This is part three in a five-part series called "The Limits of Accelerating Returns" that focuses on the limitations of Ray Kurzweil's Law of Accelerating Returns when applied to molecular biology and biomedical technology, including longevity treatments. The other articles in the series are "The Limits of Accelerating Returns," "Biology is not Digital," "Some Rates are Fixed," and "Implications of Fixed Returns."
In his essay called “Making the World a Billion Times Better,” Ray Kurzweil writes, “The approximately 23,000 genes in our cells are basically software programs, and we are making exponential gains in modeling and simulating the information processes that cracking the genome code has unlocked.” The problem is that genes are only roughly analogous to software programs, as I discussed in the previous post in this series, and simulating the information processes of the genome is non-trivial in the extreme.
Let’s take a very simple example of an artificial genome called a “repressilator” that was added to bacteria as a proof-of-concept of synthetic biology and biological simulation. The repressilator’s machinery consists of three genes, each of which turns off one of the other genes. When one of the genes is active, it turns off one of the other genes. Meanwhile, the third gene is in the processing of shutting down the first. When the first gene is turned off, the second one turns on, and stats shutting down the third gene. Thus, there’s a cycle or oscillation to the activation of the genes (the name of the construct is a contraction of “repression” and “oscillation”).
The system also includes a fourth gene that produces a visible signal so that the oscillation can be tracked. This seems like a pretty straightforward system. The scientists who developed it even did a lot of modeling to figure out how to optimize the oscillations before they went to the trouble of building the thing.
But it didn’t work the way they expected.
Given all the optimizations that went into the design, a single oscillation might be expected to take 20 minutes or so. Instead, each cycle took several bacterial generations to complete, which means that the state of the repressilator in a parent cell was passed on to its progeny. It also means that the construct wasn’t as optimized as the simulations had suggested.
A simulation, after all, is only as good as the knowledge that went into its creation. Garbage in, garbage out.
I don’t mean to disparage the repressilator’s creators. Not at all. Their work was seminal and will be taught at universities for at least the next 50 years. Rather, the disparity between the represillator’s simulated behavior and its actual behavior serves to illustrate just how noisy and complicated a lowly bacterium is and just how little we know about it.
An immense amount of data is required to accurately simulate simple biological systems. We do not, at present, have most of this data, even for the workhorse E. coli bacterium. Furthermore, it’s not easy to get.
Nonetheless, it’s reasonable to expect that we’ll have sufficiently accurate bacterial simulators within 10-20 years. That’s great if you’re designing repressilators, but if you want to extend life, as Kurzweil does, you need to simulate human physiology.
So the question is, how much more complex are you than a bacteria?
Unfortunately, there are no objective measures of organism complexity. Consider that a mosquito’s genomes has twice as much DNA as a fruit fly’s, but they have about the same number of genes. Moreover, you have only 35% more genes than either insect.
For arguments sake, let’s say a human is about 2 billion times more complex than an E. coli (2 billion is also roughly the number of years since the species split). Assuming a doubling of computing power every year, it will take about 30 years for our bacterial simulator to evolve into a human simulator.
That figure also assumes that the only thing holding back our human simulator is computing power. As we’ll see next time, that’s just not the case.
