Revising examine_pbmc68k_more.R to use k = 7 results.

analysis/examine_pbmc68k_more.R

@@ -1,77 +1,51 @@
-# For paper, highlight results with K = 11.
-#
-# Or maybe highlight K = 7 instead?
-#
+# For paper, highlight results with K = 7.
 library(fastTopics)
 library(cowplot)
 set.seed(1)
-k <- 11
+k <- 7
 load("../data/pbmc_68k.RData")
 rm(counts)
-fit1 <- readRDS("../output/pbmc68k/rds/fit-pbmc68k-em-k=11.rds")$fit
-fit2 <- readRDS("../output/pbmc68k/rds/fit-pbmc68k-scd-ex-k=11.rds")$fit
-lda1 <- readRDS("../output/pbmc68k/rds/lda-pbmc68k-em-k=11.rds")$lda
-lda2 <- readRDS("../output/pbmc68k/rds/lda-pbmc68k-scd-ex-k=11.rds")$lda
+fit1 <- readRDS("../output/pbmc68k/rds/fit-pbmc68k-em-k=7.rds")$fit
+fit2 <- readRDS("../output/pbmc68k/rds/fit-pbmc68k-scd-ex-k=7.rds")$fit
+lda1 <- readRDS("../output/pbmc68k/rds/lda-pbmc68k-em-k=7.rds")$lda
+lda2 <- readRDS("../output/pbmc68k/rds/lda-pbmc68k-scd-ex-k=7.rds")$lda
 fit1 <- poisson2multinom(fit1)
 fit2 <- poisson2multinom(fit2)
+k <- 6
+plot(lda1@gamma[,k],lda2@gamma[,k],pch = 20)
 diag(cor(fit1$L,lda1@gamma))
 diag(cor(fit2$L,lda2@gamma))
+diag(cor(lda1@gamma,lda2@gamma))
+diag(cor(fit1$F,fit2$F))
+
 n    <- nrow(fit1$L)
 rows <- sample(n,2000)
 fit1 <- select_loadings(fit1,rows)
 fit2 <- select_loadings(fit2,rows)
 
 # Try to cluster the cells.
-L <- lda2@gamma
-n <- nrow(L)
-clusters <- rep("T cells + other",n)
-names(clusters) <- rownames(fit1$L)
-clusters[L[,4] > 0.3]  <- "NK cells"
-clusters[L[,8] > 0.4]  <- "B cells"
-clusters[L[,10] > 0.1] <- "dendritic"
-clusters[L[,11] > 0.25] <- "myeloid"
-clusters <- factor(clusters)
+## L <- lda2@gamma
+## n <- nrow(L)
+## clusters <- rep("T cells + other",n)
+## names(clusters) <- rownames(fit1$L)
+## clusters[L[,4] > 0.3]  <- "NK cells"
+## clusters[L[,8] > 0.4]  <- "B cells"
+## clusters[L[,10] > 0.1] <- "dendritic"
+## clusters[L[,11] > 0.25] <- "myeloid"
+## clusters <- factor(clusters)
 
 set.seed(1)
-k <- 11
-p1 <- structure_plot(lda1@gamma[rows,],topics = 1:k,
-                     grouping = clusters[rows],gap = 10)
-p2 <- structure_plot(lda2@gamma[rows,],topics = 1:k,
-                     grouping = clusters[rows],gap = 10)
+k <- 7
+p1 <- structure_plot(fit1,topics = 1:k,gap = 10)
+p2 <- structure_plot(fit2,topics = 1:k,gap = 10)
 plot_grid(p1,p2,nrow = 2,ncol = 1)
 
-# NK cells.
-k <- 4
-dat <- data.frame(gene = genes$symbol,
-                  f0 = exp(apply(lda2@beta[-k,],2,max)),
-                  f1 = exp(lda1@beta[k,]),
-                  f2 = exp(lda2@beta[k,]))
-dat <- transform(dat,lfc = log2(f2/f0))
-subset(dat,lfc > 2.5 & f2 > 0.001)
-
-# B cells.
-k <- 8
-dat <- data.frame(gene = genes$symbol,
-                  f0 = exp(apply(lda2@beta[-k,],2,max)),
-                  f1 = exp(lda1@beta[k,]),
-                  f2 = exp(lda2@beta[k,]))
-dat <- transform(dat,lfc = log2(f2/f0))
-subset(dat,lfc > 2 & f2 > 0.0005)
-
-# dendritic cells.
-k <- 10
-dat <- data.frame(gene = genes$symbol,
-                  f0 = exp(apply(lda2@beta[-k,],2,max)),
-                  f1 = exp(lda1@beta[k,]),
-                  f2 = exp(lda2@beta[k,]))
-dat <- transform(dat,lfc = log2(f2/f0))
-subset(dat,lfc > 2 & f2 > 0.001)
+stop()
 
-# myeloid cells.
-k <- 11
+k <- 6
 dat <- data.frame(gene = genes$symbol,
-                  f0 = exp(apply(lda2@beta[-k,],2,max)),
-                  f1 = exp(lda1@beta[k,]),
-                  f2 = exp(lda2@beta[k,]))
+                  f0 = apply(fit2$L[,-k],2,max),
+                  f1 = fit1$L[,k],
+                  f2 = fit2$L[,k])
 dat <- transform(dat,lfc = log2(f2/f0))
-subset(dat,lfc > 2.5 & f2 > 0.001)
+subset(dat,lfc > 0.5 & f2 > 0.001)