Other than the obvious loss of information, I'd be interested in knowing what pitfalls come up that are specific to this case. To make things a little more concrete, take the case of a GPCR dopamine receptor. Supposedly dopamine (or one of a variety of drugs) interacts with this receptor in such a way that a different region changes conformation which in turn alters the conformation of a G-protein so that it binds GTP. This all seems to require very specific protein conformations and I can see how observing the average of many could be misleading.

Well, the main pitfall is what I already mentioned - if the conformations being averaged really are very different, this will decrease the effective resolution of the reconstruction, which will be very obvious even to an untrained eye. EM used to be notorious for producing vague blobs, in part because of the limitations of the technology (before they had direct electron detectors and had to use film), but also because the software tools (and users) weren't as good about picking out different conformations. And when the individual protein domains (often of known structure) resemble spheres in the reconstruction, it's difficult to tell that something isn't working. So it was indeed possible to generate a map that was a misleading average, and there are probably structures like that out there. But that's why everyone relied on crystallography for detailed structural information.

The good news is that at the resolution range people are using now, it should be possible to build individual structural components, but only if the particles are nearly homogeneous. So the ability to build (or dock) atomic models that clearly fit the map on the level of individual amino acids becomes a test for whether the averaging is justified.

The case you mentioned isn't really applicable, because the GPCR only assumes that conformation when bound to dopamine, and tends to work like a molecular switch. And of course if we did have a range of conformations being looked at, the reconstruction would resemble a soup can, without any atomic detail, which isn't really a publishable result. GPCRs are so small that it's currently better to use crystallography, but there are indeed structures of GPCRs in various static states at high resolution.

The remaining problems are that a) proteins aren't really static and b) the experimental methods for structural studies may induce non-physiological artifacts. I don't think (a) is that much of a problem because we have plenty of ways of studying protein dynamics and everyone is implicitly aware of this limitation anyway. The second problem is potentially worse: purification can sometimes have weird effects, crystallization packs molecules into a lattice that may not represent the native conformation, both crystallography and EM typically work at cryogenic temperatures which is known to change the structure in various ways (mostly but not always subtle), radiation damage can have side effects too. Much worse are the older "negative stain" EM structures where the proteins were covered with uranium or something similarly massive and sandwiched between thin sheets of carbon. Fortunately this is much less common now that cryo-EM has gotten so much better.

Ultimately the value of any model is determined by its ability to explain biochemical data and suggest new testable hypotheses. That's ultimately the most important way to validate their accuracy, and researchers ignore it at their peril.

I really do hope the people doing this have examined and discussed the possible pitfalls of drawing conclusions from averages of this type of data.

Yes, the pitfalls are well known, mostly because this has always been an issue with crystallography as well - it is impossible at present to determine the 3D molecular structure from a single molecule, so we are always dependent on either crystalline diffraction or averaging thousands of images to obtain the density map. (NMR has its own, well-understood problems.) The good news is that we known enough about macromolecular structure to be able to make subjective judgements very quickly based on the level of detail in the maps. (There are also plenty of higher-resolution structures of many smaller components, so we can calibrate our expectations based on known parts.) If the molecules are very heterogeneous, or the averaging is done poorly, the maps will not display known high-resolution features such as secondary structure or amino acid sidechains. For crystallography, there are also ways to calculate the deviation from the average (and I presume EM either has something analogous, or will soon). It is also common for some regions to have higher "local resolution" than others, and this can be quantified in various ways.

These methods still lose information - if, say, 10% of copies of a particular loop or sidechain are in a different conformation, this will probably not be captured. But EM experts have gotten much better about identifying clusters of similar conformations, at least on a larger scale. And in the end, the static average structure is still vastly more useful than no structure at all. Scientists can and do still publish spectacularly stupid interpretations occasionally, but these aren't due to the misuse of averaging, but rather to pure incompetence and wishful thinking.

There's another technical objection to the summary: "atomic resolution" in this context isn't the same thing as "imaging individual atoms". The actual cryo-EM images themselves are much noisier and do not have nearly this effective resolution - it is the average of many thousands of images that gives you atomic resolution electron density maps. (The same is true for X-ray crystallography, although you start with just Fourier amplitudes there, not images.) That's not a slam on the paper, which is an impressive technical achievement, but as the authors note, many conformationally homogeneous single particles (i.e. protein complexes) are required to get a map of this quality. Any differences between particles will simply be averaged out, and the more different they are, the worse the resolution.

PS. US scientists are also pursuing embryonic gene therapy (albeit using different technology). Of course, because they're not simply trying to win massive publicity, and want to actually understand the system first, they're using mouse embryos for now.

We've been hindered by what is basically a cult ideology about unborn life that we cannot do experiments like this (legally) in the west.

The fact that this experiment was done in China rather than "the West" has nothing to do with religion. The application of the CRISPR-Cas system for genetic modification was only discovered in 2012, and molecular analyses and proof-of-concept experiments - performed in the US and Europe, mostly - are being published in high-profile journals almost every month. There are, at last count, at least three companies (two in the US, one in Europe) founded by the scientists who elucidated the mechanism that have the explicit goal of human gene therapy. In fact, one group (in the US) just demonstrated in vivo genome editing (in an animal model, because only a lunatic would try this experiment in humans first).

There is no legal barrier to performing these experiments on human embryos in the US or Europe. In the US, I believe researchers are still prohibited from using NIH funding for such experiments, but that would not stop them from using private funding (and at this point, VCs and private donors are practically flinging sacks of money at this system). Their hesitation is based on concerns about the ethics of potentially lethal experimentation on unwilling test subjects. No, not the embryos, but the hypothetical live births that would result from implantation. If they're really, really lucky, the off-target effects will be silent or embryonic lethal. If they're unlucky - and given how new the system is, it's very difficult to guess what would happen - they'll wind up creating new genetic afflictions. Everyone working on the system is very excited about the potential applications to human health; no one wants to bring the field to a premature halt by rushing into human experimentation and accidentally causing severe birth defects because they didn't understand how it worked well enough.

There is a secondary issue, which is that China is almost pathetically desperate to prove it can do the same caliber of science as the West, to the extent that it's starting to throw money at non-Chinese researchers to set up labs in China, and offers large bounties for high-profile publications. (They're also known to be desperate for a Nobel prize in the sciences.) So far they've tended to just cherry-pick relatively easy, unimaginative projects following up on research done in the West (to be fair, Western scientists have done this among themselves for decades), rather than making entirely novel discoveries. Thus there is an enormous financial (and social) incentive to jump into a fast-moving field and try the obvious - but ethically dubious - application to human health.

If only it were that simple. Unfortunately, the classroom full of unvaccinated children may contain one of the few unlucky ones who have legitimate medical reasons for not being vaccinated. The fact that there are a small fraction of people like this, dependent on herd immunity for their protection, is one of the reasons for compulsory vaccination.

Rachel Carson and DDT ? http://www.21stcenturysciencet... [21stcentur...cetech.com]

If you want to be taken seriously, here's a tip: don't post links to publications from Lyndon LaRouche's organization in support of your arguments.

When your government is full of engineers, not lawyers, and when you can just ignore the flat-earth lobby instead of wasting half your funding fighting their just-because-we-can delays, you can test ideas like this.

Also useful: when your government is full of unelected bureaucrats who aren't held accountable by voters, completely dominate the news media, and stomp on any popular organization or sentiment that they don't control, and thus are free to ignore the interests of their citizens and instead spend money on wasteful, thinly-disguised military projects.

(Except, of course, that's not what's actually happening in this case - the article summary makes it sound like "OMG China will dominate space", because of course that's more interesting than "superannuated Chinese scientist spouts nonsense".)

reasonably lucrative work

I could understand Fiorina just wanting the attention - but dear God, I hope she doesn't need the money. That $20 million severance package would last most ordinary people a lifetime.

I think her target market is Republicans who want a viable female challenger to Hilary. Realistically, she's setting herself up for Sec. of Commerce, or maybe, if she's extremely lucky and does moderately well in the primaries, VP. I am no fan of hers for all of the obvious reasons, but she is a rocket scientist compared to Bachmann and Palin.

Then why is it exclusively conservatives who use the incorrect term?

These people want to put a stop to progress because they think humans are some kind of holy ground that must not be tred upon.

Actually, those people are involved in (at current count) at least two companies that have targeted therapeutic modification of humans as their primary business goal.

Bullshit:

research is needed to understand and manage risks arising from the use of the CRISPR-Cas9 technology. Considerations include the possibility of off-target alterations, as well as on-target events that have unintended consequences. It is critical to implement appropriate and standardized benchmarking methods to determine the frequency of off-target effects and to assess the physiology of cells and tissues that have undergone genome editing. At present, the potential safety and efficacy issues arising from the use of this technology must be thoroughly investigated and understood before any attempts at human engineering are sanctioned, if ever, for clinical testing. As with any therapeutic strategy, higher risks can be tolerated when the reward of success is high, but such risks also demand higher confidence in their likely efficacy.

I will continue to do work in this area and continue to work to improve humanity

Considering how little thought you've given to the potential downsides of such experiments, I'd guess that it's considerably more likely that you'll fuck up and produce a bunch of horribly malformed fetuses and live humans with fatal genetic problems. Fortunately, the ensuing lawsuits should put you out of business quickly.

I also have not ethical problems with genetic engineering on humans.

Do you have an ethical problem with genetically engineering an embryo and accidentally creating new problems that result in an individual crippled from birth, or doomed to a short and miserable life span?