Friday, August 24, 2012

The Perils of the Too Small Gene Panel

The newswires are alive with summaries of an item in Science Express which, discounting the supplementary material, isn't much bigger than the news articles describing it.  It's a nice piece showing the relevance of cancer genomics, but there's also a backstory fleshed out in the news items which is interesting.

The gist of the paper is this: a trial of everolimus, an MTOR inhibitor, in bladder cancer did not go well; very few patients showed benefit.  But, one patient did spectacularly well.  So the researchers performed whole genome sequencing and found inactivating mutations in the well-studied tumor suppressors TSC1 and NF2.  Screening 13 more bladder tumors with a panel of cancer-specific genes found 3 more cases of inactivating mutations in TSC1, plus one patient with a missense mutation of unknown significance.  In the trial, patients with TSC1 mutations stayed on trial longer than patients without the mutations.  

My first reaction was "this all makes sense" -- TSC1 and NF2 are genes which immediately suggested themselves as TOR-related. 

What's interesting from the news items is a suggestion that the original patient had been sequenced for a limited number of genes around mTOR, and that this did not include TSC1 or NF2.  Of course, the problem when doing a limited screen is picking who to include, and from my 2 second analysis on the train I would have included TSC1 and NF2.  But, that could be much easier said than done.  They are attractive, since they are tumor suppressors known to be mutated in cancer, but so are other genes in the neighborhood (such  as TSC2 or PTEN).  There are activating mutations known in the neighborhood as well, such as PI3K or various AKT family members.  Presumably it was a PCR-based (quite likely Sanger) method, in which case it can be challenging to target every exon both because you may have an "exon budget" (number of exons to be amplified) and some exons are nightmarish to amplify.  

I think the case illustrates one reason whole genome or whole exome sequence are by far the best strategies in a case such as this.  The potential payoff in understanding is huge, as you have one strong outlier patient.  The number of patients are small (though, admittedly, this is probably one success pulled from many dry holes).  Plus, these days the cost of WGS/WES is probably not much more than targeted PCR, given the costs of developing good PCR assays.  

The other potential advantage of WGS/WES, over even broad cancer-specific gene panels, for a case such as this, is that the field can change.  New oncogenes and tumor suppressors are identified periodically, perhaps even in the time period between when a panel is designed and it is used.  In a research setting to understand the basis of a clinical trial anomaly, it's particularly valuable to explore all corners, because what might not make sense today might become clear tomorrow.


Anonymous said...

People like yourself and I who do omics work and look at cancer samples from a big picture view know there are all kinds of things wrong within a tumor. We continue to take a single gene approach for very complex diseases. No wonder drug discovery successes in oncology are hard to come by. Focusing on one gene or even an oncogene panel misses other potentially important factors, such as mutations in drug metabolizing genes or drug pumps. Yes the cost is greater for WGS/WES but we're missing the forest for the trees.

Anonymous said...

Easier said than done.

A routine WGS or even WES of a cancer would currently produce a vast amount of information and no one knows what it means. Figuring out what sequence variant is of relevance is difficult enough in classical human genetics and may take weeks and months, let alone in cancer genetics. Usually you don't want to wait that long for treatment, even if it is an experimental treatment!

Also note that in cancer sequencing these days, you like to have a high coverage, because you often have a very mixed population of cells to work with. 200-500x coverage is not uncommonly requested. So if resources are limited (and of course they usually are), it does make sense to look closely and quickly at potential changes that you really understand instead of spending a lot of time looking at it all with mediocre sensitivity and not knowing what to do with the data after all.

Also, I am not sure whether in this specific case an a priori knowledge of the TSC1 and NF2 mutation status would have helped. As a retrospective explanation why the one patient did so well, this is certainly very informative. But while it might seem logical in hindsight, I am not so sure that anyone would have guessed this result beforehand. Cancer genetics and drug action are just too complex, unless you are looking at direkt binding targets of a certain drug. Noone would have excluded the other study participants a priori because of lack of TSC1 Mutation....

So currently, I would say studies are warranted which link WGS/WES results to clinical outcome in standard or new cancer therapies. Some WGS/WES studies in cancer have been done and linking the data to clinical outcome is certainly being done. But this is currently retrospektive and not yet as a tool for a priori risk stratification and therapy decision.