mm9 vs. mm10 about cistopic

[cbravo93]

Hi @JBreunig !

Regards rGREAT, I am only able to reproduce this error when my regions are in mm10 but I select mm9 as genome (or vice versa). Which one is the genome assembly of your data, mm9 or mm10? If it is mm10, this code should work:

cisTopicObject <- GREAT(cisTopicObject, genome='mm10', fold_enrichment=2, geneHits=1, sign=0.05, request_interval=10)
pathToFeather <- '/media/data/lcb/icistarget/data/rankings/feather/mm9-DHS_enh_prom_pennstate-9species.all_regions.mc9nr.feather' 
library(R.utils)
url <- "http://hgdownload.soe.ucsc.edu/goldenPath/mm10/liftOver/mm10ToMm9.over.chain.gz"
mm10Tomm9.chain <- "mm10Tomm9.over.chain"
download.file(url, destfile = paste0(mm10Tomm9.chain, ".gz"))
gunzip(paste0(mm10Tomm9.chain, ".gz"))
mm10Tomm9.chain  <- import.chain(mm10Tomm9.chain)

# Obtain liftOver dictionary (as list)
mm10_coord <- [e-mail]
mm10_to_mm9_list <- liftOver(mm10_coord, mm10Tomm9.chain)

cisTopicObject <- binarizedcisTopicsToCtx(cisTopicObject, liftOver=mm10_to_mm9_list, genome='mm9')
cisTopicObject <- scoredRegionsToCtx(cisTopicObject, liftOver=mm10_to_mm9_list, genome='mm9')
cisTopicObject <- topicsRcisTarget(cisTopicObject, genome='mm9', pathToFeather, reduced_database=FALSE, nesThreshold=3, rocthr=0.005, maxRank=20000, nCores=4)
cisTopicObject<- getCistromes(cisTopicObject, annotation = 'Both', nCores=5)

Let me know if this is not the case!

C

[JBreunig]

This is almost certainly my issue as we used mm10. Thanks for the code and help solving it.

[JBreunig]

Hi Carmen,
Your liftover code worked fine. However, I'm running out of memory at the topicsRcisTarget line:

cisTopicObject <- topicsRcisTarget(cisTopicObject, genome='mm9', pathToFeather, reduced_database=FALSE, nesThreshold=3, rocthr=0.005, maxRank=20000, nCores=4)

(Running Ubuntu 18.04 on a dual Xeon workstation with 128Gb of ram and a 32 gb swapdisk.) Memory use starts at 8GB of RAM after running the command but slowly grows until it uses both RAM and swapdisk and then creates an error. This is a dataset of 2,695 mouse tumor cells.

Initial settings:

cisTopicObject <- runModels(cisTopicObject, topic=c(20), seed=987, nCores=14, burnin = 200, iterations = 250, returnType = "selectedModel", addModels=FALSE)

cisTopicObject <- selectModel(cisTopicObject, select=20)

I should add that I've had issues with corrupted feather databases with SCENIC but it doesn't look like I can checksum the mouse feather file mm9-regions-9species.all_regions.mc9nr.feather with the sha256sum.txt file. (It takes 5 days to download each feather file.)

Any suggestions? Thanks in advance!
Josh

[cbravo93]

Hi @JBreunig !

Can you try reducing the number of cores?

Cheers!

C

[JBreunig]

Reducing the cores worked...thanks!
Best,
J