Hi Felix,

You're rigth, I move to the BioC list.

I tried your exemple, it works well.

In the exptData, do I really need all these fields ? or can I use an empty string for projectPath, fragmentDir, reSequence1 ... ?

Then, just to clarify some points ;

fc@colData is your experimental design, so in your exemple, you have 6 samples from the same viewpoint "testdata".

What about rowData ? if I'm correct, these are the measures (counts) per viewpoint ? so here you only have interacting loci per viewpoint ? Which make sense with the dim of the counts object (6 samples x 2 measures)

I'm just wondering if you can have multiple viewpoints in the same FourC object ? with potentially different number of measures ...

Thank you again,

Best wishes

N.

--------------------------------

Hi Nicolas,

this might be interesting to other users as well. So I would recommend we move the discussion to the Biocondoctor Support Forum if this is fine with you.

Have a look at the script below. This should explain all the steps you need to do to add a custom fragment reference, viewpoint information and counts.

Best regards,
Felix

--------------------------------

library(FourCSeq)

exptData <- SimpleList(projectPath=tempdir(),
                       fragmentDir="re_fragments",
referenceGenomeFile=system.file("extdata/dm3_chr2L_1-6900.fa",
package="FourCSeq"),
                       reSequence1="GATC",
                       reSequence2="CATG",
                       primerFile="",
                       bamFilePath="")

colData <- DataFrame(viewpoint = "testdata",
                     condition = factor(rep(c("WE_68h", "MESO_68h", "WE_34h"),
                                            each=2),
                                        levels = c("WE_68h", "MESO_68h", "WE_34h")),
                     replicate = rep(c(1, 2),
                                     3),
                     bamFile = "",
                     sequencingPrimer="")

fc <- FourC(colData, exptData)
fc

## this step can also be skipped by adding your fragment reference (GRanges)
##rowData(fc) <- your Fragment GRanges
fc <- addFragments(fc)
fc

## counts would by your HiC counts, here I take a random sample
cnts <- matrix(sample(1:10, prod(dim(fc)), replace = TRUE),
               nrow=nrow(fc))
cnts

## make sure they are integers otherwise you will get an error
typeof(cnts)

counts(fc) <- cnts
fc

## manually add viewpoint information because there is no primerfile
colData(fc)$chr = "chr2L"
colData(fc)$start = 6027
colData(fc)$end = 6878

## now you should be able to continue with the workflow

On 10/09/2015 06:34 PM, Nicolas Servant wrote:
Dear Felix,

I'm developing the HiTC package on Bioconductor to analyse Hi-C data.
I would like to make a test to see whether the method implemented in FourCSeq can be used on some specific Hi-C viewpoints (so 4C like information).
So actually, my question is quite simple, let's say that I have a GRanges objects with a count information extracted from HiTC, and that I would like to enter into FourCSeq to detect valid interactions.
I saw that the FourC object require many information about the pre-processing such as the BAM file, etc.
Could give any advices to create such object from a count table ?

Thank you very much
Best wishes
Nicolas

Hi Nicolas,

for exptData you can use empty strings if you do not want to provide the information. However the fields are required by the class validity check so that everything works in the normal process. Same is true for colData.

colData(fc) (better accessed this way) returns the experimental design and you have to provide the fields given in the example.

rowData is a GRanges object that contains the genomic positions of the restriction fragments. During the analysis more data is added to the mcols by FourCSeq and calls to DESeq2. Here you should be able to add "resFrag", "cutSites", y_intervals(), and all other GRanges objects generated by HiTC.

The counts are stored in "assays" (= SimpleList of matrices). "counts(fc)" is a special accessor that returns the counts "assay" of a FourC object. All viewpoints must have the same number of fragments (= nrow). So it wont work if you have different fragment references, e.g. cut by different combinations of restriction enzymes.

The subsequent analysis also assumes that all data is from the same viewpoint (replicates and different conditions). So putting two different viewpoints into the FourC object will cause problems in the downstream analysis.

Best regards,

Felix

Hi Felix,

Still trying to adapt FourCSeq to other 'C' data. I build my FourC object from scratch, with many empty fields following our previous discussion. Then when I used the getZScores, it crashed because leftSize, rightSize, leftValid, rightValid are not defined.

The line

fragData$blind = fragData$leftSize < 0 & fragData$rightSize <  0

from getDistAroundVp() crashed.

I'm not sure to really understant what these fields mean ? Do I have to fix them to a constant ?

Thank you

Nicolas

Hi Nicolas,

these settings are used for 4C libraries to filter out fragments that have do not contain a site of the second cutter at all. In this case the size is set to -1.

If you want to use all fragments then you can just set leftSize and rightSize to 1 and leftValid and rightValid to TRUE in the fragment data. This should to the trick.

Best regards,

Felix

ok. Thank you Felix. I will try it.

Still an issue with getDistAroundVp from getZScores

Error in mcols(mcols(fragData))[colnames(mcols(fragData)) %in% newCols,  :
  type / longueur incorrect (S4 / 0)

It seems that mcols(mcols(fragData)) is NULL in my case ?

Did I miss something ?

Thank you again

Nicolas

Thank you Felix. It works. If useful, I can send you the function I wrote to build a FourC object from a BED file.

Last question, when I apply the code from the vignette, I have no significant interaction, although I can see on the viewpoint plot some points higher than the dotted blue line (in both replicates). Would you have any idea ?

Thank you. Nicolas