What's really up with Roche and PacBio?

You're probably aware that Roche recently severed its 2013 partnership with PacBio and decided to go it alone without access to long read SMRT technology. Neil Gunn, Roche's head of sequencing R&D explains that Roche will now 'focus more intently on its internal R&D efforts to "drive our long term strategy, which is to be a leader in clinical diagnostic sequencing."'. This wouldn't have anything to do with the 2014 acquisition of Genia, would it?

Keith Robison at Omics! Omics! has already speculated about various reasons for Roche's break up, but I think that the developments around Genia are the primary reason for this parting of way. As Robison explains:

The obvious culprit would be Genia, the nanopore sequencing technology which Roche purchased. ... Genia has twice reported this year technical success, first with their nucleotide chemistry and then their protein engineering.

But good publications do not a commercial instrument make, and there has been no signal that a launch is imminent.  Perhaps we will hear a lot more at JP Morgan or AGBT, but Genia has burned a lot of credibility in the past with premature announcements.

This may be true, but consumers in the genomics space have been pretty forgiving:

Many of complained that ONT announced vaporware at 2012, but they've delivered devices to customers since mid-2014.  

And as of now there are a lot of working ONT devices in the field.

But back to Roche/Genia. It's tempting to speculate that Roche is building a competing platform to Oxford Nanopore's technology, which on the surface may look like what it is. However, the technologies and the way sequence is read from them is different.  On one hand, ONT's nanopores detect electrical signals generated by DNA kmers traveling through the pore, and in comparison Genia's technology reads DNA by having a polymerase send engineered tags through their pore.

How will Oxford Nanopore's MinION compare with Roche's Genia?

At this point, they look like competing technologies in the nanopore sequencing space, but without more information on the Genia it's hard to make a call. However, since the Genia relies on proprietary modified bases called NanoTags read by a nanopore, I'm going to speculate and assume that the Genia will have a higher consumable cost as result; potentially a big drawback for the Roche platform. Nevertheless, here's how they compare across a two key value drivers.

Read accuracy is a big deal. At the moment, it looks like Genia might have a hand up due to their NanoTag read system. I suspect that it easier to calling bases directly using engineered NanoTags that it is to infer bases from electrical signals generated by DNA going through a pore, as is the case with ONT. If Roche can make up the additional cost of NanoTag reagents by offering a higher read accuracy, the Genia might be worth it, particularly if the flow cell cost is kept below that of ONT's.

Cost. There are two major categories of cost sequencing teams are concerned with; operating costs and capital costs. The nice innovation that ONT has brought to the genomics field is the idea of zero (or nearly zero) upfront capital costs. You buy a flow cell for <$1000 and start sequencing.  The capital cost of a Genia is an unknown, but if it's anywhere near that of PacBio or Illumina equipment it'll be an obstacle to adoption in the field. Unfortunately, without a target purchase price (and other assumptions about the Genia's longevity) it's impossible to come up with a cost model and estimated $/gigabase.

Some guidance from Roche is needed here both in terms of cost and accuracy. I haven't even touched read length as a dimension of performance, but given ONT's extreme read lengths (up to 200kb, more commonly ~50kb), Genia needs to easily hit >30-50kb even to be considered as competition.