Enormous amounts of warnings from brow EChO.1_0.R and its matrix version as well.

[skilpinen]

Hi, we chatted briefly in Heidelberg last week. Based on that I started to do deeper analysis to our Cut & Tag peaks. I used this EChO analysis code from this repo. After creation of all required bed files I started to run these through the code but both foci and matrix version of the code generate thousands on warnings from R side.

For example foci mode gives warnings like this:
Warning messages:
1: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
pseudoinverse used at -60.63
2: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
neighborhood radius 327.63
3: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
reciprocal condition number 2.8048e-17
4: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
There are other near singularities as well. 1.0734e+05
Error in simpleLoess(y, x, w, span, degree = degree, parametric = parametric, :
NA/NaN/Inf in foreign function call (arg 7)
Calls: loess -> simpleLoess
In addition: There were 50 or more warnings (use warnings() to see the first 50)
Execution halted
There were 50 or more warnings (use warnings() to see the first 50)
Warning messages:
1: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
pseudoinverse used at -135.13
2: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
neighborhood radius 5.13
3: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
reciprocal condition number 0
4: In predLoess(object$y, object$x, newx = if (is.null(newdata)) object$x else if (is.data.frame(newdata)) as.matrix(model.frame(delete.response(terms(object)), :
There are other near singularities as well. 969.08

etc... these go for thousands of lines. Though code still produces output file but I am not certain can it be trusted given this huge amount of warnings. Matrix mode has its own warnings and fatal errors.

I am proficient in Bash/Perl/R coding so I traced this very deep into the code. In the foci mode this part of R code:
lo<-loess(frame$V2 ~ frame$V1, span=i)
generates like bulk of the warnings. Though these seems to be warnings from loess fit.

Minor ones came from is.na(as.numeric(argsL$span))==FALSE which I changed to is.numeric(argsL$span) and that I hope is fine now.

Then in matrix mode:
It gets into serious issues with this part of the code:
values<-seq(1,0.05,-0.01)
span<-c()
wss<-data.frame(matrix(ncol=3,nrow=0))
for(i in values){
if(i*nrow(frame) > 1){
lo<-loess(frame$V2 ~ frame$V1, span=i)
if(is.na(lo$s)==FALSE){
out2<-predict(lo,index)
if(sum(out2<0)==0){
rough<-sd(diff(out2))
new3<-data.frame(i,lo$s,rough)
wss<-rbind(wss,new3)
}
}
}
}
wss<-wss[!is.infinite(rowSums(wss)),]
wss$loscale<-scale(wss$lo.s)

As if i*nrow(frame) > 1 is not TRUE for any i of values it will end up giving zero rows matrix to scale function. And that is error.

Another source of errors was this line:
spanout<-data.frame(argsL$line,span,lo$s,out3$fit[index==0],out3$se[index==0])
As index in my data does not contain 0 in any case so that data.frame cannot be formed with varying length elements.

I am wondering am I doing something wrong here, or is this runnable with some specific older R version. I already tested between 4.3.0 to 3.6.3 and nothing changes, which is not too surprising. Can you please advice a bit, I can fix code yes, but I would rather not rethink this whole calculation as you have made great tool with much of thought. And I just suspect I do something wrong.

I wonder if I give you these few ready bed files, do you have environment where you could test foci and matrix runs with those?