dozmorovlab / HiCcompare

Joint normalization of two Hi-C matrices, visualization and detection of differential chromatin interactions. See multiHiCcompare for the analysis of multiple Hi-C matrices
https://dozmorovlab.github.io/HiCcompare/
Other
19 stars 3 forks source link

Warnings from simpleLoess, effect on results? #30

Closed TeiturAK closed 1 year ago

TeiturAK commented 1 year ago

Hi,

Thank you for a great tool.

I am getting these warnings when I run hic_loess:

There were 11 warnings (use warnings() to see them)
1: In simpleLoess(y, x, w, span, degree = degree, parametric = parametric,  ... : pseudoinverse used at 2
2: In simpleLoess(y, x, w, span, degree = degree, parametric = parametric,  ... : neighborhood radius 1
3: In simpleLoess(y, x, w, span, degree = degree, parametric = parametric,  ... : reciprocal condition number  0
...

From your manual (https://www.bioconductor.org/packages/release/bioc/manuals/HiCcompare/man/HiCcompare.pdf): Note: if you receive the warning "In simpleLoess(y, x, w, span, degree = degree, parametric = parametric, ... :pseudoinverse used..." it should not effect your results, however it can be avoided by manually setting the span to a larger value using the span option.

Are the errors mentioning neighborhood radius and reciprocal condition number related to the pseudoinverse and can be assumed to have little effect on the result?

TeiturAK commented 1 year ago

Setting span = 0.05 in hic_loess removes this warning, as described in this thread: https://github.com/dozmorovlab/HiCcompare/issues/16#:~:text=%22make%20span%20bigger,at%20vcu%20.%20edu

mdozmorov commented 1 year ago

Great, that's the correct solution. The issue arises if the data is sparse and having a fixed larger span adds stability to the loess procedure.