centipede fitting error

[ahorvath]

Dear Kamil,

I would like to ask your help. I'm trying to use the CENTIPEDE package for creating ATAC footprint plots with bed a file. I transformed the file to the format that the fimo file has. I got the following error:

fit <- fitCentipede( Xlist = list(DNase = cen$mat), Y = as.matrix(data.frame( Intercept = rep(1, nrow(cen$mat)) )) ) Initialization of the parameters: Error in quantile.default(x, c(TrimP, 1 - TrimP)) : missing values and NaN's not allowed if 'na.rm' is FALSE

Can you help me?

Thanks in advance. Best Regards, Attila

[slowkow]

Please share the input files and the commands you used to construct the variables.

[ahorvath]

library(devtools)
library(Rsamtools)
library(CENTIPEDE)
library(CENTIPEDE.tutorial)

sample1.bam <- "Sample1.bam"
bed <- "Cluster2_peaks.txt"
cen <- centipede_data(
        bam_file = sample1.bam,
        fimo_file = bed,
        log10p = 4,
        flank_size = 100
)
plot(cen$mat[1,], xlab = "column", ylab = "reads")

fit <- fitCentipede(
        Xlist = list(DNase = cen$mat),
        Y = as.matrix(data.frame(
                Intercept = rep(1, nrow(cen$mat))
        ))
)

I just sent you the data in email.

[slowkow]

It seems you have very few reads in your regions, so I'm not sure if CENTIPEDE will work for you.

Unfortunately, I do not provide support for the CENTIPEDE package. I suggest that you direct further questions to the authors of CENTIPEDE.

library(CENTIPEDE.tutorial)

setwd("~/work/attila/")

sample1.bam <- "Sample1.bam"
bed <- "Cluster2_peaks.txt"
cen <- centipede_data(
  bam_file = sample1.bam,
  fimo_file = bed,
  log10p = 4,
  flank_size = 100
)

plot(rowSums(cen$mat))

library(NMF)
library(viridis)
aheatmap(
  cen$mat, labRow = NA, labCol = NA, Colv = NA, Rowv = NA, color = viridis(100))

[AnnaKuchmiy]

Dear Slowkow, I have exactly the same problem following the tutorial step by step with the data provided here https://slowkow.github.io/CENTIPEDE.tutorial/ Do you aware whether there is any solution for it? Many thanks in advance, Anna

[slowkow]

You might considering taking a step back and looking at your data. Maybe it is not suitable for modeling with CENTIPEDE?

Try making a plot of the read density across the genome. Next, try annotating the positions of transcription factor motifs on the same plot. This should help to develop a sense of how the data might be analyzed further.

[zhenzuo2]

I debugged the function to see waht caused this error. It turns out to be some zeros in LambdaParList. When take log of LambdaParList, it comes to -Inf. And it will cause NA when the function multiplies LambdaParList to some other matrix. So specify LambdaParList before running the fitting should solve this problem.

[rdbcasillas]

Hey @zhenzuo2 , Thanks for the response. How did you decide the value of LambdaParList. From the documentation, it is not clear other than that it should match the length of Xlist.

[bakerwm]

Hi, @slowkow , I create the cen object using the function centipede_data, and found 0 in some positions of my ATAC-Seq data. And this will cause the above error.

The Error was fixed when I Assign the minimum Lambda value to the 0 positions. see here, fixInf branch.

It means no cut events detected at that position? could it happened in actual experiment?