Signature in the Cell is a book published in 2010 by Stephen C. Meyer. On page 283 to 286 he evaluates the Ev program. This is a review of that section of the book.
The book is available several ways:
|Jump to the relevant passages of Signature in the Cell discussed below with this link.|
|The original scientific paper by Thomas Schneider that describes Ev is available through the page "Evolution of Biological Information".|
|You can try the Evj program yourself!|
|There is a Tutorial guide for running Ev.|
Below is my point-by-point critique of Meyer's discussion
of the Ev program.
Note: the quotes here are for the purpose of commentary, criticism and scholarly reports and so fall under the fair use doctrine.
From page 283:
"Schneider claims Ev can simulate the production of nucleotide binding sites."Actually, it's 'modeling' because Ev is a 1 to 1 model to what happens in the natural world. This is confirmed by the results which show that Rsequence evolved to match Rfrequency which is a result previously observed in natural binding sites. See: Schneider et. al 1986.
"Nucleotide binding sites are very short sequences of nucleotides on the genome located upstream of specific genes."Actually they are often found inside genes and between genes.
'Schneider claims that "the program simulates the process of evolution of new binding sites from scratch."'Ev models this process using natural selection. Meyer here is indirectly implying that he thinks that the information in the binding sites is not zero, but it is clear that Rsequence is near zero at the beginning of the run, as can be seen in this graph of the evolutionary process by Ev:
"Schneider supplies Ev with a target sequence, in this case a particular sequence of nucleotide bases that function as a binding site.This is wrong. The program does not have a particular target sequence and you can show this by running it repeatedly with different random seeds or by wiping the sites out by turning off selection and then turning selection on again. The particular pattern that appears is generated randomly. The important measure is of the amount of pattern (Rsequence), and that matches the information needed to find the sites (Rfrequency). Meyer either did not read the paper or he misunderstood how the algorithm works.
"He then has the program generate a random crop of sequences of equal length."After the initiation, Ev does not create entirely random sequences as this sentence appears to imply. The genome lengths are fixed and they are mutated by changing only 1 base per generation (in the standard run). The sequences are only created entirely de novo when an Ev run starts.
"After favoring sequences that manifest the general profile of a binding site, ..."This is entirely fabricated since the program does not do this at all. Anyone can confirm this by looking at the code, which is available! Does he mean that sites have a fixed length? This is the situation found in nature since proteins that bind DNA have a finite size!
"... Ev applies a fitness function to the remaining sequences"There is no artificial fitness function or variable in the program. Instead, organisms make mistakes and compete against each other.
"The fitness function then assesses the degree of divergence between the mutated sequences and the target sequence and applies an error value to each mutated sequence."There is no target sequence!
"Then he repeats the process again and again, until finally Ev converges on the target sequence."There is no target sequence.
"Ev incorporates one additional step that Dawkins's and Küppers's simulations lack. Before Ev applies its fitness function, it applies a filter to the crop of mutated sequences. The filter favors sequences that have the general profile of a binding site. Like the fitness function, this coarser filter makes use of information about the functional requirements of binding sites to favor some sequences over others. As such, it imparts information based on knowledge that Thomas Schneider, not natural selection or the environment, has imparted into the Ev simulation. "I don't know what Meyer is talking about here. There is no second filter! This is apparently a fabrication. As Chris Adami has pointed out, natural selection absorbs environmental information. In this case Rfrequency is the measure of that environmental information and indeed Rsequence, the information in the binding sites, converges to Rfrequency by natural selection. In other words, the information inside the organism (Rsequence) evolves to match the information about the environment (Rfrequency). This result is exactly what is observed in nature, as shown by Schneider et. al 1986.
Apparently Meyer did not read the EV paper carefully enough to understand exactly what it is doing.
I had a student once who understood every step of the Ev algorithm but could not grasp the entire process.
Thanks to Paul Anagnostopoulos for bringing these passages in Signature in the Cell to my attention.
origin: 2013 Nov 08
updated: 2013 Nov 09
U.S. Department of Health and Human Services | National Institutes of Health | National Cancer Institute | USA.gov |
Policies | Viewing Files | Accessibility | FOIA