Open veramazzara opened 1 year ago
Hi @veramazzara ,
I have pushed a commit to allow the i3 model to work by taking its parameters from all observations when no references are provided. Just keep in mind that when no references are provided, the average of all observations is used as the baseline to which everything is compared, so results may be flipped to reality depending on the proportion of healthy/tumor (with a given cnv) cells. To get the update, you will need to update using the code on github with:
library("devtools")
devtools::install_github("broadinstitute/infercnv")
Regards, Christophe.
Hi Christophe, my apologies for my slow response. Thank you very much for this further commit; now I can run HMM prediction! And thank you for alerting me about the results's interpretation without the reference.
Regards,
Vera
Hi GeorgescuC, thank you for this useful tool! I am working with R version 4.2.3 and infercnv 1.14.2 but I can't be able to obtain prediction with HMM model.
I'm working without reference cells
infercnv_obj_default = infercnv::run( infercnv_object_example, cutoff=0.1, out_dir=getwd(), cluster_by_groups=TRUE, plot_steps=FALSE, denoise=TRUE, HMM=TRUE, HMM_type=c("i3"), analysis_mode="subclusters",
no_prelim_plot=TRUE,
png_res=300, num_threads = 4 )
I obtained this error:
INFO [2023-05-25 21:37:30] i3HMM_predict_CNV_via_HMM_on_tumor_subclusters(i3_p_val=0.05, use_KS=FALSE) INFO [2023-05-25 21:37:31] determine mean delta (sigma: NA, p=0.05) -> NA mean_delta: NA, at sigma: NA, and pval: 0.05 WARN [2023-05-25 21:37:31] get_HoneyBADGER_setGexpDev:: k_cells must be at least 2, setting to 2 Error in seq.default(0, gexp.sd, gexp.sd/10) : 'to' must be a finite number
Is it possible to apply HMM model without reference cells? If yes, how can fix this issue?
Thank you so much.
Vera