We used the latest devel version of minfi and used estimateCellCounts() to estimate cell proportion with EPIC data. Everything runs smoothly. But when it comes to large data (like more than 2000 samples), running the function can easily hit the memory limit. We are using cluster which is offering 256G memory.
We could run estimateCellCounts() plate by plate, but will this introduce potential biases compared with running it with the whole data?
Our long term plans (within the next 6 months) are to transition to a HDF5 backend which (should) enable processing in bounded memory. Otherwise I have no comments.
Hi,
We used the latest devel version of minfi and used estimateCellCounts() to estimate cell proportion with EPIC data. Everything runs smoothly. But when it comes to large data (like more than 2000 samples), running the function can easily hit the memory limit. We are using cluster which is offering 256G memory.
We could run estimateCellCounts() plate by plate, but will this introduce potential biases compared with running it with the whole data?
Thanks!