In its simplest sense, we imagine that learning occurs through a series of positive and negative rewards. Some actions lead to pleasure, others to pain, and it seems reasonable to expect that people will repeat the actions with pleasurable results and avoid those that ended in pain. Yet, we all know people who aren't deterred by the idea of punishment. We all know people who never seem to learn.
Could there be a physical reason, hidden in their genes?
In December 2007, Science published a study by Klein et. al. (1) where they asked if a specific genotype at a location called "DRD2-TAQ-IA" affected an individual's ability to learn from mistakes.
I'm intrigued by this question because I'm interested, in general, in the biology of learning and how this information can be applied to the practice of teaching. If we know that students with some genotypes learn better one way and students with other genotypes learn better another way, perhaps we can use this information to tailor our teaching and help all students get the most out of their time in the classroom.
Some background on the study
In this study, the participants were 26 males, who were placed into two different groups according to their genotype. The group with the A1- genotype had 12 members and the other group, with the A1+ genotype had 14.
The kind of genetic change that was used to group the different guys is called a RFLP. I have another post that describes RFLPs in more detail. In short, for this RFLP, the enzyme TaqI can cut a guy's DNA and make two fragments (A1+) if he has an A at a specific position on chromosome 11. If a guy has a G at that position, for both copies of chromosomes 11, then the enzyme won't cut the DNA, we won't see two fragments, and he gets sorted into the A1- group.
There are a few more details that are also important here. First, we all have two copies of chromosome 11. A person who has a TaqI site on both chromosomes would be called A1+/A1+, a person who only has a TaqI site on one chromosome would be A1+/A1-, and a person without any TaqI recognition sites would be A1-/A1-. As far as I can tell, this authors of this study grouped everyone with a TaqI site into the A1+ group and everyone else into the A1- group. Usually, these studies would put people into three different groups (A+/A+, A+/A-, and A-/A-) but that wasn't done here.
The study and some of the results
In one of the main experiments, Klein et. al. tested the ability of the A1+ and A1- guys to recognize certain symbols in 20 different trials and concluded that the guys with the A1+ allele were more likely to make a mistake and pick the wrong symbol. In the graph below, A was right, B was wrong and one group was better at avoiding B than the other group. In other words, the A- guys were better at avoiding mistakes (they avoided B).
That part of the study seemed fairly straightforward and seemed to be confirmed by the other experiments.
Klein and coauthors mentioned that another study found that the DRD2-TAQ-IA+ allele (polymorphism) was associated with fewer dopamine receptors on the surface of brain cells. They speculated that having more dopamine D2 receptors makes one more sensitive to negative feedback and better able to learn from mistakes. They also cited studies that claimed to link reduced dopamine receptor density with other conditions like drug addiction, obesity, and compulsive gambling. It all sounds pretty horrible.
And yet...
The problem
I found a problem with the study that seems to have been overlooked.
Many of the conclusions in the paper result from the notion that this polymorphism is located within the dopamine D2 receptor gene and somehow affects either the functioning of the dopamine D2 receptor or the levels of this protein.
But when I looked in OMIM (even last December), I found that the genetic change described by Klein isn't located in the dopamine receptor gene (3). In fact, it's possible that the DRD2-TAQ-IA polymorphism doesn't have any effect on the dopamine D2 receptor.
From OMIM (3):
Neville et al. (2004) showed that the TaqIA RFLP is located in a novel kinase gene designated ankyrin repeat and kinase domain containing-1 (ANKK1; 608774), located downstream of the DRD2 gene in chromosome band 11q23.1.
I looked at dbSNP to investigate the frequency of the A1+ allele in different populations. In dbSNP, I found that the percentage of A1+ people varies from 10% in Europeans to as high as 50% in Asian populations (4). It seems hard to believe that such a large fraction of people would have a harder time learning from their mistakes or suffer from all the other problems that were attributed to this allele.
Learning where this genetic change is really located makes me wonder if some of the many qualities that are attributed to the DRD2-TAQ-IA polymorphism are real or if those qualities were partly inferred because researchers thought this change was located in a receptor for dopamine.
As long as we're talking about learning from mistakes, perhaps someone should suggest to Klein et. al. and the reviewers from Science that they consider learning how to use OMIM.
References:
- Klein, T. (2007). Genetically Determined Differences in Learning from Errors. Science, 318(5856), 1642-1645.
- Neville, M. J.; Johnstone, E. C.; Walton, R. T. :
Identification and characterization of ANKK1: a novel kinase gene closely linked to DRD2 on chromosome band 11q23.1. Hum. Mutat. 23: 540-545, 2004.
PubMed ID : 15146457 - OMIM, dopamine receptor D2, accessed 6/17/2008
- http://www.ncbi.nlm.nih.gov/SNP/snp_ref.cgi?rs=1800497
- Log in to post comments
Wow, good research! I wonder what the authors (and the responsible editor at Science) will say...
BTW, there is evidence currently under review (re-submitted) that it may indeed be possible to be genetically incapable of feedback learning but be perfectly normal in all other forms of learning. If this is the problem of these people, I'd predict that the DRD2-TAQ-IA polymorphism is somewhere in one of the protein kinase C isoform pathways...
Thanks Bjoern,
According to Neville, Johnstone, and Walton, the TaqI polymorphism maps in ANKK1, a kinase named ankryin repeat that may function in signal transduction.
This seems to have been confirmed by a few others who are listed in the protein record.
I just came here to post this.
"Many of the conclusions in the paper result from the notion that this polymorphism is located within the dopamine D2 receptor gene and somehow affects either the functioning of the dopamine D2 receptor or the levels of this protein."
A polymorphism does not have to be "within" a gene to influence levels.
Klein et al. cite 3 papers that find an association between the polymorphism and function. Doesn't this address your concern? (I haven't read the papers so I don't know if they're good or if there's opposing research, but there is at least some data to support them.) DRD2 and ANKK1 are adjacent and transcribed in opposite directions, transcription of one could influence transcription of another.
Both the Neville paper and the OMIM record state that there's some disagreement about whether the TaqI polymorphism has any effect on dopamine D2 receptor levels. This point was ignored in the paper and the topic was discussed as if there's no disagreement at all.
I agree that the polymorphism could influence receptor levels, but it's not clear that those data exist.
Your criticism of the Klein et al paper is commendable, but misdirected.
The authors do not assert that the Taq1A polymorphism is causative of the observed phenotypic variation, and it has been established that this polymorphism (agreed, downstream of DRD2 within ANKK1) is in linkage disequilibrium with polymorphisms across DRD2 gene. This should have been discussed in the paper. Typing for this polymorphism alone is a sloppy surrogate for haplotyping across the known polymorphisms of DRD2 or, better yet, resequencing on all subjects. There were only 26 subjects, after all.
Zhang et al (PMID: 18077373) show that different variants within the DRD2 gene affect gene expression and are in linkage disequilibrium with the Taq1A polymorphism. These functional variants of DRD2 are associated with fMRI alterations during memory tests.
That Klein et al. disregard the third genotype (homozygotes for the less frequent allele) is addressed in their online supplement. The homozygotes would be expected to represent only 3% of the target population. However, with only 26 subjects in the study, the authors should have either excluded them or discussed how many were included and justified their inclusion in the A1+ population by addressing how results from the homozygotes contributed to the aggregate A1+ data.
The more appropriate criticism is that the entire field of associations between genotype and human behavior is besotted with a load of sensationalist hogwash. A good headline (Gene X Causing Aberrant Behavior Y) is too delicious to pass up, and Science is not immune. (I can imagine the articles in the local newspapers picking up on the Science press release: ?Study Shows German Men Are Genetically Unable to Learn from Their Mistakes?). A quick trawl of PubMed links DRD2 polymorphisms to: schizophrenia, drug and medication seeking behavior (heroin, cocaine, nicotine, and alcohol abuse), overeating, binge eating, anorexia nervosa, bulimia, anxiety, restless leg syndrome, ADHD, juvenile delinquency, vulnerability to violent victimization, number of sexual partners and human loving styles (I kid you not).
Methodological flaws are abundant in this literature. Many contain obvious problems in the genotyping (see Sand 2007, PMID: 18154681). Most have samples sizes that are ridiculous to support generalize, and over-mine the data for correlations. Most limit their selection of subjects to a specific co-descendent population (which may not be a bad thing, if the causative mutations are actually private to that population, but leads to chaos when SNP-defined haplotypes are used in studies on different populations.) All the studies lump together genotypic groups based on associated markers, with little or no discussion of whether a common causative mutation is responsible for the measured phenotype within the group or if multiple mutations lead to a common phenotype. And the behavioral classifications are not particularly objective or rigorous.
In the first paragraph, they say this:
I interpreted their use of the word "modulate" to mean that they think this polymorphism does affect the receptor density phenotype.
Other than that, I'd say we agree.