MassGenomics

A comment on my blog from a Chinese reader revealed something interesting: MassGenomics is blocked in China. At first I took insult to this, and then also a bit of pride, that the content of this blog caught the interest of Chinese officials. I queried my liaison with the hosting company, Chris Clark of V-fluence, and here’s what he found out:

We confirmed your blog is being blocked in China (and India). At this point, we have no way of telling why your blog was blocked. It could be a keyword in your content was on a block list or more likely, because you have a blog. China has blocked Technorati and many other blog-related properties. At this point, without knowing the exact reason your blog was blocked, there really isn’t that much we can do. I would send a note to your reader asking him to try and keep up to date with your blog using an RSS reader, access your site through a proxy server or simply e-mail him your posts. Godaddy said there wasn’t much we could do.

So, no MassGenomics for China. Unfortunately, it seems that I’m not alone in censorship. My friends in the blogosphere dug into this issue. Daniel MacArthur of Genetic Future wrote:

Turns out a lot (possibly the majority) of blogs are blocked as a matter of course by the Chinese. A bunch of the people on Sb have had their external blogs blocked for no particular reason.

Chris Clark referred me to an interesting page on Wikipedia – the Golden Shield Project. It seems that the Chinese Ministry of Public Security (MPS) has spent ten years and hundreds of millions of dollars on a massive-scale censorship and surveillance project that limits and monitors internet access in China. It’s not just blogs that are blocked – sites deemed to be “subversive” by Chinese officials are walled off as well.

This begs the question – is it coincidence that my blog is blocked, or am I deemed a threat by the Chinese government? I admit I do tend to be a rabble-rouser. As David Dooling of PolITigenomics suggested, perhaps my misdeeds are finally catching up with me.

Yesterday was one of those gorgeous spring days in Missouri – warm and windy, with everything coming into bloom.  I drove over to the WashU non-medical campus to hear the annual Viktor Hamburger memorial lecture, given this year by Dr. Sydney Brenner.  The auditorium was filled to the brim, with attendees finally sitting on the steps and standing in the back of the room.  It was unsurprising not only because he’s a Nobel laureate, but also revered among the nematode research community which in St. Louis is quite large.

It turns out that Dr. Brenner, who was born in South Africa, has a charming accent and a bit of a rambling style, which he’s certainly allowed.  He is a clear proponent of what he calls “the scientist at the bench” – the lone researcher or graduate student working late into the night – and not a fan of large-scale genomics.  Dr. Brenner joked that many people say sequencing the genome was like sending a man to the moon, and he agrees, because “it’s easy.  The hard part is bringing him back.”

The Cell as the Correct Level of Abstraction

He went on to address the complexity of understanding biology at its various levels of abstraction – gene, protein, cell, tissue, organism, etc. – and argued that the cell is the “correct level of abstraction” to talk about the genome.  The genome itself presents a problem of complexity – 20,000 genes, each encoding one or more proteins.  If the genes are the templates, then what they encode might be called “instantiations” of genes, and can vary widely.  In fact, Dr. Brenner posited that the human genome may encode as many as 150,000 “instantiations” from 20,000 genes.

Reducing Genome Complexity

So how does one reduce the complexity?  This is a problem, Dr. Brenner said, that has already been solved by nature: by assembling small groups of proteins into sub-machinery complexes (e.g. spliceosome), the complexity is reduced by modularization.  He likened it to two watch-makers, one of whom assembles watches one piece at a time, while the other assembles pieces into subcomponents before putting them all together.  The second watchmaker is the one who will ever finish a watch.

Dr. Brenner filled the hour with little metaphors like these; there were no slides.  He was like a storyteller, and the audience listened with rapt attention.

Systems Biology: What They Should Be Doing

The speaker remarked that many questions asked by scientists are “inverse problems” – using observations (data) to infer mechanisms and function underlying them.  He gave as an example fellow Nobel laureates Watson and Crick, who had sought to use X-ray diffraction data to infer the structure of DNA.  They were all three at Cambridge at the same time, it turns out.  Unfortunately, the output of diffraction experiments provide a value that is the square of what’s being measured – the phasing information is lost, so there was no way to know if it was 1 or -1.  There is not enough computing power in the world to test all of the possible combinations, not even if they started calculating 2,000 years ago.  So instead, Watson and Crick took a more direct approach – they built models, predicted what the diffraction output of each model would be, and then compared it to the actual output.

Artificial Intelligence Combined with Human Stupidity

Omics sciences, Dr. Brenner argued, instead try to measure everything (e.g. sequence the entire genome) and hope that a magical computer program will provide the answers, something he called “Artificial intelligence combined with human stupidity.”  He said that such an approach to science was low input, high throughput, no output – where you put garbage in and get garbage out.  Instead, Dr. Brenner encouraged us to start with high input, which, he said, is what’s in your head.

At a cancer genomics seminar this week we heard from Michael Tomasson of the Siteman Cancer Center, whose interest is multiple myeloma (MM), a cancer of plasma cells characterized by renal failure with bone fractures.  It’s also uncurable, with a 3-year average survival following diagnosis according to the American Cancer Society.

Why Did Patient X Do So Poorly?

Dr. Tomasson presented a case, which I’ll call Patient X, a 64-year-old man who presented with anemia and back pain.  The anemia had been observed in 2003, when the patient was turned down as a blood donor.  In 2004 he suffered a “pathological hip fracture” – in other words, a break from a light fall that shouldn’t have caused one.  High IgA levels were observed in a blood sample, and a subsequent bone biopsy revealed 56% plasma cells.  By a FISH assay the patient was found to be positive for the 4:14 translocation, which is associated with a poor prognosis in MM.  The patient didn’t respond to “salvage Rx,” which I’m guessing is an aggressive chemotherapy, and died in 2005, just a year after diagnosis.

Epidemiology of Multiple Myeloma

Like many complex diseases, risk of MM increases with age; typical onset is between 65 and 70 years of age.  It’s also more prevalent in African Americans.  Recently, researchers discovered that virtually all multiple myeloma cases are preceded by “monocloncal gammopathy of uncertain significance (MGUS),” a common, age-related medical condition characterized by an accumulation of bone marrow plasma cells derived from a single abnormal clone.  MGUS does not promise MM, however; only a small fraction (~1%) of MGUS-postitive individuals develop multiple myeloma.

Complex Genetics, Largely Unexplored

According to Dr. Tomasson, the known genetic abnormalities seen in MM fall into two categories.  One is hyperdiploid amplification of odd-numbered chromosomes.  The second category is what I’d call large-scale structural variation: the 4:14 translocation and large deletions of chromosome 13 are seen in ~85% of cases.  There are other translocations (11q13, 6p21) involving IgH (heavy chain) genes that seem to be associated with better prognosis.

The known translocation (4:14) doesn’t appear to create a fusion protein (like BCR-ABL) but does result in dysregulation of numerous transcripts.  The chromosome 13 deletion alleles vary between patients, but most seem to fall between 13q11 to 13q34, so any of 50+ genes could be affected.  Dr. Tomasson showed a genome-wide Nimblegen aCGH plot, and it as apparent to the audience that there were numerous signals of interest – deletions on other chromosomes in addition to chromosome 13.  Li Ding pointed out an apparent amplification of chromosome 7, which we’ve seen in other cancers.  I asked whether or not telomere loss was characteristic in MM, because several of the chromosomes seemed to have deletions near the ends.

The bottom line is that there are lots of areas to explore.  Chromosome 13 is certainly of interest, but so are the chromosomes involving known translocations.  If someone were to zoom in to the sequence level (whole-genome sequencing), who knows what we might find?  At present, multiple myeloma is incurable, representing 1% of cancer cases and 2% of all cancer deaths, so it seems like we have little to lose.

Genome Technology’s Daily Scan picked up a very interesting article from the Boston Globe on personal genetic screening of cancer patients, which is to begin within the year at Massachusetts General Hospital.  Their plan is to screen tumors of 5000-6000 cancer patients for 110 mutations in 13 key cancer genes.  This is an important step for personalized medicine, and also, something that we should have been doing a long time ago.

Why All Tumors Should Be Genotyped

Despite the reluctance of some clinicians to rely on genetic information for personalized treatment, and the reluctance of most insurance companies to pay for it, the body of evidence supporting personalized genetics-driven cancer treatment is substantial.  The Boston Globe article cited one anecdote about a woman diagnosed with lung cancer who wasn’t responding to surgery and chemotherapy.  A genetic test revealed that the tumors were positive for EML4-ALK, a transforming fusion gene identified in 2007.  Luckily, a pharmaceutical company was testing a drug that targets the fusion protein.

Another anecdote supporting personalized cancer treatment was told at AGBT 2009 by Marco Marra.  They’d had an 80 year old male patient with papillary adenocarcinoma, a tumor of unknown origin that was found on his tongue.  He underwent surgery, but after 4 months there were lots of lung metastases (he showed a PET-CT).  The tumors had EGFR amplifications but didn’t respond to erlotinib, an EGFR inhibitor.  Light transcriptome sequencing revealed some clear genomic alterations in the PTEN/RET pathway (PTEN was lost, RET was activated, but not mutated).  Long story short, they treated the man with the renal cell carcinoma drug sunitinib, a tyrosine kinase inhibitor that targets RET, and the patient’s response in just a few months was near miraculous.

So Mass Gen is screening 13 genes in every tumor.  I’d hoped to learn which genes/mutations were being screened, and the equipment that will be applied, but unfortunately the Boston Globe article was sparse on the details.  (Please, post a comment if you have any).  I could make some educated guesses: EGFR, KRAS, RET, TP53.  Yet 13 genes seems like a rather paltry number, considering the number of new cancer genes being identified by large cancer sequencing projects.

Who Pays for Genetic Testing?

Of course, I admit that genetic testing can be expensive, and the fact that insurance companies are already balking makes this a serious concern.  From the article:

"...representatives of the state's three major health plans said they pay
for gene testing only when it has proven medical benefits, meaning
insurers may balk at paying for some of the new testing."

As expensive as some of these cancer drugs are, one would think that insurance companies would be all for testing patients a priori to pick the best one.  With the rapid advance of DNA technologies, it won’t be long before genetic testing is a lot cheaper than chemotherapy, and maybe then the insurance companies will come around.

Systematically Correlating Mutations with Drug Response

In the meantime, it looks like we have to go out and prove the medical benefits – that means screening large numbers of patients, and cross-referencing their mutational profiles with how they respond to treatments.  I wonder, can we test something like this in vitro – culture cells from various tumors, characterize their mutations in a number of genes, and then throw various cancer drugs at them.  Compare the mutations to the drug response, especially for known drug-target pairs, and you could rapidly build a cancer pharmacogenetics database.

High Interest in Fusion Genes / Fusion Proteins

It’s curious to me that the woman with lung cancer had a recently-identified gene fusion, and the drug company (ta-da!) already had a drug in trials that targets it.  Fusion genes seem to be a hot topic – especially those that involve protein kinases, like the well-known BRC-ABL fusion in leukemia.  It seems like we should be finding lots of these, and if they make good drug targets, so much the better.