4,759 karma · joined March 29, 2010
Opinions posted here are my strictly my own, and not representative of any employer or organization that I am a part of.
https://tedpak.com
[ my public key: https://keybase.io/powerpak; my proof: https://keybase.io/powerpak/sigs/fEW8iI_3tD3E9XjbSaPJTDOtaai7TXU6j1QAezE82Xc ]
I agree, some of the opinions tossed around by incumbents in the industry are mind-blowing. My favorites: "A PDF is a weapons-grade tool for piracy." ...And on parents of sick kids looking up articles: "Do you really think a search and downloading of technical medical literature means anything?"
hops on hipster hashrocket and tribbles off into the sunset
In the second caption: "Each cell of JCVI-syn3.0 contains just 473 genes, fewer than any other independent organism."
In the main text: "In a 1995 Science paper, Venter’s team sequenced the genome of Mycoplasma genitalium, a sexually transmitted microbe with the smallest genome of any known free-living organism, and mapped its 470 genes."
Which is it? Venter would have seemed to disproven his own claim to novelty, unless we've found new genes in M. genitalium's genome since 1995.
Edit: As with most biological terms, part of the problem is the fuzziness of the definition of "gene". More recent studies claim M. genitalium has 525 genes [1], but that might be including tRNA and ncRNA regions. I'd still object to the article's poor editing. Also, let's get down to brass tacks here: we've only trimmed a 580kb genome down to 531kb (9% reduction). Clearly, life is already pretty damn efficient.
What is a bit sad is that data analysis and quantitative skills continue to be de-emphasized in the pre-clinical curriculum for the sake of rote memorization of huge lists and tables, which would seem to be less and less important as instant access to references on smartphones becomes commonplace.
1. The good is a commodity, i.e. its labeling has no effect on perceived value
2. The selling firms are perfectly competitive (i.e., no partial or complete monopolies that allow price-fixing)
3. The buyers are perfectly competitive (i.e., no purchasing groups)
4. All sellers can sell to all buyers equally easily
5. All parties in the market have equal access to information
6. All parties act rationally and based only on the marginal utility of each individual transaction
etc., etc., and a whole litany of other things that we know are rarely true. Since the theory is so "squishy" in practice, one could even protest that it does more harm than good when treated as a de facto starting point for price modeling: https://en.wikipedia.org/wiki/Supply_and_demand#Criticisms
So this model, even if it too is sadly oversimplified, may still be useful simply by starting from a different set of assumptions than whatever is typical in the field, and seeing how the modeled behavior differs from existing theories. There's basically no hope of experimental validation for any of this, though.
But yes, I would recommend that academics check out http://zenodo.org/ which not only helps with the tagging of a github repo but also provides a citable DOI.
While we're at it, we might as well point out that source code alone is useless if it doesn't come with instructions on what kind of environment is required to install and run it, as well as the actual input data files used by the authors.
The point of using an attenuated virus is that this is a "challenge model" that is supposed to simulate an exposure to the real virus closely enough that one can judge if the vaccine is likely working, without exposing the patients to the risks of infection by the full-strength virus. CNN did a much better job than Wired in explaining this. http://www.cnn.com/2016/03/16/health/dengue-vaccine-effectiv...
Therefore, the fact that the placebo group had certain levels of various symptoms (according to the article, 80% had a rash, 20% had nausea, etc.) is still important data for other scientists who are trying to judge if they should use the rDEN2Δ30 challenge model in their candidate vaccine trials. It's not intended to provide data on what natural dengue fever is like, which I agree would be unethical and unnecessary.
Yes, it does, precisely because the trial is about a human challenge model that uses an attenuated virus. It may or may not be representative of efficacy against natural infections by wildtype viruses, hence the second statement by the authors that "the true efficacy of TV003 can only be established by performing a phase 3 trial in endemic areas."
> it doesn't seem like Wired has editorialized/misrepresented at least the headline much relative to the original article
I believe the inclusion of "in a human challenge model" in the original title is quite significant to its interpretation, but it's precisely the kind of detail that would get overlooked by people unfamiliar with vaccine trials. A researcher in the field would know that human challenge models may or may not be representative of protection against natural infections, but the Wired title ignores that distinction and instead sows confusion by saying the vaccine simply "proves 100% effective".
> millions/year
You're right, I meant hospitalizations; sorry about that. Dengue hemorrhagic fever is a life-threatening condition that warrants hospitalization, and dengue fever in more vulnerable patients also can result in hospitalization.
The whole challenge in creating a dengue vaccine is that it has to protect against all four serotypes (called a tetravalent vaccine). Incomplete protection against one of the serotypes could potentially lead to an even worse infection if you are infected, through a process called antibody dependent enhancement: https://en.wikipedia.org/wiki/Antibody-dependent_enhancement This is why vaccine development against dengue has been so painstakingly slow.
The candidates in this trial were only challenged with one serotype, DENV-2. Therefore, we have no way to know if the tetravalence of the vaccine actually achieved efficacy against all four serotypes (although promisingly, antibodies against all serotypes were seen in 92% of the patients). The sample size is also quite small (N = ~20). This is why the original article states its impact conservatively as "This model may serve as an early check for dengue vaccine candidates, limiting the risk of conducting large unsuccessful trials" and furthermore "the true efficacy of TV003 can only be established by performing a phase 3 trial in endemic areas." Here's how you write a more honest headline and summary for lay people: http://www.cnn.com/2016/03/16/health/dengue-vaccine-effectiv...
I don't intend to detract from the achievement here, but I don't like when headlines oversimplify the implications of a clinical trial. It sets unrealistic expectations and decreases the public's trust in proper vaccine research.
Disclaimer: I am doing research with a lab that studies dengue virus, and we have collaborators that are working on a commercial vaccine.
I did trim the strange hash off of the medium.com URL, because I thought it was a tracker from when I got referred. Perhaps that is why more people eventually found this submission. The de-duping logic on HN is pretty bad.
The Gemfile contains the top-level dependencies that the app needs; the Gemfile.lock ensures that everybody developing or deploying is using the same gem versions for top-level and resolved transitive dependencies. Periodically one can `bundle update` to upgrade gem versions: http://bundler.io/man/bundle-update.1.html
It does continue to surprise me that distinctions like these are not handled by Python tooling, which has this absolutely sordid history around packaging, which continues aplomb in the wheel vs. egg wars... http://lucumr.pocoo.org/2014/1/27/python-on-wheels/
If more people start using PyPy for ad-hoc data analysis, it could help nudge it out of its niche into the mainstream.
Check out this infamous thread: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=30475 ...it gets pretty nasty!
The takeaway is that although overflow is technically undefined, a lot of security-related code was naively implemented by checking for overflows occurring in the most typical way (wraparound), and this check began to be removed by this sort of optimization. This made some pragmatically minded security folks quite angry, because although such behavior is technically undefined by the spec, it was the most widely used method of defending against overflows, and recognizing that such code was already everywhere in the wild they feared all the vulnerabilities that would be introduced by gcc dropping branches for i > i + proposed_increment.
R, despite being one of the first languages a budding "data scientist" might want to use, is probably not one of them for the many reasons given, among them:
- there are way too many ways to do everything
- implicit iteration (although great for statistics) makes performance issues hard to spot
- the data structures are a bit too flexible (it is Lisp-y in places), and you really need to understand them all to deploy the *apply and plyr functions effectively
- 3+ object-oriented programming systems
- non-standard evaluation. It's all over popular libraries like ggplot2, because it increases terseness, but it just looks like magic to beginners.
Basically, all the chapters listed here [0] -- which happens to be a great guide for experienced programmers to really understand R as a language -- happen to be the same reasons beginners give up too quickly.
Python, although it sufficiently nags me with its one-way-to-do-it motto and its many warts [1] to not want to use it regularly, is just well-rounded enough that it is a much better language for beginners. With Anaconda and iPython installed, I've found that a total programming beginner can actually get productive pretty quickly, even on stats and math problems.
The paragraph from NEJM's piece that drew the brunt of the scorn was the following:
A second concern held by some is that a new class of research person will emerge — people who had nothing to do with the design and execution of the study but use another group’s data for their own ends, possibly stealing from the research productivity planned by the data gatherers, or even use the data to try to disprove what the original investigators had posited. There is concern among some front-line researchers that the system will be taken over by what some researchers have characterized as “research parasites.”
The meme has already given birth to at least one parody account: https://twitter.com/dataparasite and one rather tongue-in-cheek domain purchase: http://researchparasites.com
There are some regrettable examples of cells and tissue being taken from patient specimens without their knowledge, some of which give birth to entire fields of research, and then it is later unclear whether data that could re-identify the patient's family (like genomic data) can be published [1]. It can be really hard to fully de-identify imaging and molecular data.
[1]: http://www.nature.com/news/deal-done-over-hela-cell-line-1.1...
Most of the HN crowd (and myself) would probably say that the ideal situation is for patients to own copyright and other rights in medical data generated about them, but I don't think this issue has actually been settled, either in a cultural or a legal sense [2].
[1]: http://www.hhs.gov/hipaa/for-professionals/privacy/guidance/...
It would be a very interesting future where healthcare workers have the reams of stats that professional athletes do, and where everybody in the space is literate enough to intuitively understand each number and its caveats, like hardcore sports fans. For example, an ICU nurse is bound to encounter more hospital-acquired infections (HAIs) per week than a nurse in pre-op assessment, and so averaged cases of HAI per primary unit is the proper context for that stat, much like you'd weigh recent rushing yards for an NFL linebacker against the strength of the defenses faced. Then, managers could optimize their teams to each person's strength or weakness and/or supply training in the correct areas, much like coaches in sports do. This would be a sea change in the medical culture, though, and it would encounter staunch resistance, because some of the metrics already being pushed on doctors (# pts/day) arguably incentivize worse care.
To your last point, part of my research now is analyzing electronic medical record (EMR) data, and coming from any other field you'd be shocked at how messy it can be. In a way, this is to be expected, because the "vocabularies" for medical data are huge, and even widely used ontologies have bizarre properties [1]. I really hope that we can solve the data collection problems with better technology, but as of now, the incentives for EMR vendors to overhaul the tech is pretty low, now that most US hospitals have picked a system and those vendors have everything to lose and little to gain from overhauling their user interface.
[1]: http://www.healthcaredive.com/news/the-16-most-absurd-icd-10...
Given that doctors in most other fields spend a ridiculous (by some metrics, 7:1) ratio of time doing paperwork or data entry vs. seeing the patient [1], I would suggest that there is serious opportunity cost in adding more self-measurements and checklists to most doctors' workload. Surgeons are likely an outlier that perhaps benefit most from being regressed to the mean.
[1]: http://well.blogs.nytimes.com/2013/05/30/for-new-doctors-8-m...
I'm trying to tell you that your chosen definition of "profitable" is utterly meaningless in a discussion of economics, but you have completely closed your mind to this idea. You don't seem to understand inflation, interest rates, opportunity cost, or the difference between accounting profit and economic profit [1]. Therefore, I can't proceed with this conversation.
[1]: https://www.khanacademy.org/economics-finance-domain/microec...
Please see any [1] of [2] these [3] to understand why it isn't as simple as "Y - X > 0", considering things like where the repayments are coming from, comparing the nominal annualized return to Treasury bond rates, all the fraud that will never be accounted for, etc., etc. Even pro-bailout articles [4] have to include weasel phrases like "on a risk-adjusted basis, even [such and such proposed Y - X] isn’t that big a profit, given the huge downside the government had," speaking directly to my point of how it's hard to properly account for risk, which is what any other lender would have done. Arguably, since there weren't any competitors to the US government for providing bailout money, the interest rate we use to judge whether it was "profitable" should be sky high.
[1]: http://www.nationalreview.com/article/395822/overselling-tar...
[2]: http://www.huffingtonpost.com/2012/04/25/tarp-profit-a-myth_...
[3]: http://www.forbes.com/sites/halahtouryalai/2012/04/25/dont-b...
[4]: https://www.washingtonpost.com/business/economy/bailout-high...
Here's an accounting trick for you. Why don't you give me a $1000 loan for twenty years, and since you don't know me, the risk of default is, let's say, pretty high. I'll repay you $1001 in 2036—heck, just to be generous, I'll do it in inflation-adjusted dollars. When I do that, I will be happy to have you tell me about the "profit" that you made off of your investment.
[1] http://web.archive.org/web/20090110184334/http://www.uiowa.e...
[2] http://web.archive.org/web/20090207100935/http://cop.senate....