September 30, 2005

Remembering A Long Life

A bit of a self-plug: Joćo Pedro de Magalhćesa has written a review paper on cognitive ageing where I have played a minor role:

Joćo Pedro de Magalhćesa & Anders Sandberg, Cognitive aging as an extension of brain development: A model linking learning, brain plasticity, and neurodegeneration. Mechanisms of Ageing and Development, Volume 126, Issue 10, October 2005, Pages 1026-1033

This is linked to the model in chapter 7 in my thesis. The basic idea is that mental ageing is the result of a development program that just continues long after we have grown up. To maximize the stored information in the brain at reproductive age the learning rate needs to decrease from childhood and onwards, but since there is no evolutionary pressure to add a stop mechanism this learning rate reduction continues long after it is adaptive.

It will be fun to see if this hypothesis works out; if it holds true we need to find ways of adjusting brain learning rates as we extend lifespan. A too high learning rate will erase old memories, a too slow learning rate will prevent new information from arriving. We will probably need some way of re-training valued older memories to keep them, a kind of life-long memory consolidation.


Posted by Anders3 at 09:58 AM | Comments (718)

September 29, 2005

You Are What You Eat

I attended the Livsmedel, genetik och genteknik conference, organised by the Gene Technology Advisory Board (and all the other usual suspects). This time the theme was food and genetic technology. By a nice non-coincidence Eudoxa released a report on GMOs at the same time.

The debate seems to have cleared up - at least when listening to the speakers. Rather than weighing for or against GMOs in general, they were going into the pros and cons of particular strains. This is a very important shift, moving the debate away from naively saying yes or no to the entire technology to a more mature look at practical benefits and risks. But the old guard was still present, mostly represented by the green party in the audience. While the stakeholder organisations have refined their discussion, this might not be reflected in the public debate.

Public views on GMOs remain negative in the EU, and policies reflect it. Most of the studies on GM attitudes are based on what people say they think they would do. Given that it is nearly impossible to buy any GM food if you want it in Sweden, and there are no GMO farming, much of the discussion become rather theoretical. People say they might want to buy GMO products if they are good for the environment or healthier, but I wouldn't be surprised that price and other features would have a larger effect if they were actually available.

There was an unstated assumption in the air that GMOs are unstoppable/the future. Which seems reasonable given how they are embraced by the US, Canada, Brazil, China and India. The EU might end up as a non-competitive GM-free zone in the world economy. And as long as there is no way of experiencing GM products attitudes will remain locked. On the other hand, EU is probably more split when it comes to GMO attitudes than Brussels likes to think.

It would be interesting to look at attitudes to ornamental uses. Would people find a GM tulip acceptable?

Several talks underscored the vast power of traditional breeding in making high yield crops and animals. Milk and egg production has doubled in 40 years, and pigs produce 10-15% more in just 20 years (as consumer demands for fat content have changed the shape of pigs have also changed noticeably in just a few generations). I think many see this as just an attempt to normalize GMOs, but it really shows something far more important: we already live surrounded by altered organisms, but these are not strongly regulated. Add a single gene to a crop and you will have to fill out literally tons of paperwork. Hybridize and mutate a crop wildly and then culture it, and you can sell it immediately. If one thinks GMOs could be risky one should really try to stop traditional breeding. That genes may be moved between species in gene technology is probably less risky since they are documented and monitored changes. Who knows what long silent genes might be activated when two plants are hybridized?

A fun lecture held by Jonas Blomberg discussed whether DNA can be absorbed from food. My intuition (and every other colleauge I asked) was that this would not be possible, but Blomberg et al. showed (Uptake of amplifiable fragments of retrotransposon DNA from the human alimentary tract, Mol Genet Genomics. 2003 Dec;270(4):362-8) that one could detect rabbit retrovirus DNA a few hours after eating a rabbit dish. The uptake seems to be extremely slight, but it is still surprising. My guess is that one could combine the numbers in this study with the experiments in using bare DNA for gene therapy to further calculate the probability of actually taking up genes from food.

Of course, this result is not a demonstration that GMOs are risky, but rather that food is (very very slightly) risky. Just imagine if a cell gets the gene for solanine production from the potatoes we eat!

Ewa Rabinowicz had a very interesting lecture on the agri-economics of GMOs. She showed that most growers have gained from GMO; in the case of Bt cotton US farmers have gained $105 million, the agbio companies $80M and consumers (through lower prices) $45M. Growers elsewhere lost $15M through competition. GMO farming in China have apparently especially helped the very small farmers (<1 ha). This is because biotech tends to be scale neutral: mechanization helps the big farmers through economies of scale, but GMOs help both similarly (however, heavy regulation of course creates economies of scale - it is easier for big agricorps to hire lawyers and interface people than small ones). Dividing between GM and non-GM products (which labeling forces) also destroys economies of scale, and since it might be hard to transfer the costs this causes it might weaken the market altogether. The fears of GM monopolies have not yet materialized; maybe because it is an early stage and there are always alternative crops to shift back to, but it could simply be that the fears have been too large. She also argued persuasively for how GMOs could help the third world break vicious circles of poverty, low productivity and environmental destruction. She was especially hopeful about helping farming in marginal areas.

Göran Hallmans pointed out the complexities of nutrigenomics: finding the kinds of food that are healthy for you is hard, even with gene maps. Today such data is found using large epidemological studies that often produce contradictory or weak results. This of course risks ending up in the domain of untrustworthiness even with impeccable science. Here we will likely need new research methods, perhaps based on volunteer people using ubiqitious computing to monitor their eating and habits and then pooling it together for data mining.

Sven Ove Hansson discussed food ethics, suggesting two principles to guide food policy. The first, based on autonomy, is to respect the individual's choice of what to eat. The second, similar to the medical ethics idea of treatments having to be in the interest of the patient, is to erect "shared barriers" for what food may be morally sold and distributed. These barriers are in the interest of all people in a society. Given that we live in secular and liberal democratic societies they cannot be based on religion or ideology; hence they need to be based on issues of safety and quality based on scientific considerations. He then went on discussing the proper form of precautionary thinking necessary for this.

But the problem in this area is that policies originally made using uncertain knowledge need to adapt as science gathers new data. But laws are very slow to change, and negative scientific results have far less impact than scary warnings, even if the majority of studies do show safety. Hence there is a risk of excessive caution and conservatism.

Much GMO resistance is probably based on anticapitalist ideas of big companies. I wonder if not GMOs would be more accepted if they arrived from small garage companies. But the heavy regulations ensure that only the big boys are allowed to play. Legally at least, and only right now. In a few decades biohacking will be so easy that protecting the environment will require other means than just regulations. Most likely we need active shields or ecological immune systems.

All in all, it seems very few of the risks ascribed to GMOs are specific to GMOs themselves. But people prefer to focus on the new and hubristic so that they do not have to think about the risks and strangeness of ordinary food production. Agriculture is after all becoming a technical service rather than a way of life or cultural symbol.

Posted by Anders3 at 10:25 PM | Comments (1279)

September 10, 2005

Technology Frees the Body

teknikenbefriarkroppen.pngHealth Consumer Powerhouse has published the report Tekniken befriar kroppen by me, Waldemar Ingdahl and Alexander Sanchez. It deals with emerging technologies in the health care system and how they might stimulate - or force - organisation changes. So far it is still just in Swedish, but there is an english summary.

As I see it, modern medicine is entering a paradoxical world where the technical abilities are accelerating, but the ability to implement these in the old healthcare systems is constant or even diminishing. Information medicine and biotech medicine hold enormous potential to improve life, but the highly regulated and infelxible health care systems are slow to absorb them. Sometimes this is due to regulatory uncertainty such as the one surrounding tissue engineering (is it a drug, device, transplant or something else?) or stem cells. Sometimes it is organisatiorial, such as the problem of reorganising the diagnosis-treatment chain to use early 3D medical imaging, networked e-health and multidisciplinary treatment teams. Sometimes it is overregulation and risk aversion, which is forcing the pharmaceutical industry to put its money on safe bets rather than real innovation.

In the long run medical advances will likely get out to us all anyway. But it makes a lot of difference to me if I get better care now or ten years after I'm dead. There is a real need for policymakers to realise that we need more experimentation and freedom in the health area.

Posted by Anders3 at 02:32 PM | Comments (1114)

September 06, 2005

Keys of the Heart

ECG to identify individuals by Steven A. Israel, John M. Irvine, Andrew Cheng, Mark D. Wiederhold and Brenda K. Wiederhold. (Pattern Recognition Volume 38, Issue 1 , January 2005, Pages 133-142)

It sounds like a plot device in a Mission: Impossible episode: "But that is not all, team. The vault is protected by a heartbeat recognition system."

The system works by first measuring the heartbeat. It then filters out noise and drift with a bandpass filter. It is usually hard to avoid getting some of the 50/60
Hz hum from power outlets and slow drift up and down as the contact between the electrode and the skin changes. Then it finds the heartbeat pattern, stretches it to compensate for fast or slow beats (which depends on stress) and measures the time between different parts of the beat (detected by changes in curvature of the time-voltage curve). This produces 15 numbers that characterize the beat, and can be used in a standard machine learning method to create a classifier that identifies people using their beat.

According to the paper this works pretty well, both for different electrode placements and heart rates. It gets the person right in 87% (or better) of the cases, and handles different stress levels and placements well. The authors are of course suggesting that it is a viable biometric.

I don't think it would work as a stand-alone biometric. Even getting a 99% recognition rate means that every hundredth person will fail the test. Weighing together several biometrics (iris or retinal scans, face recognition, fingerprint etc) would make identification more robust. The study also tested the system for about a dozen subjects. Given that there were 15 features used to identify them, it is not a very hard learning problem. I think it will be much less effective on larger groups, where people are more likely to have similar beats. There isn't that much room for individual differences in heartbeats, at least not among healthy people.

Heartbeats are relatively easily recorded, and it seems one could steal one with the right equipment. Record the beat, place an electrode (with insulation beneath to hide the real heartbeat) on the chest, link it to a player sending the signals back and apply the biometric electrode when trying to get in. This can be avoided by having a person do the biometric check, but if you have guards at the access control, why bother much with the biometrics in the first place?

Posted by Anders3 at 05:48 PM | Comments (889)