After I submitted the previous article on this topic (here), the science editor of the New York Times, Mark Hammond, posted a blog (August 12) with some comments on the Helicos BioSciences (HCLS- Nasdaq) items in the press, namely his paper (here) and The Los Angeles Times (here). The title of Mark Hammond’s blog article is “Inflated claims for the $50,000 genome”. You can read all of Hammond's discussion here.
Additional news and analysis, with emphasis on investment opportunities, and including an interview late yesterday (Aug. 12) with Steve Lombardi, President of Helicos BioSciences, was published this morning (here).
There are several items in Hammond’s piece that add to what appeared in my article published by Seeking Alpha yesterday (here). I will list and discuss topics from Hammond’s post one at a time.
1. Hammond wrote (quotes):
a. This attention is not entirely unjustified, as Quake's genome is somewhat different from the other seven that have gone before: it was the first to be sequenced using "third generation" technology, which works with single molecules of DNA rather than with samples that first have to be amplified.
b. The team used a sequencer called the Heliscope, made by Helicos BioSciences, a company co-founded by Quake. This, it was claimed, substantially cut the costs of producing the sequence (Quake's genome cost about $50,000, compared to about $1m for Watson's and $10m for Venter's), and allowed it to be done by just three people in a small lab.
c. Quake's sequence is indeed an important milestone in genomics, a staging post on the route to the $1,000 genome. Third generation sequencing platforms like Helicos's (rival technologies are being developed by Oxford Nanopore and Pacific Biosciences) are going to be critical to getting the costs of sequencing down to that level, so it becomes affordable for medical purposes. Some of the coverage, however, and the PR material that prompted it, has been at least a little disingenous.
I have problems with this sequence on some counts, while find it provides very valuable additional information on others.
a. The implication of this statement is that “third generation” technology is not a significant advance. I am not sure the implication is justified. Is the third generation not a significant advance over the second? Maybe Hammond is referring to the error rate differences, which I will discuss later.
b. Later in the post Hammond offers some information to support discussion about whether $50,000 really did substantially cut the cost of sequencing the genome for an individual. Information he presents indicates $50,000 was not less than others can do this for today.
c. The other technologies cited are research efforts. There is nothing commercially available form these two sources, nor can I find any timeline for producing commercial scale-up on either of their websites.
2. Hammond wrote (and I quote):
a. First, there's the cost. A genome for $50,000 -- which Quake compared to the price of a sports car -- sounds pretty good when it would have set you back millions a couple of years ago. Yet far from "lowering the cost" of sequencing, as the New York Times headline declared, cheaper genomes are already available.
b. In June, Illumina launched a commercial genome-sequencing service with a price tag of $48,000, using standard second-generation sequencing -- a proven technology that will, for the moment at least, get you better results. A Helicos genome costs about the same for a more experimental technology, and isn't yet available on the open market. Complete Genomics even offers a cheaper service, though to institutions, not consumers: it has charged the Broad Institute $100,000 for five genomes.
c. The $50,000 figure also ignores the cost of buying a Helicos machine: it sold its first instrument for close to £1m, and though it reduced the price by a quarter this year, $750,000 is hardly cheap. David Dooling at Politigenomics also points out that substantial depreciation costs are left out. Another hidden cost is that of lab modifications that are needed to house the Heliscope: as the sequencer must be mounted on a slab of granite to ensure microscope stability, the whole thing weighs so much that structural reinforcements are sometimes required before it can be installed. It's misleading to describe this as a budget, three-person job.
I have several things to point out here:
a. The implication in the New York Times article that Hammond criticizes, is that that third generation technology processes have the promise of drastically reducing the cost of obtaining an individual genome. Hammond criticizes the $50,000 demo cost without commenting on the lower cost potential. Does he feel that second generation processes also have that potential? Since they are quite mature, I find it not likely that they have the $1,000 (or even $5,000) potential for an individual genome.
b. Helicos has not estimated what the cost potential for their machine is. It is clear that they think it is much lower than the $50,000 demo cost. While no one is offering individual genome sequences, Steve Lombardi, the President of Helicos, told me in an interview late yesterday (Aug. 12) that seven machines have been delivered to date. It is not known if any of these will offer services to the public.
c. I discussed the capital cost/processing cost ratio in some detail in the previous Seeking Alpha article. It is not clear what went into the $50,000 cost for Dr. Quake’s individual genome sequence. If I divide $50,000 into twelve parts (3 people for 4 weeks), the cost is $4,167 per week per individual. I do not think it is reasonable that salary plus overhead for these three people would be over $215,000 per year each. If that is the case, the cost would certainly come down in a routine processing environment to less than half of $50,000. I expect that the labor content of the $50,000 could be of the order of $20,000 - $25,000 with $25,000 - $30,000 applied to materials, supplies, facilities, maintenance, and capital depreciation. This seems to be in the same ballpark as David Dooling at Politigenomics, the reference provided by Hammond. I do not see that Dooling offers anything to cause skepticism regarding the cost figures presented by Dr. Quake.
Although the title of Dooling’s article hints at derision (“Another Rich White Guy Sequences Own Genome”), his discussion does have a lot of meat. And he correctly cautions about the speculative nature of HCLS stock, although again can not resist the temptation to make his point sarcastically.
3. Hammond wrote:
a. There are also technical issues with the Quake sequence, which are set out in excellent analysis pieces from Daniel MacArthur at Genetic Future and Kevin Davies at BioIT World. It covers just 90 per cent of the genome, and the error rate is significantly higher than with second-generation sequences.
b. Don't get me wrong, the Quake genome is an important step forward, which proves in principle that at least one of the third-generation sequencing methods works at an acceptable cost which should rapidly fall. But while it might well be the future, this technology isn't ready to go head-to-head with established sequencing platforms just yet. As Daniel MacArthur notes, some of the coverage seems to have bought the Helicos PR a little too uncritically.
There are both clearly stated and confusing points made by Hammond here:
a. Hammond has confused error rate and extent of genome mapping in this paragraph. Reading MacArthur’s paper, I found that Helicos’ demo did not attempt a more complicated analysis process (it was run six months ago) to compare to the process run by Illumina that provides a larger genome coverage (99.9%). The demo by Helicos covered 90% of the available genome. MacArthur speculates that Helicos may have since run the more complete process. This discusses the area of extent of mapping (or coverage). One thing that is confusing: Is the 90% referring to all of the genome or to the 92% characterized by the Human Genome Project? Is the 99.9% referring to the 92%? If the two reference points are 99.9% of the 92% and 90% of 100%, the two coverages are much closer than it appears at first blush.
MacArthur also wrote the following about errors: “As a result of this depth of coverage and the generally low rate of base-swapping errors (as opposed to deletion errors), their accuracy for calls of single-base variants (SNPs) seems quite reasonable. They could call 97% of SNPs with 99% accuracy, which is still worse that second-generation approaches but not terrible for a rough draft genome.”
In my previous Seeking Alpha article I pointed out that Quake represented his error rate as 5 out of 100,000, and the desired level was specified as 1 out of 100,000. I can not relate this directly to MacArthur’ statement above, but it seems to me that 5 out of 100,000 "sounds" considerably better than “97% of SNPs with 99% accuracy”. Has Dr. Quake misrepresented something or is my ignorance the problem, not being able to reconcile MacArthur’s accounting methods with Quake’s?
b. Hammond is probably right: this technology may be the future. We shouldn’t have to wait long to find out. With seven machines already sold, research and development results should be made public soon. The proof will be in the pudding, not by what news reporters and bloggers write.
More was published this morning (here).
Disclosure: No positions in stocks discussed.