The PacBio Vega - More Tech Details

In my previous post I covered PacBio’s new bench top instrument - The Vega. Based on the specs it seems like an interesting option for small labs.

Optical Isolation and Continuous Operation

In this post I wanted to think through a few of the technical details. Firstly the Vega approach is kind of neat. While not explicitly described in the public information I’ve seen the Vega seems to be divided into essentially to isolated platforms:

Users can therefore load reagents and flow cells into the “user” part of the instrument. These will then be sucked away into an optically isolated part of the instrument where the run proceeds. During the run users can load up and schedule a second run. This suggest the instrument can run with very little down time¹.

The reagents/chip itself are tidied away into a compact consumable:

As far as I can tell, only the chip itself makes its way into the optically isolated platform, when the run completes the Vega spits it out into a little waste bin.

The Chip

There are no pictures of the chip itself yet (and it appears to be shipped hidden away in that consumable). But it appears to use the same chip as the Revio.

Grabbing the Vega data release and generating a heat map, shows a seemingly identical layout to the Revio… but a bit… odd (see this follow up post for more details).

PacBio said the Vega uses a chemistry that’s somewhat similar to the pre-SPRQ Revio chemistry currently. This should bring some cost savings, and perhaps suggests that there may be room to push Vega throughput in the future, perhaps close to the 120Gb offered on the Revio.

Compute

An open question on the Vega is how Pacbio managed to cram all that compute into a tiny little box. A few folks have speculated that a lot of the Revio cost was associated with the compute.

In my previous post I speculated that the baseline compute really shouldn’t be a barrier to building a lower cost PacBio style sequencer. Getting Deep Consensus to fit into such a small form factor was likely more challanging.

I previously speculated that PacBio had squeezed n public Deep Consensus implementation down so it ran on one or two GPUs. For the Vega they must have gone further than this. It seems likely that they can get away with a single Nvidia GPU, or embedded solution.

Overall it always seemed like Deep Consensus was letting you squeeze out a little more from a platform that was already producing high quality data².

Summary

Overall there it doesn’t feel like there’s anything revolutionary in the Vega. It appears to use the same technological approach as the Revio. However PacBio appear to have spent some time considering the requirements of smaller labs.

This seems like good news, the Vega should be a reasonably low risk acquisition given that the approach and data has been validated on the Revio already.

But that near $1M technology is now available to small research labs.

PacBio reviewed a version of this article prior to publication.

1. Unlike other instruments there shouldn’t be any downtime between runs for washes etc.

2. Unlike other platforms where larger models are a requirement to generate acceptable quality basecalls.