predictSmooth vs plotSmoothers

[flde]

Hello tradeSeq-Team,

I plotted the gene expression with plotSmoothers and my own function where I use predictSmooth. However, for some genes the results are similar (e.g. Kit) but for others (e.g. Cd47) the smoothed signal looks quite different.

The only thing I noticed is that both function use .getPredictRangeDf() and predictGAM(). But plotSmoothers then removes (?) the offset yhat <- c(exp(t(Xdf %*% t(beta)) + df$offset) while that happens later with predictSmooth?

I would highly appreciate it if you can help me to understand why the smoothed lines differ between the two functions.

Best wishes,
Florian

[koenvandenberge]

Hi @flde,

You may want to share your own function to allow us to look into this or take a look at plotSmoothers yourself.

Also it looks like the points can have here have different on your two plots, did you set the same seed prior to running fitGAM? It may just be a plotting thing.

[flde]

Hello @koenvandenberge,

Many thanks for your help! I did a check with your data from the workshop. The problem was that I did not take the log1p from the output of predictSmooth. I attached the jupyter notebook with the workflow in case someone else has the same question.

smoother.zip

Best,
Florian