Cannot get result from TIMECOURSE (R PROGRAM)

Question

[Hushus]

[This is a follow-up to my previous post]

Hello world,

I have a set of genes whose RNA expression is catalogued in development (FlyBase temporal RNA-seq). I want to find which genes go up and down at time point A_07 and possibly cluster genes to their patterns also.

I put this into R and used the "Timecourse package" https://bioconductor.org/packages/release/bioc/vignettes/timecourse/inst/doc/timecourse.pdf

I've tried to replicate the script as close as I can including labelling the columns accordingly.

Here is the following matrix:

I used the mb.long function and the result:

The results come out to be "NaN" with no error message by R. I've tried expanding the prior.df function from 1-5 but nothing changes this result. Whats going wrong?

[cpad0112]

Please post some example data (not image).

[cpad0112]

I created a dummy data set for this. Please go through the code below and try to compare with your code. Probably the data structure in R might be incorrect. Check your data types. I added comments to the code. Simulated data has 4 times points (1-4) and two conditions (Normal and tumor represented by A and B respectively). Gene names are 4 (gene1-4). Hence data frame is 4x8. Compare your data frame with final data frame in the example (name: total_exp)

$ library(timecourse)

## Create 4 normal sample microarray names
$ normal=paste("A_",seq(1:4),".cel", sep="")

## Create 4 tumor sample microarray names
$ tumor=paste("B_",seq(1:04),".cel", sep="")

## create row names to be used for dataframe
$ genes=paste("gene",seq(1:4),sep="")

## Create expression data for 4 normal samples for 4 time points
$ normal_exp=matrix(rnorm(16,100,25),4,4)

## Create expression data for 4 tumor samples for 4 time points
$ tumor_exp=matrix(rnorm(16,100,50),4,4)

## Assign column names to normal expression data
$ colnames(normal_exp)=normal

## Assign row names to tumor expresion data
$ colnames(tumor_exp)=tumor

## Append two data frames next to each other
$ total_exp=cbind(normal_exp, tumor_exp)

## Make gene names as row names for entire data
$ row.names(total_exp)=genes

## Run mb.long. Times= time points for each category (normal-4, tumor-4), reps= total
## number of conditions multiplied by number of rows (observations)    
$ out1=mb.long(total_exp, times=4, reps=rep(2,4))

## Run mob.long with more control. For times and reps look above. Rep.grp= Each
## condition is repeated by number of replicates (4) time.grp= each sample number 
## (1:4) is repeated condition times (2 here)
$ out2 <- mb.long(total_exp, times=4, reps=rep(2,4), rep.grp=rep(c("Normal", "Tumor"), each=4), time.grp=rep(c(1:4),2))

Answer 1

[Hushus]

I only have 1 biological condition and only 1 replicate. Do you need multiple replicates?

Here is the data: http://m.uploadedit.com/bbtc/1512924884525.txt

I used your code and got the following. "out1" is giving me the exact same thing and when I try to plot it, nothing shows. The plot image is for "out2"

Thank you so much for your time

[cpad0112]

Not necessary. It is just an example. But, in general, more the replicates, better would be your statistical power in deducing biological function. One biological condition and one replicate? But text file supplied has 12 cel files. Does each cel file represent one time point? i.e 12 cel files =12 time points x 1 experiment?

[Hushus]

12 time points x 1 experiment. [I have a list of 100 genes, this is just a sample]

So I dont expect a plot from plotProfile with my results? How can I get a clustered image like this one:

[cpad0112]

I tried one replicate per time point using fruitfly data. mb.long fails to run. I tried the same using simulated data and again it fails. I am not sure if it is possible to run timecourse package function with 1 sample per timepoint and one condition .

[Hushus]

Hmm must be so...

Do you know of another way of producing a similar plot as the one I recently posted? Im gonna start a new thread on this