Multiplexing cost calculator
Sample ‘multiplexing’, i.e. pooling cells from different samples together and running a single experiment, has significant potential benefits for single cell experiments. The 'demuxlet' algorithm (Ye lab, UCSF), leverages genetic polymorphisms to demultiplex pooled cells from different genetic backgrounds, while the 'Cell hashing' approach (Satija and Technology/Innovation labs, NYGC), accomplishes similar goals with barcoded antibodies. Both approaches also enable robust detection of cross-sample doublets, as they will exhibit multiple sample barcodes.
By identifying and discarding doublets, multiplexing enables the ‘super-loading’ of commercial droplet-based single cell platforms, which can greatly reduce costs. We provide a multiplexing cost calculator below, which models the costs of library prep and sequencing for different experimental designs.
|We aim to recover 20,000 single cells. By multiplexing 8 samples together, running one 10x lane yields a non-identifiable multiplet rate of 2.9% and a total cost of ~$4,700 We aim to recover 20,000 single cells, without multiplexing. To achieve a similar non-identifiable multiplet rate, we need to spread the cells across 6 10x runs, with a total cost of ~$14,000.|
|Number of cells desired||Number of 10x lanes||Number of multiplexed samples|
|Desired sequencing depth||Cost of 1M reads||Cost of 10x lane||Multiplet read factor|
- Expected loading and multiplet rates are extrapolated from the Chromium 3' v2 User Guide. Specifically, we approximate the cell recovery rate at ~57%, and use a linear approximation to the multiplet rate of ~4.6e-06 * [# cells loaded]. Exact calculations are present in the source code of this page.
- A 'cell' (i.e. number of cells desired, reads/cell), refers to a non-discarded data point: singlets + non-identifiable multiplets.
- Sequencing costs are approximately based on NextSeq (~$1500 / 400M reads). We add in ~30% extra cost to account for unaligned reads, adapters etc.
- Multiplet read factor is based on empirical observations that doublets contain, on average, ~80% increase in UMI, and therefore require greater sequencing depth.
This website was created by Christoph Hafemeister and Rahul Satija at the New York Genome Center. We would like to thank Peter Smibert and Marlon Stoeckius of the NYGC Technology Innovation Lab, and Jimmie Ye and Meena Subramaniam of the Ye Lab, UCSF for helpful discussions regarding this tool. Technologies used: plotly, jQuery, jQuery UI, Underscore.js
For questions or comments email email@example.com