Error in if (tvar < 0) { : argument is of length zero

[lydiayliu]

Error in if (tvar < 0) { : argument is of length zero
Calls: run_clonal_ASCAT -> recalc_psi_t -> is.segment.clonal -> <Anonymous>
Execution halted

As the error suggests, it happens in fit.copy.number -> run_clonal_ASCAT -> recalc_psi_t -> is.segment.clonal

After a long debugging session, I believe the cause of the issue is that somehow in is.segment.clonal, the segment is calculated to have negative nMajor even after nMinor gets adjusted to 0.01.

battenberg/R/clonal_ascat.R

Lines 384 to 403 in 4368667

	# check for big shifts in nMajor - if there's a big shift, we shouldn't trust a clonal call
	nMajor.saved = nMajor
	## to make sure we're always in a positive square:
	#if(nMajor < 0) {
	# nMajor = 0.01
	#}
	#
	#if(nMinor < 0) {
	# nMinor = 0.01
	#}
	#DCW - increase nMajor and nMinor together, to avoid impossible combinations (with negative subclonal fractions)
	if(nMinor<0){
	if(BAFreq==1){
	#avoid calling infinite copy number
	nMajor = 1000
	}else{
	nMajor = nMajor + BAFreq * (0.01 - nMinor) / (1-BAFreq)
	}
	nMinor = 0.01
	}

This results in the following two values
test.BAF_levels = c(NA, NA)
whichclosestlevel.test = integer(0)
going in to the function where the error accors
calc_Pvalue_t_twotailed( BAF.size, BAFreq, BAF.sd, test.BAF_levels[whichclosestlevel.test], maxdist_BAF)

battenberg/R/clonal_ascat.R

Lines 77 to 100 in 4368667

	calc_Pvalue_t_twotailed <-function( sample_size, sample_mean, sample_SD, mu_pop, max_dist) # kjd 18-12-2013
	{
	tvar = ( sample_mean - mu_pop ) * sqrt( sample_size ) / sample_SD
	if( tvar < 0 )
	{
	lower_tail_prob = pt( tvar , df = sample_size - 1 , lower.tail = TRUE )
	}else
	{
	lower_tail_prob = 1 - pt( tvar , df = sample_size - 1 , lower.tail = TRUE )
	}
	pval = 2 * lower_tail_prob
	#DCW 250314
	if(abs(sample_mean - mu_pop)<max_dist){
	pval = 1
	}
	return( pval )
	}

The error gets triggered, because tvar cannot be calculated as mu_pop is logical(0)

I'm trying to resolve this issue by adding a new condition that sets pval = 0 if length(test.BAF_levels[whichclosestlevel.test]) == 0

battenberg/R/clonal_ascat.R

Lines 433 to 440 in 4368667

	#270713 - problem caused by segments with constant BAF (usually 1 or 2)
	if(BAF.sd==0){
	pval=0
	}else{
	#pval[i] = t.test(BAFreq,alternative="two.sided",mu=BAF_levels[whichclosestlevel])$p.value
	#pval = t.test(BAFreq,alternative="two.sided",mu=test.BAF_levels[whichclosestlevel.test])$p.value
	pval = calc_Pvalue_t_twotailed( BAF.size, BAFreq, BAF.sd, test.BAF_levels[whichclosestlevel.test], maxdist_BAF)
	}