Why cancer diagnostics aren't simple.
Alright, I didn't know exactly how to title this post, which is going to be linked to a few different boards. There is a study out in Nature (Nature and Science are the big two for scientific publishing) that looks at whole genome sequencing in the context of breast cancer. Actually, a subset of breast cancer, 'oestrogen-receptor-positive breast cancers'. The subset was used presumably because these are harder to treat and may have a similar underlying identifiable cause that could help identify them early.
What they did was sequence 50 tumors (with 30X coverage) and of course also healthy tissue from the same 50 subjects (always have to do this control to see changes). To cut to the chase, here is the title from Nature News.
Decoding of ten trillion bases yields no simple patterns or silver bullets.
Such a study, done in the context 'two clinical trials of oestrogen-lowering drugs known as aromatase inhibitors', showed 'of the 1,700 gene mutations they found in total, most were unique to individual patients' tumours, and only three occurred in 10% or more.'
So much for establishing an simple diagnostic, or predictive marker. Or for the use of sequencing as the diagnostic tool, which too often yields an 'isn't that interesting' result rather than a validated A or B choice.
Anyway, here is the news story:
Now I love that they did this. Seeing more of this as sequencing gets cheaper is going to happen in my opinion, particularly in clinical settings. It is good news for companies like Illumina, Life Technologies, and potentially Pacific Biosciences who are driving down the cost of whole genome sequencing. Also Complete Genomics who is looking to offer this as a service.
But it doesn't mean easy answers are going to start popping up left and right. Some will prove fruitful, others potentailly frustrating. Diagnostic test development and validation is a tough path, even in the context of a clinical setting. But .. even if one of those > 10% signals is very predictive of drug efficacy, it is a huge advantage on when to give (and not give) a certain therapy. Genomic Health, whose business is currently built around diagnostic genomic testing for breast cancer treatment (and which I own) has really achieved something impressive, but not easily duplicated.
"The results are complex and somewhat alarming, because the problem does make you sit down and rethink what breast cancer is," says Ellis, leader of the breast cancer programme at the university's Siteman Cancer Center.
I'll say. Not to mention how to recognize it early and determine the right therapeutic approach when it appears.
Nonetheless, he says, there is reason for optimism — not least because careful analysis of the data, combined with what is already known about the functions of the affected genes, yields a wealth of new therapeutic possibilities.
I can see all of pharma rolling their eyes. They would call this an expense of new therapeutic possibilities, many of which won't pan out as leading to effective treatments.
Ellis says that the complexity of their results indicates that when it comes to developing therapeutics "very clearly the only way forward is the genome-first approach. No single blockbuster drug will answer the problem of endocrine-therapy resistance".
I struggle to have this same conclusion. Genome first should be tried, but the results argue that other strategies may prove more fruitful. Perhaps IDing an overexpressed protein receptor that can be exploited, but whose increased level (decreased turnover) isn't predicted by the genome but perhaps it is from mRNA levels (express-ome?). Ideally a single drug that would help the greatest % of the endocrine-therapy resistance population can indeed be found, and who it should be given to also determined. I don't care whether the test is genomic or not, and worry a little that such researchers who have only hammers are seeing all problems as nails, because they aren't.
Finally: This was one of two mutations already associated with breast cancer that occurred frequently in the 50 tumours: PIK3CA was found in 43% of samples and the tumour suppressor TP53 turned up in about 15%. All told, about half the cancers carried a combination of these three mutations — leaving half with cancers arising from varying constellations of much rarer mutations.
If any one of these most 'common' changes or the group together prove predictive of the efficacy of the aromatase inhibitors, then the entire effort is alarge success (which then needs to be confirmed in another trial). It seems we don't yet know that important piece.
Home Coverage Fool