title

author

date

output

PCA_UMAP

N. Alcala

9/6/2022

html_document

keep_md
true

R Code to compare embeddings on TCGA expression data

load libraries

library(tidyverse)
library(SummarizedExperiment)
library(patchwork)
library(umap)
library(ade4)
library(keras)
library(cluster)
library(readxl)

Load expression data from TCGA

load("AIML-WG/data/TCGA-MESOTranscriptome_ProfilingFri_Jul_29_11:39:27_2022.RData")
MESO.Rnaseq.SE = data
# clean histology
MESO.Rnaseq.SE$primary_diagnosis[MESO.Rnaseq.SE$primary_diagnosis=="Epithelioid mesothelioma, malignant"] = "MME"
MESO.Rnaseq.SE$primary_diagnosis[MESO.Rnaseq.SE$primary_diagnosis=="Mesothelioma, biphasic, malignant"] = "MMB"
MESO.Rnaseq.SE$primary_diagnosis[MESO.Rnaseq.SE$primary_diagnosis%in%c("Mesothelioma, malignant","Fibrous mesothelioma, malignant")] = NA
table(MESO.Rnaseq.SE$primary_diagnosis)

## 
## MMB MME 
##  22  58

load("AIML-WG/data/TCGA-BRCATranscriptome_ProfilingFri_Jul_29_12:11:29_2022.RData")
BRCA.Rnaseq.SE = data

load additional clinical data from TCGA papers

MESO_S1 = read_xlsx("AIML-WG/data/TCGA_TableS1.xlsx",sheet=2)
# match barcodes 
MESO_S1$barcode = MESO.Rnaseq.SE$barcode[sapply(MESO_S1$TCGA_barcode, function(x) grep(x = MESO.Rnaseq.SE$barcode,pattern = x) )]
MESOorder = sapply(MESO.Rnaseq.SE$barcode,function(x){res=which(MESO_S1$barcode==x);if(length(res)==0){res=NA};return(res)})
# remove samples not reported in study
MESO.Rnaseq.SE = MESO.Rnaseq.SE[,!is.na(MESOorder)]
MESO_S1 = MESO_S1[MESOorder[!is.na(MESOorder)],]

make survival groups

## MESO
MESO.Rnaseq.SE$survival_group = cut(MESO.Rnaseq.SE$days_to_death,breaks = quantile(MESO.Rnaseq.SE$days_to_death,c(0,0.33,0.67,1),na.rm=T))
MESO.Rnaseq.SE$survival_group[which(as.numeric(MESO.Rnaseq.SE$survival_group)<3 & MESO.Rnaseq.SE$vital_status=="Alive")] = NA

## BRCA
BRCA.Rnaseq.SE$paper_days_to_death = as.numeric(BRCA.Rnaseq.SE$paper_days_to_death)

## Warning: NAs introduced by coercion

BRCA.Rnaseq.SE$survival_group = cut(BRCA.Rnaseq.SE$paper_days_to_death,breaks = quantile(BRCA.Rnaseq.SE$paper_days_to_death,c(0,0.33,0.67,1),na.rm=T))
BRCA.Rnaseq.SE$survival_group[which(as.numeric(BRCA.Rnaseq.SE$survival_group)<3 & BRCA.Rnaseq.SE$paper_vital_status=="Alive")] = NA

Get expression in TPM

MESO.Rnaseq.expr.tpm = assay(MESO.Rnaseq.SE,"tpm_unstrand")
BRCA.Rnaseq.expr.tpm = assay(BRCA.Rnaseq.SE,"tpm_unstrand")

MESO.Rnaseq.expr.tpm.logmat = t(log(MESO.Rnaseq.expr.tpm[order(apply(MESO.Rnaseq.expr.tpm,1,var),decreasing = T)[1:5000],]+1))
BRCA.Rnaseq.expr.tpm.logmat = t(log(BRCA.Rnaseq.expr.tpm[order(apply(BRCA.Rnaseq.expr.tpm,1,var),decreasing = T)[1:5000],]+1))

Analysis

PCA of log(tpm+1) expression of 5000 most variable genes

pca.MESO = dudi.pca(MESO.Rnaseq.expr.tpm.logmat,scannf = F,nf=10)
pca.BRCA = dudi.pca(BRCA.Rnaseq.expr.tpm.logmat,scannf = F,nf=10)

UMAP of log(tpm+1) expression of 5000 most variable genes

umap.MESO = umap(MESO.Rnaseq.expr.tpm.logmat)
umap.BRCA = umap(BRCA.Rnaseq.expr.tpm.logmat)

Autoencoder of of log(tpm+1) expression of 5000 most variable genes

# set models
modelMESO <- keras_model_sequential()
modelBRCA <- keras_model_sequential()
modelMESO %>%
  layer_dense(units = 50, activation = "linear", input_shape = 5000) %>%
  layer_dense(units = 2, activation = "linear", name = "bottleneck") %>%
  layer_dense(units = 50, activation = "linear") %>%
  layer_dense(units = 5000)
modelBRCA %>%
  layer_dense(units = 50, activation = "linear", input_shape = 5000) %>%
  layer_dense(units = 2, activation = "linear", name = "bottleneck") %>%
  layer_dense(units = 50, activation = "linear") %>%
  layer_dense(units = 5000)
# view model layers
summary(modelMESO)

## Model: "sequential"
## ________________________________________________________________________________
## Layer (type)                        Output Shape                    Param #     
## ================================================================================
## dense_2 (Dense)                     (None, 50)                      250050      
## ________________________________________________________________________________
## bottleneck (Dense)                  (None, 2)                       102         
## ________________________________________________________________________________
## dense_1 (Dense)                     (None, 50)                      150         
## ________________________________________________________________________________
## dense (Dense)                       (None, 5000)                    255000      
## ================================================================================
## Total params: 505,302
## Trainable params: 505,302
## Non-trainable params: 0
## ________________________________________________________________________________

# compile model
modelMESO %>% compile(loss = "mean_squared_error", optimizer = "rmsprop")
modelBRCA %>% compile(loss = "mean_squared_error", optimizer = "rmsprop")

# fit model
modelMESO %>% fit(x = MESO.Rnaseq.expr.tpm.logmat, y = MESO.Rnaseq.expr.tpm.logmat, epochs = 200,verbose = 1)
modelBRCA %>% fit(x = BRCA.Rnaseq.expr.tpm.logmat, y = BRCA.Rnaseq.expr.tpm.logmat, epochs = 200,verbose = 1)

# evaluate the performance of the model
mse.aeMESO <- evaluate(modelMESO, MESO.Rnaseq.expr.tpm.logmat, MESO.Rnaseq.expr.tpm.logmat)
mse.aeBRCA <- evaluate(modelBRCA, BRCA.Rnaseq.expr.tpm.logmat, BRCA.Rnaseq.expr.tpm.logmat)
mse.aeMESO

##    loss 
## 1.61935

mse.aeBRCA

##      loss 
## 0.5589278

intermediate_layer_modelMESO <- keras_model(inputs = modelMESO$input, outputs = get_layer(modelMESO, "bottleneck")$output)
intermediate_layer_modelBRCA <- keras_model(inputs = modelBRCA$input, outputs = get_layer(modelBRCA, "bottleneck")$output)
AE.MESO <- predict(intermediate_layer_modelMESO, MESO.Rnaseq.expr.tpm.logmat)
AE.BRCA <- predict(intermediate_layer_modelBRCA, BRCA.Rnaseq.expr.tpm.logmat)

plot

We create tibble objects for plotting

embeddings.MESO.tib = tibble(PCA=pca.MESO$li, UMAP = umap.MESO$layout, AE= AE.MESO, data=as.data.frame(colData(MESO.Rnaseq.SE) ) )
embeddings.BRCA.tib = tibble(PCA=pca.BRCA$li, UMAP = umap.BRCA$layout, AE=AE.BRCA, data=as.data.frame(colData(BRCA.Rnaseq.SE) ) )

Histopathology

Plot first 2 PCA axes. For MESO, we use the histopathological types as colors; for BRCA, we use the molecular subtypes

ggMESO_PCA  = ggplot(embeddings.MESO.tib, aes(x=PCA$Axis1,y=PCA$Axis2 ,col=MESO.Rnaseq.SE$primary_diagnosis)) + geom_point() + theme_bw() + xlab("PC1") + ylab("PC2")
ggBRCA_PCA  = ggplot(embeddings.BRCA.tib, aes(x=PCA$Axis1,y=PCA$Axis2 ,col=BRCA.Rnaseq.SE$paper_BRCA_Pathology)) + geom_point() + theme_bw() + xlab("PC1") + ylab("PC2")

We plot the UMAP

ggMESO_UMAP = ggplot(embeddings.MESO.tib, aes(x=UMAP[,1],y=UMAP[,2] ,col=MESO.Rnaseq.SE$primary_diagnosis)) + geom_point() + theme_bw() + xlab("UMAP1") + ylab("UMAP2")
ggBRCA_UMAP = ggplot(embeddings.BRCA.tib, aes(x=UMAP[,1],y=UMAP[,2] ,col=BRCA.Rnaseq.SE$paper_BRCA_Pathology)) + geom_point() + theme_bw() + xlab("UMAP1") + ylab("UMAP2")

We plot the autoencoders

ggMESO_AE = ggplot(embeddings.MESO.tib, aes(x=AE[,1],y=AE[,2] ,col=MESO.Rnaseq.SE$primary_diagnosis)) + geom_point() + theme_bw() + xlab("AE1") + ylab("AE2")
ggBRCA_AE = ggplot(embeddings.BRCA.tib, aes(x=AE[,1],y=AE[,2] ,col=BRCA.Rnaseq.SE$paper_BRCA_Pathology)) + geom_point() + theme_bw() + xlab("AE1") + ylab("AE2")

We arrange the plots with patchwork

((ggMESO_PCA+guides(color="none"))+ ggMESO_UMAP+guides(color="none")+ (ggMESO_AE + guides(col=guide_legend(title="Histology"))) )/((ggBRCA_PCA+guides(color="none"))+ ggBRCA_UMAP+guides(color="none")+ (ggBRCA_AE + guides(col=guide_legend(title="Histology"))) )

Molecular groups

ggMESO_PCA  = ggplot(embeddings.MESO.tib, aes(x=PCA$Axis1,y=PCA$Axis2 ,col=MESO_S1$iCluster_k4.5types)) + geom_point() + theme_bw() + xlab("PC1") + ylab("PC2")
ggBRCA_PCA  = ggplot(embeddings.BRCA.tib, aes(x=PCA$Axis1,y=PCA$Axis2 ,col=BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50)) + geom_point() + theme_bw() + xlab("PC1") + ylab("PC2")

We plot the UMAP

ggMESO_UMAP = ggplot(embeddings.MESO.tib, aes(x=UMAP[,1],y=UMAP[,2] ,col=MESO_S1$iCluster_k4.5types)) + geom_point() + theme_bw() + xlab("UMAP1") + ylab("UMAP2")
ggBRCA_UMAP = ggplot(embeddings.BRCA.tib, aes(x=UMAP[,1],y=UMAP[,2] ,col=BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50)) + geom_point() + theme_bw() + xlab("UMAP1") + ylab("UMAP2")

We plot the autoencoders

ggMESO_AE = ggplot(embeddings.MESO.tib, aes(x=AE[,1],y=AE[,2] ,col=MESO_S1$iCluster_k4.5types)) + geom_point() + theme_bw() + xlab("AE1") + ylab("AE2")
ggBRCA_AE = ggplot(embeddings.BRCA.tib, aes(x=AE[,1],y=AE[,2] ,col=BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50)) + geom_point() + theme_bw() + xlab("AE1") + ylab("AE2")

We arrange the plots with patchwork

((ggMESO_PCA+guides(color="none"))+ ggMESO_UMAP+guides(color="none")+ (ggMESO_AE + guides(col=guide_legend(title="Molecular group"))) )/((ggBRCA_PCA+guides(color="none"))+ ggBRCA_UMAP+guides(color="none")+ (ggBRCA_AE + guides(col=guide_legend(title="Molecular group"))) )

Clinical groups

ggMESO_PCA  = ggplot(embeddings.MESO.tib, aes(x=PCA$Axis1,y=PCA$Axis2 ,col=MESO.Rnaseq.SE$survival_group)) + geom_point() + theme_bw() + xlab("PC1") + ylab("PC2")
ggBRCA_PCA  = ggplot(embeddings.BRCA.tib, aes(x=PCA$Axis1,y=PCA$Axis2 ,col=BRCA.Rnaseq.SE$survival_group)) + geom_point() + theme_bw() + xlab("PC1") + ylab("PC2")

We plot the UMAP

ggMESO_UMAP = ggplot(embeddings.MESO.tib, aes(x=UMAP[,1],y=UMAP[,2] ,col=MESO.Rnaseq.SE$survival_group)) + geom_point() + theme_bw() + xlab("UMAP1") + ylab("UMAP2")
ggBRCA_UMAP = ggplot(embeddings.BRCA.tib, aes(x=UMAP[,1],y=UMAP[,2] ,col=BRCA.Rnaseq.SE$survival_group)) + geom_point() + theme_bw() + xlab("UMAP1") + ylab("UMAP2")

We plot the autoencoders

ggMESO_AE = ggplot(embeddings.MESO.tib, aes(x=AE[,1],y=AE[,2] ,col=MESO.Rnaseq.SE$survival_group)) + geom_point() + theme_bw() + xlab("AE1") + ylab("AE2")
ggBRCA_AE = ggplot(embeddings.BRCA.tib, aes(x=AE[,1],y=AE[,2] ,col=BRCA.Rnaseq.SE$survival_group)) + geom_point() + theme_bw() + xlab("AE1") + ylab("AE2")

We arrange the plots with patchwork

((ggMESO_PCA+guides(color="none"))+ ggMESO_UMAP+guides(color="none")+ (ggMESO_AE + guides(col=guide_legend(title="Survival group"))) )/((ggBRCA_PCA+guides(color="none"))+ ggBRCA_UMAP+guides(color="none")+ (ggBRCA_AE + guides(col=guide_legend(title="Survival group"))) )

Compute metrics

For histopathological types

## MESO
silMESO.PCA  = silhouette(as.numeric(as.factor(MESO.Rnaseq.SE$primary_diagnosis))[!is.na(MESO.Rnaseq.SE$primary_diagnosis)],
           dist = dist(embeddings.MESO.tib$PCA[!is.na(MESO.Rnaseq.SE$primary_diagnosis),]))
silMESO.UMAP = silhouette(as.numeric(as.factor(MESO.Rnaseq.SE$primary_diagnosis))[!is.na(MESO.Rnaseq.SE$primary_diagnosis)],
           dist = dist(embeddings.MESO.tib$UMAP[!is.na(MESO.Rnaseq.SE$primary_diagnosis),]))
silMESO.AE   = silhouette(as.numeric(as.factor(MESO.Rnaseq.SE$primary_diagnosis))[!is.na(MESO.Rnaseq.SE$primary_diagnosis)],
           dist = dist(embeddings.MESO.tib$AE[!is.na(MESO.Rnaseq.SE$primary_diagnosis),]))
## BRCA
silBRCA.PCA  = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$paper_BRCA_Pathology))[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Pathology)],
           dist = dist(embeddings.BRCA.tib$PCA[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Pathology),]))
silBRCA.UMAP = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$paper_BRCA_Pathology))[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Pathology)],
           dist = dist(embeddings.BRCA.tib$UMAP[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Pathology),]))
silBRCA.AE   = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$paper_BRCA_Pathology))[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Pathology)],
           dist = dist(embeddings.BRCA.tib$AE[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Pathology),]))

For molecular groups

## MESO
silMESO.PCA.mol  = silhouette(as.numeric(as.factor(MESO_S1$iCluster_k4.5types)), dist = dist(embeddings.MESO.tib$PCA))
silMESO.UMAP.mol = silhouette(as.numeric(as.factor(MESO_S1$iCluster_k4.5types)), dist = dist(embeddings.MESO.tib$UMAP))
silMESO.AE.mol   = silhouette(as.numeric(as.factor(MESO_S1$iCluster_k4.5types)), dist = dist(embeddings.MESO.tib$AE))
## BRCA
silBRCA.PCA.mol  = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50))[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50)],
           dist = dist(embeddings.BRCA.tib$PCA[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50),]))
silBRCA.UMAP.mol = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50))[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50)],
           dist = dist(embeddings.BRCA.tib$UMAP[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50),]))
silBRCA.AE.mol   = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50))[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50)],
           dist = dist(embeddings.BRCA.tib$AE[!is.na(BRCA.Rnaseq.SE$paper_BRCA_Subtype_PAM50),]))

For clinical variables

## MESO
silMESO.PCA.clin  = silhouette(as.numeric(MESO.Rnaseq.SE$survival_group)[!is.na(MESO.Rnaseq.SE$survival_group)], dist = dist(embeddings.MESO.tib$PCA[!is.na(MESO.Rnaseq.SE$survival_group),]))
silMESO.UMAP.clin = silhouette(as.numeric(MESO.Rnaseq.SE$survival_group)[!is.na(MESO.Rnaseq.SE$survival_group)], dist = dist(embeddings.MESO.tib$UMAP[!is.na(MESO.Rnaseq.SE$survival_group),]))
silMESO.AE.clin   = silhouette(as.numeric(MESO.Rnaseq.SE$survival_group)[!is.na(MESO.Rnaseq.SE$survival_group)], dist = dist(embeddings.MESO.tib$AE[!is.na(MESO.Rnaseq.SE$survival_group),]))

## BRCA
silBRCA.PCA.clin  = silhouette(as.numeric(BRCA.Rnaseq.SE$survival_group)[!is.na(BRCA.Rnaseq.SE$survival_group)],
           dist = dist(embeddings.BRCA.tib$PCA[!is.na(BRCA.Rnaseq.SE$survival_group),]))
silBRCA.UMAP.clin = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$survival_group))[!is.na(BRCA.Rnaseq.SE$survival_group)],
           dist = dist(embeddings.BRCA.tib$UMAP[!is.na(BRCA.Rnaseq.SE$survival_group),]))
silBRCA.AE.clin   = silhouette(as.numeric(as.factor(BRCA.Rnaseq.SE$survival_group))[!is.na(BRCA.Rnaseq.SE$survival_group)],
           dist = dist(embeddings.BRCA.tib$AE[!is.na(BRCA.Rnaseq.SE$survival_group),]))

Plot metrics

sill = tibble(embedding=rep(c("PCA","UMAP","AE"),2*3),variable=rep(c("Histological types","Molecular groups","Survival groups"),each=3*2), 
              Cohort = rep(rep(c("MESO","BRCA"),each=3),3),
              silhouette=c(mean(silMESO.PCA[,3]),mean(silMESO.UMAP[,3]),mean(silMESO.AE[,3]),
                           mean(silBRCA.PCA[,3]),mean(silBRCA.UMAP[,3]),mean(silBRCA.AE[,3]),
                           mean(silMESO.PCA.mol[,3]),mean(silMESO.UMAP.mol[,3]),mean(silMESO.AE.mol[,3]),
                           mean(silBRCA.PCA.mol[,3]),mean(silBRCA.UMAP.mol[,3]),mean(silBRCA.AE.mol[,3]),
                           mean(silMESO.PCA.clin[,3]),mean(silMESO.UMAP.clin[,3]),mean(silMESO.AE.clin[,3]),
                           mean(silBRCA.PCA.clin[,3]),mean(silBRCA.UMAP.clin[,3]),mean(silBRCA.AE.clin[,3])
                           ))

ggplot(sill, aes(x=embedding,y=silhouette,fill=Cohort) ) + geom_bar(stat="identity",beside=T,position=position_dodge()) + 
  theme_bw() +ylab("Mean silhouette width") + facet_grid(.~variable)

## Warning: Ignoring unknown parameters: beside

Session Info

sessionInfo()

## R version 4.1.2 (2021-11-01)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: CentOS Linux 7 (Core)
## 
## Matrix products: default
## BLAS/LAPACK: /usr/lib64/libopenblasp-r0.3.3.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats4    stats     graphics  grDevices utils     datasets  methods  
## [8] base     
## 
## other attached packages:
##  [1] readxl_1.4.1                cluster_2.1.4              
##  [3] keras_2.9.0                 ade4_1.7-18                
##  [5] umap_0.2.8.0                patchwork_1.1.1            
##  [7] SummarizedExperiment_1.24.0 Biobase_2.54.0             
##  [9] GenomicRanges_1.46.1        GenomeInfoDb_1.30.1        
## [11] IRanges_2.28.0              S4Vectors_0.32.4           
## [13] BiocGenerics_0.40.0         MatrixGenerics_1.6.0       
## [15] matrixStats_0.62.0          forcats_0.5.1              
## [17] stringr_1.4.1               dplyr_1.0.10               
## [19] purrr_0.3.4                 readr_2.1.2                
## [21] tidyr_1.2.1                 tibble_3.1.8               
## [23] ggplot2_3.3.6               tidyverse_1.3.1            
## 
## loaded via a namespace (and not attached):
##  [1] bitops_1.0-7           fs_1.5.2               lubridate_1.8.0       
##  [4] httr_1.4.4             tools_4.1.2            backports_1.4.1       
##  [7] bslib_0.4.0            utf8_1.2.2             R6_2.5.1              
## [10] DBI_1.1.3              colorspace_2.0-3       withr_2.5.0           
## [13] tidyselect_1.1.2       compiler_4.1.2         cli_3.4.1             
## [16] rvest_1.0.3            xml2_1.3.3             DelayedArray_0.20.0   
## [19] labeling_0.4.2         sass_0.4.2             scales_1.2.1          
## [22] askpass_1.1            tfruns_1.5.1           digest_0.6.29         
## [25] rmarkdown_2.16         XVector_0.34.0         base64enc_0.1-3       
## [28] pkgconfig_2.0.3        htmltools_0.5.3        highr_0.9             
## [31] dbplyr_2.2.1           fastmap_1.1.0          rlang_1.0.6           
## [34] rstudioapi_0.14        farver_2.1.1           jquerylib_0.1.4       
## [37] generics_0.1.3         jsonlite_1.8.0         tensorflow_2.9.0      
## [40] RCurl_1.98-1.8         magrittr_2.0.3.9000    GenomeInfoDbData_1.2.7
## [43] Matrix_1.5-1           Rcpp_1.0.9             munsell_0.5.0         
## [46] fansi_1.0.3            reticulate_1.26        lifecycle_1.0.2       
## [49] whisker_0.4            stringi_1.7.8          yaml_2.3.5            
## [52] MASS_7.3-58.1          zlibbioc_1.40.0        grid_4.1.2            
## [55] crayon_1.5.1           lattice_0.20-45        haven_2.5.1           
## [58] hms_1.1.2              zeallot_0.1.0          knitr_1.38            
## [61] pillar_1.8.1           reprex_2.0.2           glue_1.6.2            
## [64] evaluate_0.15          modelr_0.1.9           vctrs_0.4.1           
## [67] png_0.1-7              tzdb_0.3.0             cellranger_1.1.0      
## [70] gtable_0.3.1           openssl_2.0.3          assertthat_0.2.1      
## [73] cachem_1.0.6           xfun_0.33              broom_1.0.1           
## [76] RSpectra_0.16-1        ellipsis_0.3.2

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

PCA_UMAP.md

PCA_UMAP.md

R Code to compare embeddings on TCGA expression data

load libraries

Load expression data from TCGA

load additional clinical data from TCGA papers

make survival groups

Get expression in TPM

Analysis

PCA of log(tpm+1) expression of 5000 most variable genes

UMAP of log(tpm+1) expression of 5000 most variable genes

Autoencoder of of log(tpm+1) expression of 5000 most variable genes

plot

Histopathology

Molecular groups

Clinical groups

Compute metrics

For histopathological types

For molecular groups

For clinical variables

Plot metrics

Session Info

Files

PCA_UMAP.md

Latest commit

History

PCA_UMAP.md

File metadata and controls

R Code to compare embeddings on TCGA expression data

load libraries

Load expression data from TCGA

load additional clinical data from TCGA papers

make survival groups

Get expression in TPM

Analysis

PCA of log(tpm+1) expression of 5000 most variable genes

UMAP of log(tpm+1) expression of 5000 most variable genes

Autoencoder of of log(tpm+1) expression of 5000 most variable genes

plot

Histopathology

Molecular groups

Clinical groups

Compute metrics

For histopathological types

For molecular groups

For clinical variables

Plot metrics

Session Info