Closed mengwill closed 1 year ago
As with any correlation matrix the sign of the eigenvector is arbitrary. So you can always multiply it by -1 and flip the sign. To know what the correct sign is for the original eigenvector code, you need additional information beyond Hi-C data (active genes, methylation, etc.) and you flip the sign to correlate appropriately.
Our newer code, POSSUMM, will correlate the vectors with housekeeping genes.
Are you sure this is an issue? Github Issues is reserved for situations that require changes to the codebase. Questions, discussions, etc. should be posted to the forum: aidenlab.org/forum.html When we use the juicer eigenvector function, we found the signs of the vectors are opposite in 2 out of 8 samples. These are hiC datasets. The tracks illustrate the eigenvectors of chromosome 22.