0.0374 0 0
## [52,] GENE1105 0 0 0 0.0354
## [53,] GENE2166 -0.0315 0 0 0
## [54,] GENE2303 -0.0305 0 0 0
## [55,] GENE123 0.0302 0 0 0
## [56,] GENE1387 0.0276 0 0 0
## [57,] GENE2146 0 0 0 0.0229
## [58,] GENE788 -0.0225 0 0 0
## [59,] GENE335 0.0164 0 0 0
## [60,] GENE1207 0 0 0 0.016
## [61,] GENE567 -0.0112 0 0 0
## [62,] GENE1353 0 0 0 0.0084
## [63,] GENE714 0 0 0 0.0081
## [64,] GENE2144 0 0 0.0024 0
## [65,] GENE910 0 0 0 3e-04Try heterogeneity analysis, with class “BL” taken to be the normal group
## 123456789101112131415161718192021222324252627282930## 1234Fold 1 :123456789101112131415161718192021222324252627282930
## Fold 2 :123456789101112131415161718192021222324252627282930
## Fold 3 :123456789101112131415161718192021222324252627282930
## Fold 4 :123456789101112131415161718192021222324252627282930
## Fold 5 :123456789101112131415161718192021222324252627282930
## Fold 6 :123456789101112131415161718192021222324252627282930
## Fold 7 :123456789101112131415161718192021222324252627282930
## Fold 8 :123456789101112131415161718192021222324252627282930Identify better threshold scalings
## Initial errors: 2.60000 0.10000 3.53333 2.16667 Roc 9.66305
## Update 1
## 123456789101112131415161718192021222324252627282930
## Errors 3.03333 0.10000 4.06667 3.86667 Roc 9.39042
## Update 2
## 123456789101112131415161718192021222324252627282930
## Errors 3.40000 0.06667 4.50000 5.50000 Roc 9.16457
## Update 3
## 123456789101112131415161718192021222324252627282930
## Errors 3.40000 0.10000 4.36667 4.13333 Roc 8.95462
## Update 4
## 123456789101112131415161718192021222324252627282930
## Errors 3.70000 0.10000 4.66667 5.73333 Roc 8.56622
## Update 5
## 123456789101112131415161718192021222324252627282930
## Errors 3.70000 0.26667 4.56667 4.16667 Roc 10.05755
## Update 6
## 123456789101112131415161718192021222324252627282930
## Errors 4.00000 0.23333 4.80000 5.73333 Roc 9.23474
## Update 7
## 123456789101112131415161718192021222324252627282930
## Errors 4.00000 0.36667 4.80000 4.16667 Roc 10.91507
## Update 8
## 123456789101112131415161718192021222324252627282930
## Errors 4.40000 0.30000 4.83333 5.73333 Roc 10.09324
## Update 9
## 123456789101112131415161718192021222324252627282930
## Errors 4.40000 0.40000 4.83333 4.16667 Roc 11.41245
## Update 10
## 123456789101112131415161718192021222324252627282930
## Errors 4.73333 0.36667 4.86667 5.73333 Roc 10.92883## 123456789101112131415161718192021222324252627282930## 1234Fold 1 :123456789101112131415161718192021222324252627282930
## Fold 2 :123456789101112131415161718192021222324252627282930
## Fold 3 :123456789101112131415161718192021222324252627282930
## Fold 4 :123456789101112131415161718192021222324252627282930
## Fold 5 :123456789101112131415161718192021222324252627282930
## Fold 6 :123456789101112131415161718192021222324252627282930
## Fold 7 :123456789101112131415161718192021222324252627282930
## Fold 8 :123456789101112131415161718192021222324252627282930Interactive analysis
Begin by typing “1” to select pamr.train, and then after that computation is done, pick “2” for pamr.cv Typically, you would go through steps 3 through 8, to generate plots and gene lists. Along the way, in some of the steps you are asked for a threshold value: this value you choose visually from the plot created by pamr.plotcv. Menu Choice 9 is optional.
References
The tutorial 1
The tutorial 2
The PNAS paper containing Khan dataset
The Khan dataset
The JCO paper implemented the pamr approach
Session Info
sessionInfo()## R version 3.3.2 (2016-10-31)
## Platform: x86_64-apple-darwin13.4.0 (64-bit)
## Running under: macOS 10.13.3
##
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] pamr_1.55 survival_2.41-3 cluster_2.0.6
##
## loaded via a namespace (and not attached):
## [1] Rcpp_0.12.14 lattice_0.20-35 digest_0.6.13 rprojroot_1.3-2 grid_3.3.2 backports_1.1.2 magrittr_1.5 evaluate_0.10.1 stringi_1.1.6 Matrix_1.2-12 rmarkdown_1.9 splines_3.3.2 tools_3.3.2 stringr_1.3.0 yaml_2.1.16 htmltools_0.3.6 knitr_1.20