Low cost nanopore arrays

In my last post I discussed the fabrication processes used in nanopore arrays. This prompted a few interesting discussions with folks involved in MEMS fabrication. Using existing capacity the best estimate I received for fabrication costs using the process described is ~$20 per device (Flongle-like 10mm squared die). If this is true, it may suggest that Oxford Nanopore’s yields are relatively low leading to a $45 COGS (at their overall margin of ~50%).

Existing MEMS fabrication capacity for the processes used (from pureplay MEMS fabs) is probably somewhat limited. This limited supply likely contributes significantly to this $20 cost per device.

One route to a low cost per device could therefore be to build out your own fab capacity. For the kinds of diagnostic applications I’m interested in, we’d be looking at >1M units a month. Building out such a fab would likely be very expensive. I suspect significantly more than $200M, perhaps much more.

But with such a facility, it seems at least plausible that you could reach a single digit dollar COGS. And just maybe, you could get close to the qPCR COGS of $1 or $2.

Other approaches

The silicon substrate array approach seems to be most appropriate for scaling out arrays to 1000s or 10000 pores. But if we’re content with the 126 pores on a Flongle might other approaches be a better fit?

One approach might be to fabricate silver electrodes on printed circuit boards directly. Element have demonstrated this in their single channel solid state nanopore systems:

Silver immersion PCBs are readily available and not particularly expensive. Silverr (AgCl) electrodes are pretty much the standard way of creating electrodes in ion channel experiments. Silver however oxidizes relatively easily, making it less desirable in a semiconductor fabrication environment.

However, on a PCB this shouldn't be as much of an issue. These boards can support features down to about 125 microns. Allowing a square millimeter for each well, we can still fit 100 wells into a 10 square millimeter PCB. Last time I ordered immersion silver, boards this size board cost me ~50 cents, with next day turn around. Much lower prices are likely available.

Like the Flongle this would need to interface with an external ASIC. If you’re lucky, off the shelf current sense amplifiers might work for this. Parts are available, which support up to 256 channels per chip. But failing that, a custom ASIC would also need to be developed.

This approach seems plausible to me, but if you have any thoughts on why this would/would not work or better cost estimates than I’ve shown above, please get it touch!