In general, targeting 20,000 reads per cell is a good place to start for sequencing depth optimization. However, the ideal sequencing depth will be dependent on your sample type and experimental goals. For example, if you are aiming to sequence rare cell types in a heterogenous population, you will likely need to sequence much deeper than this. Alternatively, if you are analyzing a homogenous population, you may not need to sequence as deeply.
We have several datasets on the Parse Biosciences website which can serve as a helpful reference for sequencing depth across different sample types.
The flexibility of sublibrary creation in the Parse Biosciences workflow allows you to allocate different numbers of cells or nuclei into each sublibrary. We recommend creating one small sublibrary (e.g., 500 or 1,000 cells) that can be sequenced more deeply to assess library complexity and sequencing saturation. In turn, this can inform how to optimize sequencing depth for the rest of the sublibraries in the experiment.