About this Author
DBL%20Hendrix%20small.png College chemistry, 1983

Derek Lowe The 2002 Model

Dbl%20new%20portrait%20B%26W.png After 10 years of blogging. . .

Derek Lowe, an Arkansan by birth, got his BA from Hendrix College and his PhD in organic chemistry from Duke before spending time in Germany on a Humboldt Fellowship on his post-doc. He's worked for several major pharmaceutical companies since 1989 on drug discovery projects against schizophrenia, Alzheimer's, diabetes, osteoporosis and other diseases. To contact Derek email him directly: Twitter: Dereklowe

Chemistry and Drug Data: Drugbank
Chempedia Lab
Synthetic Pages
Organic Chemistry Portal
Not Voodoo

Chemistry and Pharma Blogs:
Org Prep Daily
The Haystack
A New Merck, Reviewed
Liberal Arts Chemistry
Electron Pusher
All Things Metathesis
C&E News Blogs
Chemiotics II
Chemical Space
Noel O'Blog
In Vivo Blog
Terra Sigilatta
BBSRC/Douglas Kell
Realizations in Biostatistics
ChemSpider Blog
Organic Chem - Education & Industry
Pharma Strategy Blog
No Name No Slogan
Practical Fragments
The Curious Wavefunction
Natural Product Man
Fragment Literature
Chemistry World Blog
Synthetic Nature
Chemistry Blog
Synthesizing Ideas
Eye on FDA
Chemical Forums
Symyx Blog
Sceptical Chymist
Lamentations on Chemistry
Computational Organic Chemistry
Mining Drugs
Henry Rzepa

Science Blogs and News:
Bad Science
The Loom
Uncertain Principles
Fierce Biotech
Blogs for Industry
Omics! Omics!
Young Female Scientist
Notional Slurry
Nobel Intent
SciTech Daily
Science Blog
Gene Expression (I)
Gene Expression (II)
Adventures in Ethics and Science
Transterrestrial Musings
Slashdot Science
Cosmic Variance
Biology News Net

Medical Blogs
DB's Medical Rants
Science-Based Medicine
Respectful Insolence
Diabetes Mine

Economics and Business
Marginal Revolution
The Volokh Conspiracy
Knowledge Problem

Politics / Current Events
Virginia Postrel
Belmont Club
Mickey Kaus

Belles Lettres
Uncouth Reflections
Arts and Letters Daily
In the Pipeline: Don't miss Derek Lowe's excellent commentary on drug discovery and the pharma industry in general at In the Pipeline

In the Pipeline

« "The Smarter You Are, The More I Will Hate You" | Main | What Layoffs Have Done »

September 26, 2011

Predicting Toxicology On A Chip?

Email This Entry

Posted by Derek

Predicting toxic drug effects in humans - now, that's something we could use more of. Plenty of otherwise viable clinical candidates go down because of unexpected tox, sometimes in terribly expensive and time-wasting ways. But predictive toxicology has proven extremely hard to realize, and it's not hard to see why: there must be a million things that can go wrong, and how many of them have we even heard of? And of the ones we have some clue about, how many of them do we have tests for?

According to Science
, the folks at DARPA are soliciting proposals for another crack at the idea. The plan is to grow a variety of human cell lines in small, three-dimensional cultures, all on the same chip or platform, and test drug candidates across them. Here are the details. In keeping with many other DARPA initiatives, the goals are rather ambitious:

DARPA is soliciting innovative research proposals to develop an in vitro platform of engineered tissue constructs that reproduces the interactions that drugs or vaccines have with human physiological systems. The tissue constructs must be of human origin and engineered in such a way as to reproduce the functions of specific organs and physiological systems. All of the following physiological systems must be functionally represented on the platform by the end of the program: circulatory, endocrine, gastrointestinal, immune, integumentary, musculoskeletal, nervous, reproductive, respiratory, and urinary.

The request goes on to specify that these cell cultures need to be able to interact with each other in a physiologically relevant manner, that distribution and membrane barrier effects should be taken into account and reproduced as much as possible, and that the goal is to have a system that can run for up to four weeks during a given test. And they're asking for the right kinds of validation:

Proposers should present a detailed plan for validating integrated platform performance. At the end of each period of performance, performers are expected to estimate the efficacy, toxicity, and pharmacokinetics of one or more drugs/vaccines that have already been administered to humans. Proposers should choose test compounds from each of the four categories listed below based on published clinical studies. These choices should also be relevant to the physiological systems resident on the platform at the time of testing and should include at least one test compound that was thought to be safe on the basis of preclinical testing but later found to be toxic in humans.
i. Drugs/vaccines known to be safe and effective
ii. Drugs/vaccines known to be safe and ineffective
iii. Drugs/vaccines known to be unsafe, but effective
iv. Drugs/vaccines known to be unsafe and ineffective

Now, that project is going to keep some people off the streets and out of trouble, for sure. It's a serious engineering challenge, right off the bat, and there are a lot of very tricky questions to get past even once you've got those issue worked out. One of the biggest is which cells to use. You can't just say "Well, some kidney cells, sure, and some liver, yeah, can't do without those, and then some. . ." That's not how it works. Primary cells from tissue can just die off on you when you try to culture them like this, and if they survive, they (almost invariably) lose many of the features that made them special in their native environment. Immortalized cell lines are a lot more robust, but they've been altered a lot more, too, and can't really be taken as representative of real tissue, either. One possibility that's gotten a lot of attention is the use of induced stem cell lines, and I'd bet that a lot of the DARPA proposals will be in this area.

So, let's stipulate that it's possible - that's not a small assumption, but it's not completely out of the question. How large a test set would be appropriate before anyone puts such a system to serious use? Honestly, I'd recommend pretty much the entire pharmacopeia. Why not? Putting in things that are known to be trouble is a key step, but it's just as crucial that we know the tendency of such an assay to kill compounds that should actually get through. Given our failure rates, we don't need to lose any more drug candidates without a good reason.

We're not going to have to worry about that for a while, though. DARPA is asking for people to submit proposals for up to five years of funding, contingent on milestones, and I still cannot imagine that anyone will be able to get the whole thing working in that short a period. And I think that there's still no way that any system like this will catch everything, of course (and no one seems to be promising that, fortunately). A system sufficient to do that would be like building your own in vitro human, which is a bit out of our reach. No, I'd definitely settle for just an improved look into possible tox problems - every little bit will definitely help - but only if it doesn't set off too many false alarms in the process.

Comments (24) + TrackBacks (0) | Category: Toxicology


1. Jose on September 26, 2011 8:27 AM writes...

I want Santa to bring me a pink pony, too!

Permalink to Comment

2. johnnyboy on September 26, 2011 9:32 AM writes...

oh lawl. I hope those nutjobs at DARPA are very, very, very patient. What they're asking is essentially to create a living , multi-organ organism in a dish. Which, btw, you can't, and probably won't be able to for the next, oh I don't know, 100 years ? Or maybe 1000 ? And not only must you be able to manage that, but that mini-frankenstein must also be predictive of drug and vaccine toxicity, and to be useable in an industrial context... sure, no problem mate. Hmm, how many trillions did you say you had to support this research ?

Look, we're already using real actual live multi-organ organisms for toxicity testing, they are called rodents and dogs. And with 50 years of experience and billions of rats and mice used, we still don't have a good hold of their predictive ability, although we have a general feel that they're not bad, and generally useful. I understand the will to get better predictivity, but what DARPA seems to think, that just using human cells will solve everything, is just ludicrous. And don't tell me 'well they helped invent the internet, so why not this ?' For all the wonders of microchips and phone lines, the complexity of what they're asking here is several dozen orders of magnitude than making email possible.

Permalink to Comment

3. milkshake on September 26, 2011 10:13 AM writes...

of course this is only a cover story. What DARPA is really after is grafting engineered human tissue onto a superalloy endoskeleton

Permalink to Comment

4. John Schilling on September 26, 2011 10:13 AM writes...

DARPA does like to tackle the hard problems, don't they? I like the general idea, but "cell cultures need to be able to interact with each other in a physiologically relevant manner" strikes me as far out of reach for a first-generation system.

Still, if all you get is a quick screen against many cell types in isolation, that would I think be quite useful. In fact, I think I'd rather have the best possible data on drug effects vs single-cell cultures in isolation than a half-assed attempt at measuring more complex interactions.

And I might rather have the Mark I version use mouse, rat, or dog cells than human. We aren't going to be approving drugs for human use based on this DARPA gizmo for quite some time; there are going to be animal trials first. So, build a device that screens out drugs that never would have had a chance in the animals. And validate it by showing how precisely it does (or does not) predict the results of animal testing.

Permalink to Comment

5. Pete on September 26, 2011 10:23 AM writes...

My understanding is that it's difficult enough to just maintain many (most?) human cell types in culture. It would seem prudent to solve those problems before trying to culture human cells in physiologically relevant assemblies.

When looking at some toxicities (e.g. cardiotoxicity resulting form ion channel blockade) one will need to work very hard to persuade individual cells to organise themselves into a physiologially relevant assembly.

Permalink to Comment

6. RespiSci on September 26, 2011 10:29 AM writes...

If such an ambitious endeavor were to be successful, not only would be weeding out of bad actors earlier in the development process, but also we could identify actors that aren’t toxic in the human system although there may be findings in one or both animal species from the preclinical toxicology studies arguing for continued development of such candidates. So among the list of drug candidates to test, besides “include at least one test compound that was thought to be safe on the basis of preclinical testing but later found to be toxic in humans.� may also want to include some drugs that had some animal toxicology issues which weren’t manifested in clinical studies. (Question: how public is this knowledge?)

“At the end of each period of performance, performers are expected to estimate the efficacy, toxicity, and pharmacokinetics of one or more drugs/vaccines that have already been administered to humans.� Whoa…....assessing toxicology is going to be challenging enough, but is this request to assess efficacy on the chip too? Besides creating a healthy system on the chip to monitor toxicology is there also a need to replicate the disease state (cancer/infection etc) in the chip system to show that in fact the drug/vaccine is working? Truly an ambitious undertaking.

Permalink to Comment

7. simpl on September 26, 2011 10:36 AM writes...

One of the best chemical production interactions I wver saw with research guys (ours and an academic group) was an inventory of missing safety data for standard chemicals and intermediates. The toolkits around were good enough to set priorities, i.e. predict when the risks were low, and when an effect was worth testing for. Some results will become visible for all in the REACH database, latest 2018.

Permalink to Comment

8. wwjd on September 26, 2011 10:52 AM writes...

Is this to help develop/detect chemical weapons?

Permalink to Comment

9. johnnyboy on September 26, 2011 11:17 AM writes...

@4: there already are cell-based safety assays in common usage. For example, the hERG assay is human-cell based, and is commonly used as a pre-IND safety evaluation. Other cell-based assays are used for other aspects, eg micronucleus test for genotoxicity, etc... Other assays are getting developed for specific issue, eg direct toxicity to hepatocytes. But these are all very targeted, simplified, and predictive of only one very focused toxic effect (and even with such a relatively simple models, their predictivity for the clinical setting can still be somewhat in doubt). What DARPA is asking here is completely different, it is reconstructing organ-like arrays of multiple cells per tissue, with all the different organoid interconnected and interacting together. Which is why it's totally pie-in-the-sky stuff.

Permalink to Comment

10. drsnowboard on September 26, 2011 11:26 AM writes...

Wouldn't it be easier just to feed drugs to death row inmates? Except obviously the ethnic spread might be a little off..

(NB bad taste comment yes, serious , no)

Permalink to Comment

11. pete on September 26, 2011 12:34 PM writes...

@9 johnny

"pie-in-the-sky" -- Yep, if the criterion for acceptance is a multicellular mass that faithfully replicates the physiology of organ X.

But at least I give them credit for holding out a carrot to folks developing sophisticated cell culture systems, and for acknowledging that these systems can elicit behaviors that transcend what we've seen in planar cell culture. Safe to say that the art & science of sophisticated cell culture has been evolved a good deal in the past couple of decades so I'd like to believe that maybe something positive, Tox Pred-wise, can be accomplished within the relatively short time frame.

My main concern would be the economics of doing such Tox screens. They may easily break the bank allotted for any given compound.

Permalink to Comment

12. David Formerly Known as a Chemist on September 26, 2011 12:34 PM writes...

This is actually the big idea, reach for the sky type of projects that I like to see government funding. Better than the little esoteric R01 grants that feed narrow-minded faculty and trains the next generation of Indo-China scientists. Look what can be accomplished with grand vision: man on the moon, Manhattan Project, etc. This would clearly be a grand leap forward.

Permalink to Comment

13. CMCguy on September 26, 2011 1:45 PM writes...

This is an interesting and potentially useful concept which hope does result is supportive technology yet wonder how would be in practice. What level of False Positives and False Negatives are likely to be encountered and the bigger questions are how well will these cell based models actually be validatable? Based on (currently limited?) "in silico Tox models" I have seen compounds rejected on of hand even with animal studies suggesting no major issue or conversely other compounds being developed while ignoring a negative prediction from a model. Drug tox is often linked to ADME properties and frequently impact multiple biologic systems so the greatest challenge may be to make "relevant to the physiological systems". I agree with Derek that to best prove value would have to test against the entire pharmacopeia plus would add as many compounds that "died" in clinical development as possible and even then would worry about the "exceptions" both ways. Great if it works out but that is indeed is a big if.

Permalink to Comment

14. And D on September 26, 2011 2:23 PM writes...

One problem with the DARPA proposal is the relative complexity of the predictive tools proposed - "an array of engineered tissue contructs", when weighed with the degree of success such an approach is likely to have.

Not sure that a toolbox of tissues would yield anything that would be that much better than attempting to predict "toxicology risk" from molecular descriptors/simple measurements, which is an approach already advocated to some extent.

OK, you will never totally be sure if your highly liophillic, high dose, high molecular weight...etc...etc. potential candidate will prove to be unacceptably toxic, but with a big enough dataset, we could perhaps do a better job of weighing up relative risks.

Permalink to Comment

15. Paul on September 26, 2011 3:06 PM writes...

DARPA is intended to do avant garde research of this kind. As I understand it, they are happy if 1% of their projects succeed. Those successes, and spin offs, pay for everything else.

Permalink to Comment

16. hn on September 26, 2011 3:32 PM writes...

DARPA gives long term goals and is not expecting results in an R01 time frame. If you want ambitious, check out their 100 year starship project.

Permalink to Comment

17. Pete on September 26, 2011 5:52 PM writes...

The hERG assay is, as johnnyboy has noted, human cell-based. However, it measures hERG blockade but not toxicity. If you want to observe this type of toxicity you will need to use an animal model. The links between in vitro and in vivo assays for toxicity are not always as transparent as one would like. There is a tendency to see toxicity purely in terms of affinity and to ignore free (unbound) concentration, a point which I’ve discussed this in my most recent blog post entitled ‘Dans la merde’).

Permalink to Comment

18. BCP on September 26, 2011 8:00 PM writes...

I applaud DARPA for trying to encourage further innovation here. The main issue I have with this is that the more troubling kinds of human toxicity that are cripplingly expensive for drug developers to deal with and a public health hazard in the real world are those rare, idiosyncratic events that get picked up in post market meta-analyses or large P3 trials with hundreds or thousands of patients. By definition, if you have a signal that shows up in 1 or 2% of the population, then 98-99% of people who receive your drug may not give you any signal that you have a problem. All of those data are from intact, living, breathing, multi organ systems too.

This seems to be the complete opposite of the concept of predicting efficacy with in vitro system, where you are starting from an expectation that everyone who receives your drug *should* be able to benefit. Obviously hERG inhibition is a good counter example, but it seems rare that we ever understand the molecular basis for human toxicities.

Permalink to Comment

19. JK on September 26, 2011 8:57 PM writes...

While very ambitious, I don't think that progress in this direction is as pie-in-the-sky as some commenters seem to think. See e.g.

which looks to me encouraging.

In some ways it won't be as physiological as animal testing, but the up side is the possibility of using human cells and higher throughput.

Permalink to Comment

20. TX Raven on September 27, 2011 1:24 AM writes...

How come we didn't have all this crap 30 years ago and we discovered safe drugs?
Why should the "low hanging fruits" be safer than higher hanging ones?
Or is it that our reduced risk tolerance leads us to expect "perfect" drugs that do no harm to any living species, including cell cultures?
Now, that is a new concept!

Permalink to Comment

21. johnnyboy on September 27, 2011 8:01 AM writes...

re #12 et al, "This is actually the big idea, reach for the sky type of projects that I like to see government funding."

Personally, what I like to see government fund is its own existing institutions, to a level where they can actually function. Is it normal that it takes one year for FDA to review a NDA ? Is it normal that in my field of medical devices, FDA hasn't been able to even produce final guidelines as to what they actually want for a safety package ? It's nice to be all starry-eyed, but it makes more sense to finish building your house's roof before you start construction on a new skyscraper.

Permalink to Comment

22. simpl to johhnyboy on September 27, 2011 10:40 AM writes...

Be thankful the FDA takes its time issuing guidelines in your area. They would tend to be overly prescriptive in a growing area. You could enjoy the ride, and your relative expertise. Or you could influence the FDA, by submitting what you think is a reasonable package.

Permalink to Comment

23. HenrikOlsen on September 28, 2011 3:30 PM writes...

Even if noone makes the goal, this is a project that's almost certain to spin off interesting results and new knowledge as people try to solve the problems involved.

Permalink to Comment

24. abm on October 20, 2011 8:16 AM writes...

While ambitious, the DARPA project is too complex to be achieved in 5 years. This is nothing like creating the Internet. It would takes 10-20 years and groups of teams working together on this technology. It would have been preferable if the project was disassembled into more pieces and had multiple groups working on this. It is as big as NIH funding the DNA sequencing program. Also, by offering multiple awards, how will standardization processes be adopted?

Permalink to Comment


Remember Me?


Email this entry to:

Your email address:

Message (optional):

The Last Post
The GSK Layoffs Continue, By Proxy
The Move is Nigh
Another Alzheimer's IPO
Cutbacks at C&E News
Sanofi Pays to Get Back Into Oncology
An Irresponsible Statement About Curing Cancer
Oliver Sacks on Turning Back to Chemistry