online_iNMF code check

[niehu2018]

Source: https://github.com/welch-lab/liger/blob/57b9ffa5a720f7c028850dea493c1da6657a3a8c/R/rliger.R line 1629: processed = !is.null(X_new[[i]]@scale.data) This line should be : processed = !is.null(object@scale.data[[i]]) ?? When I give a liger object with multiple dataset, online_iNMF failed, the reason was "out of bounds".

[cgao90]

Hi Niehu,

To help further investigate, could you share the code snippet associated with this error message?

Best, Chao

[niehu2018]

Here are the code and log

code: library(rliger) stim = readRDS("pbmcs_stim.RDS") ctrl = readRDS("pbmcs_ctrl.RDS") ctrl = createLiger(list(a = ctrl, b = stim), remove.missing = F) # this may not make any sense, I just want to simulate my situation stim = createLiger(list(stim = stim), remove.missing = F)

stim = normalize(stim) stim = selectGenes(stim, var.thresh = 0.1, do.plot = F) stim = scaleNotCenter(stim) stim = online_iNMF(stim, k = 20, miniBatch_size = 5000, max.epochs = 5) stim = quantile_norm(stim) stim = runUMAP(stim)

ctrl = normalize(ctrl) ctrl@var.genes = stim@var.genes ctrl = scaleNotCenter(ctrl) comb = online_iNMF(stim, X_new = list(ctrl = ctrl, stim = stim), k = 40, max.epochs = 1)

log: New dataset 1 already preprocessed. New dataset 2 already preprocessed. Error in X_new[[i]] : subscript out of bounds

The pbmcs_stim.RDS and pbmcs_ctrl.RDS were downloaded from https://github.com/welch-lab/liger

[cgao90]

Hi, the error here is caused by the fact that the stim object is being used as the 1st and 3rd input at the same time. This would confuse the algorithm make it think there are x2 number of cells in stim . The following should work:

ctrl = createLiger(list(b = ctrl_data), remove.missing = F) 
stim2 = createLiger(list(stim2 = stim_data), remove.missing = F) 
#normalization, assign variable genes, scaling, ...
comb = online_iNMF(stim, X_new = list(ctrl = ctrl, stim2 = stim2), k = 20, max.epochs = 1)