Change find cutoff in PCA for Stress and NE to first local minima after global maxima

r.scripts/sc_PC.Score.NE.R

@@ -96,7 +96,7 @@ ts_1 <- ts(d1$y)
 #ts_2 <- ts(d2$y)
 tp1 <- turnpoints(ts_1)
 #tp2 <- turnpoints(ts_2)
-pit1 <- min(d1$x[tp1$pits][d1$y[tp1$pits]==min(d1$y[tp1$pits])])
+pit1 <- min(d1$x[tp1$pits][d1$x[tp1$pits]>d1$x[tp1$peaks][d1$y[tp1$peaks]==max(d1$y[tp1$peaks])]])
 #pit2 <- d2$x[tp2$pits][length(d2$x[tp2$pits])]
 #abline(v=pit1)
 

r.scripts/sc_PC.Score.Stress.R

@@ -75,13 +75,14 @@ plot(plot)
 dev.off()
 
 #find first turning point on a KDE of Stress1
+histo <- hist(GetCellEmbeddings(object=sc10x.Group,reduction.type="Stress",dims.use=1),breaks=10,prob=TRUE,plot=TRUE)
 d1 <- density(GetCellEmbeddings(object=sc10x.Group,reduction.type="Stress",dims.use=1),n=1000)
 #d2 <- density(GetCellEmbeddings(object=sc10x.Group,reduction.type="Stress",dims.use=2),n=1000)
 ts_1 <- ts(d1$y)
 #ts_2 <- ts(d2$y)
 tp1 <- turnpoints(ts_1)
 #tp2 <- turnpoints(ts_2)
-pit1 <- min(d1$x[tp1$pits])
+pit1 <- min(d1$x[tp1$pits][d1$x[tp1$pits]>d1$x[tp1$peaks][d1$y[tp1$peaks]==max(d1$y[tp1$peaks])]])
 #pit2 <- min(d2$x[tp2$pits])
 
 #plot clusters
@@ -98,6 +99,8 @@ plot <- plot+geom_vline(xintercept=pit1,color="red",lwd=2.5)
 plot(plot)
 dev.off()
 
+
+
 #subsample all cells (+Stress) to better visualize their clustering
 if (!is.na(opt$s)){
   rnd <- sample(1:ncol(sc10x.Group@data),opt$sv)