I have integrated 11 seurat objects that I need to merge before downstream analysis, I combined the first 4 (2.6GB total) and saved them as an rds object just fine but every time I combine the rest (>5GB) the session aborts and PC even crashed. I found that Rstudio can handle just between 2-4 GBs, is there a way to integrate these objects? And would performing the quality control and dimensionality reduction for each object before merging reduce the size? Or is scanpy in python more convenient in that case?
You simply need more computing resources.
And if RStudio is not handling the data, don't use it, just use the CLI to run command or script.
I agree that you need more computing resources. This is definitely not a task to carry out on a pc. Do you have access to a high throughput computing system? With more resources, R would be up to the task.
Dimensionality reduction before merging will increase the size of each object significantly and makes the integration even more computationally intensive. First, you may want to increase the memory size (memory.size(max = xxx)). Second, after data integration you may want to extract the "data" slot from "integrated" assay and read it in a separate Seurat object with relevant metadata. So this way you can decrease the size of your Seurat object and do the Dim reduction then.
Of course, merging 11 scRNA-seq or so objects in PC is hardly possible if not impossible.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts