Awesome. The alternative biofuel part of the talk starts at 13 minutes. I highly recommend watching it all.
Comments
allianus
I understand totally random attempts at solutions won't work very well if the range of possibilities is virtually unlimited compared to how fast you can test them. No one said that's how Venter was generating his trials. In real life we use whatever means are at our disposal to pre-select likely candidates for trial. If you have no criteria whatsoever to reject/accept candidates then that could be a big problem.
jehu
It seemed to me that Venter kept stressing that what they were doing was design and not so-called “applied evolution.”
Absolutely. Only a moron would use a blind watchmaker when a sighted one could be employed instead. That should go without saying.
DaveScot
June 1, 2008
June
06
Jun
1
01
2008
05:20 AM
5
05
20
AM
PDT
Sad to say, beautiful theories (and even beautiful rhetoric) must come into the real world, at which point they find themselves confronted with stubborn facts. The problem with using random throughput screening for drug discovery is that it is random. It may produce perfectly useful compounds, but the drug company has no way of knowing if they are useful until it tests them; since there are dozens of potential therapeutic areas where they may be useful, the very randomness of the process makes it too costly for practical use. It was popular at one time but has now been abandoned. So it is with evolution. Random mutation cannot lead to any positive gains without a specific end in mind, no matter how many billions of years are merrily tacked on.allanius
June 1, 2008
June
06
Jun
1
01
2008
02:53 AM
2
02
53
AM
PDT
DaveScot,
That's right. It seemed to me that Venter kept stressing that what they were doing was design and not so-called "applied evolution." He mentioned that all life comes from life as well. So they are taking the lego blocks that life provides and rearranging them a bit.
There were a few companies in the nineties that sought as a business model to develop novel enzymes using high throughput screening and "applied evolution." Last time I checked these companies were no longer around and they failed to develop a single novel enzyme. They did succeed at optimizing enzymes of a known trait.
If we ever get a novel enzyme it will come through design based on the knowledge we have of how other related enzymes function. High throughput screening may allow for the testing of many different designs. If function is found, the designed enzyme can be optimized by artificial selection. But so far I don't think we have even be able to do that.
It sounds like Venter is taking the smart approach and looking for existing genes that perform the function he wants.Jehu
May 31, 2008
May
05
May
31
31
2008
11:50 AM
11
11
50
AM
PDT
Jehu
My understanding is that Venter isn't using randomly generated DNA. He's using genes found in nature. A large part of his research, remember, consisted of two circumnavigations of the globe sampling microbial life from the ocean depths and shotgun sequencing their DNA to assemble a gene library. I think he mentioned a catalog of 30 million genes in the video. Most of the genes are organized into closely related families and what he's doing is trying out different family members in an effort to optimize certain biological processes (like converting cellulose to sugar and sugar to ethanol). So he isn't doing a blind search of 10^2000 (or whatever) possible genes but rather a search through families of 10^2 or 10^3 biologically active genes.
The technology that enables all this is, as far as can tell, following the general rule of Moore's Law for semi-conductors - a doubling of cost/performance every 18 months. Therefore we can expect bio-engineering to improve at the same pace that computer engineering has done before it. A lot of this hinges off miniaturization. Tests that used to be done by the score in racks of test tubes with a lot of time required by a technician are now automated and happen by the thousands on a wafer the size of a dime. The technologic progression in all this is very much like what I experienced in the computer industry from 1975 to 2000. Faster, smaller, cheaper, better... with a doubling every 18 months.DaveScot
May 31, 2008
May
05
May
31
31
2008
10:20 AM
10
10
20
AM
PDT
DaveScot,
Well from what I understand, in practice the high throughput screening by itself hasn't yielded much results for the cost. It is more effective to determine what constraints to put on the throughputs as much as possible. My guess is that the reason is that the space of useless structures is so large compared to the space of useful structures.Jehu
May 31, 2008
May
05
May
31
31
2008
09:58 AM
9
09
58
AM
PDT
allianus
We're talking about higher throughput at the same cost - making the screening process faster and more efficient.DaveScot
May 31, 2008
May
05
May
31
31
2008
08:48 AM
8
08
48
AM
PDT
Increasing the throughput simply makes the process more diffuse. The problem is there's no way to know if a compound is useful till you use it. And there are a heck of a lot of potential therapeutic areas. Hence adding to the sheer volume of new compounds actually makes the drug discovery process less profitable. The random selection approach is so 90's. Today they combine the two approaches--they use the shotgun approach with a specific therapeutic end in mind (when they're not trying to piggback off existing compounds). But then the drug analogy is germane. Without specific goals in mind, evolution produces random, undifferentiated results. And evolution, unlike Craig Venter, doesn't have a mind.allanius
Jehu
re; high throughput drug screening
If it works, but not very well, then it can be improved by increasing the throughput.
This isn't rocket science. It's simple troubleshooting. You can find a solution by completely understanding the problem or by trying out different potential solutions if you don't completely understand the problem. Better understanding leads to finding solutions faster but sometimes better understanding takes so long that it's easier to just work on increasing the speed at which you can try out potential solutions. In practice the latter approach usually leads to better understanding. It's a matter of hypothesis testing - the faster you can formulate and test hypotheses the faster you can do science (or engineering, or really any kind of problem solving).
Of course some things can't be done this way. For instance we can't build better space shuttles by adding things to them and seeing if what we added makes them crash or fly better. But outside such constraints the shotgun method is a very fruitful and essential means of investigation.
DaveScot
May 31, 2008
May
05
May
31
31
2008
03:06 AM
3
03
06
AM
PDT
In the context of drug discovery, a review of high throughput screening in Nature a few years back made clear that simple mass screening of random compounds doesn't work very well. The article was clear that significant design work was required before the screening phase. "Therefore, the time spent on obtaining relevant information, rather than sheer capacity of synthesis and testing, will determine the success of the research programme." This seems to parallel many discussions we have had about trying to get a novel function by filtering stochastic inputs. The inputs have to be highly constrained to get a function.
http://www.cmbi.ru.nl/edu/bioinf4/articles/nrd1086_fs.pdfJehu
May 31, 2008
May
05
May
31
31
2008
02:09 AM
2
02
09
AM
PDT
Jehu is correct that it's not novel. It's called the shotgun approach and is a common method in all kinds of problem solving when there are unknowns to deal with - spray the problem with lots of possible solutions (like a shotgun blast) and see which works and/or which works the best. Even job hunters do it when they send out lots of resumes all at once. Jehu is incorrect that it's not fruitful.
DaveScot
May 31, 2008
May
05
May
31
31
2008
01:33 AM
1
01
33
AM
PDT
Basically, he’s relying (as he always has done) on fast technologies to create lots of data, which then has to be filtered and sorted to get something useful out of it. Selection is a powerful tool for doing that.
Well, that approach is not novel and has not really been fruitful. I think that Venter is attempting to design the data that they then optimize through selection.
I would agree that selection does play a role. I would point out that it is artificial selection and not natural selection. The concept of artificial selection pre-exists Darwin through the field of animal husbandry. Also, Darwin lays claim to the origin of species not just mere natural selection. Natural selection in itself is not Darwinism.Jehu
May 31, 2008
May
05
May
31
31
2008
12:49 AM
12
12
49
AM
PDT
Does anyone here think that evolutionary thinking will contribute to his work in any meanigful way?
Venter does. He mentioned selection - that's evidently how he's envisioning sorting through the large number of DNA sequences that he's planning on creating.
Basically, he's relying (as he always has done) on fast technologies to create lots of data, which then has to be filtered and sorted to get something useful out of it. Selection is a powerful tool for doing that.Bob O'H
May 30, 2008
May
05
May
30
30
2008
11:17 PM
11
11
17
PM
PDT
As well, I really have to ask, Does anyone here think that evolutionary thinking will contribute to his work in any meanigful way? As far as myself, I only see dead-ends and wasted time and money for any evolutionary inspired paths he may choose to take in his ventures, But as far as "Intelligent Design" is concerned, ID will give him a more realistic and fruitful path to follow since it will highlight what he likely can and cannot do with his cutting edge pursuits.
Many here may argue that evolutionary thinking is critical to his science but I do not see it being necessary at all.bornagain77
May 30, 2008
May
05
May
30
30
2008
06:23 PM
6
06
23
PM
PDT
Very exciting and frustrating. Exciting to see cutting edge science, frustrating that he, to put it mildly, glosses over the mounumental hurdles for "novel" species to appear in his many obligatory bows to king Darwin.
To highlight his problems;
The human genome is proven by ENCODE to most likely be 100% poly-functional, Thus it is an all or nothing venture for genome sequencing. He even admitted as much when he was talking of his gene knock-out work with genetalia (though very reservedly). And in his work transplanting a genome to a "new" cell. Though there may be a certain severely limited ability for the genome to accept new genes into its genome it is fairly clear to see that any insertion into a 100% poly-functional code will not allow what he envisions for flexibilty.
He has yet to demonstrate such flexibility of the genome, and while his work tinkering with what already exists is worthy of applause his dreams for a "new evolutionary future are unrealistic to what we currently know of genome constraints.bornagain77
May 30, 2008
May
05
May
30
30
2008
06:00 PM
6
06
00
PM
PDT
Venter: "This is building on about three and a half billion years of evolution."
This man goes into a trance when he makes this statement, as if he was supposed to state it. He rightly speaks as a genetic engineer: "we designed," we included," etc.
His is a religio-philisophic asertion until he addresses his own team, whereupon the strict disciplines of engineering prevail.
Where are those space aliens, SETI?toc
May 30, 2008
May
05
May
30
30
2008
04:33 PM
4
04
33
PM
PDT
Very interesting talk. Whatever one may say about Venter, his team is definitely tackling exciting questions.
I particularly liked the discussion about design, starting about 4:45 in. He gives the obligatory nod to traditional evolutionary theory a few times in his talk, but let's be very clear: everything they are doing, and every time he had to describe the processes and challenges they are tackling, he talked about "design."
Venter's work is good engineering -- based on the understanding of existing systems and careful application for new uses. It owes absolutely zero to the historical gloss of traditional evolutionary theory, whatever one may think of that.Eric Anderson