dzhu

"It's remarkable that non [sic] of the investor or regulators paid half as much of attention."

Actually, the regulators paid just as much, if not more attention to Theranos - the article itself asserts this:

> On August 25, 2015, months before the Journal story broke, three investigators from the F.D.A. arrived, unannounced, at Theranos’s headquarters, on Page Mill Road, with two more investigators sent to the company’s blood-testing lab in Newark, California, demanding to inspect the facilities.

According to someone close to the company, Holmes was sent into a panic, calling advisers to try to resolve the issue. At around the same time, regulators from the Centers for Medicare and Medicaid Services, which regulates laboratories, visited the labs and found major inaccuracies in the testing being done on patients. (The Newark lab was run by an employee who was criticized for insufficient laboratory experience.) C.M.S. also soon discovered that some of the tests Theranos was performing were so inaccurate that they could leave patients at risk of internal bleeding, or of stroke among those prone to blood clots. The agency found that Theranos appeared to ignore erratic results from its own quality-control checks during a six-month period last year and supplied 81 patients with questionable test results.

Benchling's awesome - definitely a well-designed (and much needed!) tool for biologists. They've done a great job with adding new features, from CRISPR design to gel analysis, and their latest electronic lab notebook update is a compelling step towards becoming the central hub for designing, storing and sharing experiments and data.

Actually, they DID collect clinical response data even though it was a Phase Ib trial. Specifically, as the NYT piece mentions, they measured both cognitive function (via the mini-mental state exam aka MMSE and the clinical dementia rating scale aka CDR) and plaque removal (using one of Lilly's imaging agent) in relation to dosage. The data shows consistently improving cognitive function and plaque removal as dosage increases (albeit at the cost of additional side effects).

Now this is certainly atypical for Phase 1b data but it does (to some extent) justify Biogen increasing in almost $10B in market cap. In fact, the Phase 1b data is compelling enough that Biogen is going straight to Phase III trials.

While it is healthy to be skeptical about early trial results and subsequent market movement, I think it's clearly misplaced here.

They chose to encode one bit per base (rather than two) to give them flexibility in encoding the binary values to avoid particularly tricky sequences for DNA synthesis or sequencing (such as ones that have many repeating bases, e.g. AAAAA, which are known as homopolymers, or ones that have high GC content).

"Then you repeat the test."

What if the test is inherently flawed?

"look at what this drug did to my grandma"

Uh, that is purely anecdotal, n of 1 data, which is not scientifically valid evidence. This is what scientifically valid evidence begins to look like: http://www.nejm.org/doi/full/10.1056/NEJMoa1310669, which suggests the _opposite_ of what 23andMe says about warfarin sensitivity.

The point isn't that we shouldn't have tests for things such as warfarin sensitivity or statin sensitivity; it's simply that we need to have clinically VALIDATED tests for these medical conditions. Otherwise, without a clear understanding of the analytic and clinical validity of these tests, it's possible that many false positives or false negatives arise.

There's a reason why other genetic tests (e.g. Genomic Health's Oncotype DX or Myriad's BRCAnalysis) haven't been pulled by the FDA (in fact, they're not even regulated by the FDA!) - that's because they've gathered sufficient data to support the clinical outcomes and utility claims that they've made. Put more simply, these other molecular diagnostics have actually verified their science. 23andMe has not - that's why they're being regulated.

Warfarin sensitivity is _precisely_ why 23andMe needs to be regulated; 23andMe has NOT clinically (or analytically) validated any of the claims around the specific SNPs they've identified. All 23andMe does is link to research papers that may suggest that there are gene variants that are linked to warfarin sensitivity; however, these have not been validated to be clinically significant.

In fact, just last month, the New England Journal of Medicine published a landmark study around the pharmacogenomics of warfarin dosing, which found that this very genetic data does NOT affect/improve warfarin dosing (http://www.nih.gov/news/health/nov2013/nhlbi-19.htm, original paper: http://www.nejm.org/doi/full/10.1056/NEJMoa1310669). In light of this, this warfarin sensitivity information that you've obtained from 23andMe is almost certainly a false positive, and if you were on a prescribed warfarin treatment, 23andMe would've likely resulted in a change in your behavior that would be strictly detrimental to your health.

Ah, this is the original article (everything else uses this source as their reference). I forgot that my university has a Science subscription; I'll change it to the abstract, which should be viewable, instead.