Bhattacharyya coefficient calculations returning NAs

[tasarte]

I am calculating Bhattacharyya coefficients for different sets of structures using ensemble NMA, using the following code:

modes <- aanma(pdbs, rtb=TRUE, reduced=TRUE)
covs <- cov.enma(modes)
bc <- bhattacharyya(modes, covs=covs)

In at least one example that I have seen, there are NAs in the Bhattacharyya coefficient matrix. Clearly, these happen with specific structures, but there is no obvious reason to me that these should be problematic (e.g., they have good overlap with other structures, structures with breaks in connectivity have already been filtered out, etc.). They also don’t lead to NAs consistently. What could cause this problem?

I have attached the aforementioned matrix, which also contains the names of the structures.

‌

[Xinqiu Yao]

Hi,

‌

Please provide a short example for us to reproduce the problem.