[Chipper]

Sorry to be pedantic, but the chip in the image clearly isn't square (measures up using the high tech method of "ruler against screen" with edge ratio of 2x1.6) that aside, back to the sensor size (I won't use "pixels" to describe them again, perhaps they should be called sexels (sensor elements)).

Have just been back and reworked (post coffee) some of the sensor dimensions. Using images of sensor cross-section posted previously here, the sensors on the 318 chip are ~4.1um edge-to-edge, with ~3.4 um diameter wells, and ~0.7um well to well spacing.

For Proton I, the sensor size (at ~170M per chip) must be ~2um2. Assuming a square sensor (not necessarily a good approximation, but let's go with it for now), the sensor width would be ~1.4um. If the ratio of well width to sensor spacing is preserved relative to the 318 chip, then the well would be ~1.16um wide, and the intersensor spacing would be ~0.25 um.

For Proton II, the sensor size (at ~660M per chip) must be ~0.5um2. Assuming a square sensor (still probably not the best approximation...), the sensor width would be ~0.7um. If the ratio of well width to sensor spacing is preserved relative to the 318 chip, then the well would be 0.58um wide, and the intersensor spacing would be 0.125 um.

This all assumes that there will be no increase in cross-talk between wells or sensors as the density increases.

Even with this best case scenario, the well width on the Proton II will be ~0.6um, which would require the use of 0.2-0.3um diameter beads. Perhaps that isn't so unreasonable?

0.2um beads seems very small given that they do not use the nano beads for SOLiD.

Wouldn't it make more sense for them to make a four-layer chip since they do not need any optics? That way they could use the same beads and essentially the same chips (The 660 M wells on Proton II is exactly 4x the proton I's 165M).

[krobison]

0.2um beads seems very small given that they do not use the nano beads for SOLiD.

Wouldn't it make more sense for them to make a four-layer chip since they do not need any optics? That way they could use the same beads and essentially the same chips (The 660 M wells on Proton II is exactly 4x the proton I's 165M).

I wonder how practical it is to load & deliver reagents to such a chip -- certainly an interesting idea, but with its own challenges.

[low99]

Actually the price on an ION is more like $250K by the time you add the One Touch and server that you need for each ION unit....LT tends to only talk about the price of the sequencer and never mentions the hidden costs of all the other equiptment

[BBoy]

I shouldn't plan for any NGS instrument to last longer than 1 year, maybe 2? I manage a NGS Core facility and looking at pricing to possibly purchase an Ion Proton.

[snip]

For 2012 there are 252 work days in the year.

Aaah, such an academic view of the world :-) There is a reason why extremely capital-intensive industries always operate 24x7x365 with >80% utilization of the equipment.

I wonder how practical it is to load & deliver reagents to such a chip -- certainly an interesting idea, but with its own challenges.

Yes, loading and reagent delivery is going to be a bear, to say the least. In addition they use transistors to read, and stacking transistors is nowadays done the crude way - you make two wafers and then you bond them together. There are very few commercial products that use this technology, mostly cell phone SOCs where footprint is of paramount importance. In the case of Ion it does not seem to buy you much - you now have two chips with more than 2x the manufacturing cost and less yield. The only benefit is the same horizontal footprint. Not to mention that supporting the data rates required to get this much signal off a small area is going to put severe stress on the I/O.

[Chipper]

Yes, loading and reagent delivery is going to be a bear, to say the least. In addition they use transistors to read, and stacking transistors is nowadays done the crude way - you make two wafers and then you bond them together. There are very few commercial products that use this technology, mostly cell phone SOCs where footprint is of paramount importance. In the case of Ion it does not seem to buy you much - you now have two chips with more than 2x the manufacturing cost and less yield. The only benefit is the same horizontal footprint. Not to mention that supporting the data rates required to get this much signal off a small area is going to put severe stress on the I/O.

I don't see how the amount of signal would be less with a single layer chip of the same size? I didn't say it would be easy to manufacture, but I thought the manufacturing cost would be lower if a single wafer design could be used for both chips, and that making the sub-micrometer beads compatible with long reads and sustained signals would be much more difficult. Obviously the layers / chips would have to be separated to allow reagent delivery, and also that having something capable of doing a $1000 genome / run has a tremendous marketing value...

[BBoy]

I don't see how the amount of signal would be less with a single layer chip of the same size?

All I/O of the current design is pushed to the periphery. There are only so many pins to feed signal/power/clock/etc in/out. There are ways to work around that particular problem but they cost money, yield, and complexity. Perhaps a parallel with current CMOS image sensors would be illustrative. The circuitry of an image sensor is very similar, as is the pixel size. At 3-5um pitch all the action is on the front side and the signal must be fed out on the periphery. The highest pixel count 35mm sensor on the market nowadays is from Nikon, 36 MPx. The need to get this much data off the chip is limiting the frame rate of the sensor to 4fps. That is about 150 Mbs, in line with what you get when streaming HD 1080p video at 60fps. In order to accurately acquire the base incorporation signal Ion is probably sampling in the 10s of Hz range. With a 10 MPx sensor they would need to be getting data off chip at the same rate as the Nikon D800. Double that via die stacking and you quickly get into serious numbers territory. Of course, I am pulling all this out of my behind, as I have absolutely no real information on what they are doing, I am just guessing.

[Geneus]

Of course, I am pulling all this out of my behind, as I have absolutely no real information on what they are doing, I am just guessing.

Your honesty gives you credibility, IMO.

[Elcannibal]

The ccd analogy to commercial cameras is not perfect. First, there ae tons of research grade cameras out there that have very unique speed sensors. Chemically etched ccd's and ultra high speed sensors using only ram transfer mode. The biggest difference here is that a digital sensor looking for an electron change has almost no depth and can really transfer fast. Most CCD sensors have to collect RGB AND also have to collect in linearity (8-10-12-16 bit sensors). Here, the only requirement is a threshold signal, a true digital transfer, sensitivity and linearity are probably completely removed for speed. If a movie image had only a black and white digital display it would be screaming fast resolution, at the expense of shades of grey...

[epistatic]

"For the second half of 2012, we continue to expect solid growth in our Ion Torrent platform, as we begin shipping Ion Proton Sequencers in September"

Any date set for release? Very interested in data quality and yield.

[ymc]

"For the second half of 2012, we continue to expect solid growth in our Ion Torrent platform, as we begin shipping Ion Proton Sequencers in September"

Any date set for release? Very interested in data quality and yield.

http://www.genomeweb.com/sequencing/oppenheimer-bullish-life-techs-ion-proton-instrument-anticipated-launch-nears

well, it is only good for exome for the sept one if it comes out. why not get MiSeq which can do the same but with better accuracy and proven platform.

wake me up when they come out with the one that can do whole genome 30x

[asaleh]

can anyone link run data.

I figure that R&D will need time to improve specs even after launch, but I would still like to look at a run.

[belfield]

Single cell sequencing

Can anyone tell me if the Ion Torrent system will be suitable for single cell whole genome sequencing? If not what system would be the best for such an experiment?

Thx

[steinmann]

Doing single cell genome sequencing is more about the library prep than the sequencer you use. In principle the Proton should do the job just as wells as the HiSeq. The more important question you should ask yourself is whether doing single cell genome sequencing is a good idea in the first place and what we learned from those two ridiculous cell papers.