…and this is a 
The first few talks this morning focused primarily on policy as illuminated by science; only the third talk was pure science.
Chiechanover's talk was on both the history and future of drug research, which he characterized in terms of three major revolutions in the last century.
The first revolution was a period of accidental discoveries in 1930s-1960s, where the discovery of a useful drug comes first, by observation of therapeutic effects, followed by chemical isolation, and only at the end (if at all), is the mechanism of action worked out. He gave the example of aspirin. Willow tree bark used for pain relief since at least Aristotle, but the active agent (salicylin) was only isolated in the 19th century by Buchner, and it was initially useless medically: it was water-insoluble and extremely bitter. Gerhard acetylated it to make it soluble near the end of the 19th century (but didn't take advantage of it as a clinical tool), and Hoffman (who made Bayer rich) repeated the acetylation and turned it into a useful drug, using it to treat his father, who was sick with arthritis. It's mechanism, as an inhibitor of prostaglandin synthesis, was not discovered until the late 20th century. The drug also prevents platelet aggregation, so is also being used in heart disease prevention, and its anti-inflammatory action may also make it a preventative for some cancers. However, it is a story of complete serendipity.
Another example of fortuitous discovery was Fleming's penicillin, which was a major factor in nearly doubling human lifespan in about a century, and antibiotics in general opened up the potential for all kinds of life-saving procedures, such as surgery.
The Second revolution occurred in the 1970s-2000s, and was planned. The key innovation here is high throughput, brute force screening of large libraries of chemical compounds, which he compared to "fishing in a swamp". We have no idea what we'll find, but there is the expectation that some compound will be found that will have a useful effect. It is a procedure that still relies serendipity, we've merely elevated the chances of finding something, and of exploiting it rapidly.
The example given was the work of Akira Endo, who knew that fungi were resistant to parasites, presumably because they contained agents to suppress bacterial cell wall synthesis by inhibiting cholesterol production. His work led to the discovery of statins, which has become a $20 billion/year industry for reducing to cholesterol levels in patients with heart disease. It has also been found to reduce the probability of heart attacks in patients who only have susceptibility for heart disease, and is now being used as a preventative in healthy patients (which is always a great way to vastly increase profits). It may also help with Alzheimers and malignancies, by mechanisms not currently known.
The third revolution is ongoing. The new strategy is understanding the mechanism first, followed by targeted design. He illustrated the problem with current pharmaceuticals by pointing out that men with prostate cancer and women with breast cancer are treated with the same tools: imaging technology, histology, and chemotherapy. These are different diseases! At the same time, two women may be diagnosed with breast cancer, but one will be estrogen sensitive and the other will be estrogen insensitive, which means that an effective treatment for one may be a lethal waste of time for the other. We aren't treating the disease specifically, but are using a one-treatment-fits-all formula for for general disease. What we need is a molecular diagnosis of tumors to fit treatment plans individually.
He thinks we are entering an era of personalized medicine. One example he gave was herceptin, which is an antibody targeted for the EGF receptor. People with a mutant, constitutively active EGF receptor are susceptible to certain kinds of cancers, so this is a very useful drug for down-regulating EGF activity. But for people with wild-type receptors, it's completely useless. The utility of this drug relies on diagnosis by PCR of specific alleles to find candidates for drug use.
He sees great promise in cheaper whole genome sequencing as an important tool for personalized medicine, and is looking forward to the days of the thousand-dollar genome. He also advocates a systems approach: interdisciplinary action will be needed to put together useful solutions.
The problems he foresees with personalized, targeted therapies are:
Multigenic deseases. Most of these diseases and susceptibilities aren't going to be the product of single alleles, but of multiple, interacting genes. That means answers won't be simple, but will require an understanding of combinatorial effects.
Malignancies are typically the product of genomic instability. They are moving targets.
Complications of human experimentation. We can't just pin down patients and run them through a series of carefully controlled trials, so working out the effective details of personalized medicine are going to be hard.
Lack of good animal models. A mouse is not a human. We can't do the necessary experiments on people, but at the same time we can't entirely trust the results of animal experiments.
Costs and legal liabilities. Medicine is done for profit. How do you pay for tools that work on tiny percentages of the population?
Bioethical problems. Information has repercussions. How do individuals cope with the knowledge that they might have, for instance, elevated susceptibility to breast cancer? How will it affect their relationships with family and spouses (or prospective mates)?
This was a good talk, but very, very general. I'm hoping we get some more scientific meat in other talks.
By the way, the way the meetings are run at Lindau is a little different than I'm used to — there is no Q&A afterwards! However, what they do instead is schedule small group meetings with the Nobelist speakers and the "Young Investigator" group of attendees, to which people here as press (like me!) are not invited, which is too bad from my point of view, but is probably a very good way to give people with more direct interests good access to the speaker.
Medicine & Health
Comments
Posted by: Mozglubov | June 30, 2009 6:15 AM
That sounds like an interesting talk, although I think the issue of cost is a very important one (as Steve Novella (I think it's him, at least, who I remember saying this) often points out, we currently know how to do a lot more health care than we can afford).
Posted by: Phodopus
|
June 30, 2009 6:16 AM
I am jealous, the Lindau meetings are quite some fun! Though the lack of Q&A is certainly not the only peculiarity there :). I found that talking to other participants was at least as interesting as the discussions with laureates themselves who are sometimes hard to access as they tend to be surrounded by crowds.
Posted by: Chris Davis
|
June 30, 2009 6:21 AM
Hmm. When do we get to the revolution where the discovery of a disease mechanism leads directly to a set of computer-generated molecules being produced specifically to bind to the sites and substances involved.
Seems to me - in my abysmal ignorance - that being able to predict accurately the shape and properties of proteins and other long biological chains will allow us to start using Nature's own lego blocks as the FSM intended.
Posted by: Jason | June 30, 2009 6:29 AM
Wow, thank you for writing this up, PZed. This sounds like a fantastic conference. Even though you say it wasn't detailed enough for your tastes, I found the information covered here very interesting.
I look forward to reading about the rest of the talks you are attending!
Posted by: africangenesis | June 30, 2009 6:30 AM
Personalized medicine also offers opportunities to save costs by avoiding treatments that will be ineffective or have undesirable side effects. The pharmaceutical industry is often criticised for producing too many "me too" drugs, but I actually appreciate having a range of side effect profiles and metabolic pathways to select from. Unfortunately centralized formularies often eliminate or penalize these options.
Did Chiechanover have anything interesting to say about molecular docking software and computer rather than experimental screening of molecular databases?
Posted by: apostrophe nazi | June 30, 2009 6:41 AM
PZ: "Its mechanism, as an inhibitor of prostaglandin synthesis..."
Posted by: David Marjanović, OM | June 30, 2009 8:42 AM
LOL. Was it televised, too? ;-)
Insane.
True.
Forget about it. He'll never learn that. Regard it as a personal quirk like ERV's war on apostrophes in general.
Posted by: James Sweet | June 30, 2009 9:16 AM
Be very careful with phraseology here. While I more or less agree with that statement, and think that personalized medicine has a lot of potential (though it's a minimum of ten years off, probably more than that, due to the difficulties that were listed), the "one-size-fits-all" is a rallying cry of alties and woomeisters who want to hawk alternative or unproven treatments, or sucker people into delaying or forgoing vaccination.
It's like the old argument that the existence of God is a 50/50 shot because we can't prove it either way. Exploiting a similar fallacy, alties correctly assert that the standard treatments offered in conventional medicine are not always the best solution for an individual patient, and then use this to promote the idea that no treatment is inherently better than any other treatment.
I don't disagree with the talk, I'm just saying, I hope the guy is careful about phraseology. Alties quote mine at least as much as Creationists, and I'd hate to see an anti-vax website saying something like, "Nobel laureate Aaron Ciechanover has said that one-treatment-fits-all conventional medicine is a fundamentally flawed!"
Posted by: Kausik Datta | June 30, 2009 9:22 AM
Gasp! He did not mention the fourth revolution? One that undermines everything his first three have achieved or stand for? The unprecedented rise of pseudoscientific, evidence-lacking, misleading and potentially dangerous, complimentary and alternative medicine?
[/snark]
Posted by: Kausik Datta | June 30, 2009 9:26 AM
Gaah! Make that 'Complementary'
Must.Have.Morning.Coffee.
Posted by: Paul Browne | June 30, 2009 10:20 AM
Good point James (#8), I've come across such claims many times in altie commentary, the "flipping a coin" fallacy is particularly strong amongst animal rights supporters.
As to Chris Davis (#6), I'd give it about 40 years.
I'm not sure about the use of the word "revolution", since that implies that the previous methodologies were swept away when the new one came along, whereas in reality these revolutions have overlapped to a large degree...is the first revolution even over?
I think this is important because it affects how we think of personalized medicine is that it doesn't all have to happen at once, even as we discover the knowledge and develop the technologies necessary to develop truely personalized medicine we can improve medicine by incorporating aspects of it.
As for pre-clinical evaluation of personalized medicine, there are some very interesting developments in GM animals, 3D tissue culture and stem cell research (especially iPS) that will help fill the gap over the coming years.
I have a big problem with the sentence "He illustrated the problem with current pharmaceuticals by pointing out that men with prostate cancer and women with breast cancer are treated with the same tools: imaging technology, histology, and chemotherapy. These are different diseases!"
Imaging technology and histology are diagnostic tools used to identify to identify the type of cancer, while there are many different types of chemotherapy available for different cancers with the choice of drug(s) being dependent on the results of the earlier diagnosis. While molecular diagnosis will complement these more established tools they are unlikely to replace them. To imply that using these same tools in different cancers is somehow a "one size fits all" approach is getting to close for comfort to the kind of guff you'd expect from Mike Adams.
Posted by: Paul S | June 30, 2009 11:17 AM
To Chris (#3):
That revolution has been happening for 20-25 years and is known as computer-aided drug discovery (CADD) or the more generic molecular modeling. The problem turns out to be far more complex than anyone thought. Atomic-level structures are known and publically available for roughly 25000 unique proteins (based on a 95% identity cutoff), see the RCSB for more info. However, there are few serious shortcomings.
1) There are only a handful of integral membrane proteins with known structures. (Background: Most of the structures are solved by first crystallizing the proteins, which borders on an art. These crystals are then bombarded by X-rays and the diffraction of said X-rays recorded. Through some mathematics, these patterns of dots, think of the famous DNA diffraction patterns with many more spots, are converted into 3-D atomic level structures. Membrane proteins are exceedingly difficult to crystallize.) Many such proteins are key drug targets such as GPCRs. These are key players in cell signaling. GPCR work is generally done based on using the 3(?) current structures as bases for modeling the one of interest. This introduces its own errors.
2) Solved protein structures are a static model of the protein. In real life, proteins are dynamic, constantly undergoing movements of various types. There has been a lot of work in this area, but it still remains a difficult problem. Also, many proteins will undergo a dramatic shift in their shape upon binding of certain compounds (known as induced fit). In fact, this is the way GPCRs transmit extracellular signals into the cell. The prediction of such movements has met with very limited success. In some cases, a protein that has been well studied with multiple structures with multiple ligands was found in a very different conformation when one particular complex structure was solved.
3) One key component of making a new drug is that is has to get into the body and not kill the patient. Accurate prediction of such properties (ADMET: Absorption, Distribution, Metabolism, Excretion, and Toxicity) is a huge challenge and is a significant line of on going research. Even the development of cell-based assays for prediction is problematic. While such predictive systems exists, surprises still occur once the compound is given to whole organisms. (This is one reason why animal trials are vitally important.)
Posted by: J_w23 | June 30, 2009 12:43 PM
A supervisor of one of my undergraduate internships was prof drs Gert Vriend. He's working for many years now on protein modeling software to predict the structure of potential drugs. He's quite far right now and had it integrated into a program called Yasara (designed by one of his Phd. students). It's a brilliant software package and got to work with it very often. But actually predicting the ligand structure of a possible drug is proven to be extremely difficult; mainly because the protein structures needed for modeling are mostly available in crystal structure only. That means devoid of water.
Posted by: antistokes
|
June 30, 2009 12:54 PM
One more point about crystal structures that my phd adviser kept hammering in at lab meetings: there is also selection for proteins that are able to be crystallized. Working in a TB drug discovery lab, you rapidly learn that there are a loooot of proteins (NOT just membrane proteins, although those are really troublesome) that are very, very difficult to even, for example, get to the concentrations required for crystallization-- they just from inclusion bodies in the cells you're employing for overproduction and degrade. The protein database represents only those special proteins that we can make in extremely high concentrations. (Not that it's not useful!- it's our best shot at many structure based drug designs at the moment-- just, it is selective.)
Posted by: africangenesis | June 30, 2009 12:56 PM
Intense X-ray sources are being constructed to attempt infering structure from a single molecule. I wonder if they will be able to handle a little water?
Subscription may be required:
http://www.nature.com/nmeth/journal/v6/n1/full/nmeth0109-8a.html
Posted by: Paul S | June 30, 2009 1:03 PM
Actually, that's not really true. Protein crystals are generally 30-70% water. (I think I've seen some references to some being upwards of 80%.) We only see the water molecules that are quite stable in the crystal structures (referred to as crystalographic waters), but often times there are several hundred of these for a modest size protein. The real issue with waters is that they often move around during binding and predicting this, along with which remain when ligand is bound and which are displaced is non-trivial. (Sorry to harp on the waters, but a chunk of my graduate work looked at modeling of protein-water interactions.) In some cases, e.g., HIV protease, a particular water can play a key role in understanding the protein-drug interactions.
Posted by: antistokes
|
June 30, 2009 1:08 PM
heh how much simulation time did that take up? :)
(my SO does computational simulations on viral capsid assembly and he's always bitching about the waters).
Posted by: Pikemann Urge | June 30, 2009 4:54 PM
Drug research is not science - it's more than that. It's technology. Get your disciplines right, dagnabbit.
Posted by: J-w23 | June 30, 2009 7:31 PM
@ Paul S #16
You're right, I just might have simplified too much. The dynamic environment a protein resides in should not be underestimated. Simulating the dynamics of water molecules surrounding any kind of protein are just extremely complicated; but are most often essential in understanding protein function.
Posted by: Heraclides | June 30, 2009 8:01 PM
If I was using aspirin as an example, I'd also point out that the crystal structure of the protein-drug complex, and hence the opportunity to design alternative drugs wasn't done until relatively recently. It's relevant as crystal structures can be used to build potential inhibitors, etc., which can then be screened. This attempts to focus the screening on likely compounds, rather than the "brute force" approach of mass screening. This approach is one of the "modern" approaches, so it's another example of how aspirin was "discovered" in the "opposite" to one approach used today. This approach in some ways lies between your second and third approaches.
(Reading the comments, I see #12 has also picked up on this. Sorry for the rehash, but at least I'll spare repeating the issues with it, as these have already been touched on. It's not quite as negative as Chris makes out, there are approaches to mitigate these problems to various extents too.)
Posted by: DLC | June 30, 2009 9:01 PM
I wish the computing power was there for full on human modeling for drug testing, but it just isn't doable at present. Of course, I'd love to be proved wrong about it.
Posted by: Orac
|
July 1, 2009 8:04 AM
Actually, the quest for "personalized" medicine is nothing that new, as has been pointed out. Computer-aided drug design and "targeted" therapies have been in development for decades. In my field, the first big targeted therapy was Herceptin. In any case, lately the Next Big Thing has been systems biology. It holds a lot of promise, but its advocates often sell it with pie in the sky sorts of claims, mainly because most systems biologists have no idea how therapies and drugs are actually validated and tested:
http://scienceblogs.com/insolence/2007/03/the_individualization_of_medical_treatme_1.php
Of course, another problem, as has been pointed out, is that personalized medicine will be a lot more difficult to demonstrate efficacy, mainly because by definition fewer patients will be getting drugs "targeted" to their biology and genetics. This fragments the number of patients who can benefit from any given therapy and makes clinical trials a nightmare.