Identifying knots on multi-curve plot

[fluentin44]

Hi,

Im trying to look at differential expression between two knots, ideally at the point where the five lineages diverge using earlyDERes <- earlyDETest(gam, knots = c(2, 3)) but cant reliably distinguish the knot number for each curve - is there anyway of overlaying each curve on the graph one by one?

Alternatively, when using yhatSmooth <- predictSmooth(gam, gene = var_genes[1:50], nPoints = 50, tidy = FALSE) would there be any way to indicate / identify where each point may lay on each curve?

Thanks,
Matt

[koenvandenberge]

Hi Matt,

This is a good question and we should probably allow for a better visualization that does this for you.
For now, what I can suggest is just working with the ggplot2 object and tweaking how you plot the knot points.
For example, below I use the example from plotGeneCount to just make the points larger so you can easily count them. Alternative ways are of course possible.

Hope this helps.

set.seed(97)
library(slingshot)
library(tradeSeq)
data(crv, package="tradeSeq")
data(countMatrix, package="tradeSeq")
rd <- slingReducedDim(crv)
cl <- kmeans(rd, centers = 7)$cluster
lin <- getLineages(rd, clusterLabels = cl, start.clus = 4)
crv <- getCurves(lin)
counts <- as.matrix(countMatrix)
gamList <- fitGAM(counts = counts,
                  pseudotime = slingPseudotime(crv, na = FALSE),
                  cellWeights = slingCurveWeights(crv))
p=plotGeneCount(crv, counts, clusters=cl, models=gamList)

dat3 <- p$layers[[3]]$data #knots data
p  + geom_point(data=dat3, size=10, col="black")

[fluentin44]

Hi,

Brilliant, ill have a play.

Thanks very much!

[greengarden0925]

Hi Matt,

This is a good question and we should probably allow for a better visualization that does this for you. For now, what I can suggest is just working with the ggplot2 object and tweaking how you plot the knot points. For example, below I use the example from plotGeneCount to just make the points larger so you can easily count them. Alternative ways are of course possible.

Hope this helps.

set.seed(97)
library(slingshot)
data(crv, package="tradeSeq")
data(countMatrix, package="tradeSeq")
rd <- slingReducedDim(crv)
cl <- kmeans(rd, centers = 7)$cluster
lin <- getLineages(rd, clusterLabels = cl, start.clus = 4)
crv <- getCurves(lin)
counts <- as.matrix(countMatrix)
gamList <- fitGAM(counts = counts,
 pseudotime = slingPseudotime(crv, na = FALSE),
 cellWeights = slingCurveWeights(crv))
p=plotGeneCount(crv, counts, clusters=cl, models=gamList)

dat3 <- hlp$layers[[3]]$data #knots data
p  + geom_point(data=dat3, size=10, col="black")

Where the object "hlp" from? Thank you for replying. I have the same question of identifing knots.

[koenvandenberge]

Hi @greengarden0925

That should have been object p, my mistake. I have now updated the above comment.

[greengarden0925]

Hi @greengarden0925

That should have been object p, my mistake. I have now updated the above comment.

Thank you for the clarification.