hic-pro 3 breaks ice normalized matrixes and sparsetodense.py

[ChenfuShi]

Hi, I updated to hic-pro version 3.0.0 but now the ice normalized matrixes are not compatible anymore with the sparsetodense script.

the ICE matrixes seem to have a new format, the old format had 3 columns with binX, binY and counts. The new format has 3 rows and they seem to be all floating point numbers. This affects only the ice normalized matrixes and not the raw matrixes, I imagine this is caused by some bug in the ice normalization script.

The file is too big to upload but if you need it i can upload it somewhere else.

Thanks!

[nservant]

Hi, The iced version you are using is bugged ! please update the package. Sorry for that N

[ChenfuShi]

i was using 0.5.4 which is the one that is set in your environment.yml file! I'll update it and see what happens

[nservant]

Yes, my fault. I'll update it. Sorry N

[ChenfuShi]

Thanks!

[nservant]

Of note, another user reported me that the latest iced version 0.5.8 may also have different outputs compared to the previous ones. The two first columns are expected to be one-based indices, while in the 0.5.8 they are zero-based. If you upgrade to 0.5.8 and observe the same thing, please let me know ... so that I can contact the iced developer. Thanks

[ChenfuShi]

yes it looks like they are 0-based

[esebesty]

I ran into this issue while trying to convert the ice normalized matrix to the format needed by TopDom. Looks like a simple matrix transpose with R on the ice normalized matrix solves the issue.

library("data.table")

icematrix   <- fread(file = "data/sample_1000000_iced.matrix", header = FALSE)
icematrix_t <- t(icematrix)
icematrix_d <- as.data.frame(icematrix_t)

write.table(icematrix_d, file = "data/sample_1000000_iced_corr.matrix", sep = "\t", quote = FALSE,
            col.names = FALSE, row.names = FALSE)

The new matrix looks like this:

1       1       3590.51297981505
2       1       1465.5227894145
3       1       270.030342746284
4       1       158.666018009291
5       1       90.5154390031429
6       1       56.8238616628458

and its accepted by the sparseToDense.py script.

BTW, it might be useful to add an option to the script and provide an N-by-(3+N) or N-by-(4+N) output matrix, where the first 3/4 columns are chr, from, to coordinates, or id, chr, from, to.

[nservant]

Fixed in iced 0.5.9 which has been added in the conda env of HiC-Pro 3.1.0