Creating a 5 or 10-fold cross validation based on sequence similarity of proteins

Question

[rafi.zon]

Hi there,

As the post title states I am trying to find an approach to construct a 5 or 10 fold cross validation dataset, applied to all of the currently available human proteins in Swiss-Prot (20.421 proteins).
Ideally, in each of the folds there should be the most similar proteins in terms of their sequence identity.
What can be a way to divide the proteins into the respective cross validation sets based on similarity?

Answer 1

[trent]

I wrote how I did this with CD Hit here, hopefully it's helpful: https://www.trenthauck.com/cd-hit-cross-validation