Even though scientists are generally pretty smart folks, they do make mistakes. There is a common misunderstanding about genetics held by quite a few real biologists, as well as plenty of laypersons. Here is the fact: when you study a structured population, such that alleles are not randomly distributed among all individuals, a correlation between a gene and a phenotypic trait does not mean that the gene is the underlying cause of the trait. For example, in humans, there is a positive correlation between the S allele of hemoglobin and dark skin. Does this mean the hemoglobin gene influences skin color? No, it just means that people whose recent ancestors lived in sub-Saharan Africa often have the S allele (due to selection for malaria resistance) and dark skin (possibly due to selection for defense against solar radiation), while people from elsewhere rarely have that allele and tend to be lighter in color. Stated this way, the logic might seem obvious, but there are a surprising number of biologists who unwittingly ignore this reasoning, as I will explain.
Consider, for example, this paper on the genetics of parasite resistance in mouse lemurs that came out in the journal Evolution last year. Evolution is a fairly prestigious journal, with a rigorous peer-review process, yet somehow no one noticed a rather glaring mistake. The authors hypothesized that variation at the major histocompatibility complex (MHC) in mouse lemurs (Microcebus murinus) was responsible for variation in resistance to roundworm parasites. MHC encodes proteins that present parasite-derived antigens to white blood cells so that the immune system can realize there is an invader and launch an attack. MHC shows extremely high diversity in many species, probably due to balancing selection, and many studies have reported correlations between MHC genotype and infection. So, it was a reasonable hypothesis. To test it, they measured worm infection in 67 lemurs and genotyped them at MHC. They found one allele correlated with susceptibility and two alleles correlated with resistance. However, they grouped several populations together to do this, and their data clearly show that allele frequencies are significantly different among populations. Furthermore, even within “populations” there are fewer heterozygotes than expected in a randomly-mating population (in what population geneticists call Hardy-Weinberg equilibrium), probably because the authors actually sampled multiple distinct populations without realizing it (the so-called Wahlund effect). Ironically, the authors noticed departure from Hardy-Weinberg expectations in a previous study of the same lemurs, but they failed to realize how that affects their conclusions in this study. Basically, one population might suffer from more parasites because they just happen to live in a habitat with more parasites, or they are more susceptible due to some completely different gene. Since MHC allele frequencies are different among populations, a spurious correlation would result. Is it reasonable to suppose that MHC influences parasite resistance in mouse lemurs? Yes. Has this been empirically demonstrated in this study? No.
The disturbing thing is, I was only able to figure this out because the paper included the genotype and infection data for every individual lemur. Other studies might make the same mistake, and it would be impossible to catch them at it since they don’t report their raw data. Notably, one of the authors of the lemur study has written several similar papers on other species, and there’s no way to tell if the same kinds of errors were made.
Here’s another example with a happier ending. Some colleagues of mine study the snail, Biomphalaria glabrata, that is the intermediate host for the human flatworm parasite Schistosoma mansoni. When they found that a resistant strain of the snail produces more hydrogen peroxide than a susceptible strain, they concluded that the amount of hydrogen peroxide a snail produces helps determine its resistance. They sequenced a gene involved in hydrogen peroxide production, and found that allele frequencies differed between the two strains. Thus, they inferred, this gene influences resistance to the parasite. Of course, they were making the same mistake that the lemur researchers made. Fortunately, they discussed their results with me and other members of my lab, and we encouraged them to look for a correlation within a single strain of the snail. Since the strain was in Hardy-Weinberg equilibrium, a correlation would show that the gene really was linked to the phenotype. As a matter of fact, the correlation remained even when a single strain was analyzed, and they published this much more conclusive evidence. It’s not the final word, because the real causative gene might be chromosomally linked to the gene they’re looking at. Furthermore, the best evidence for gene-trait correlations comes from crossing studies where you can control the genetic background. Still, they found a real correlation, whereas their original comparison between strains could easily have been meaningless.
These are just some examples I’ve noticed in the area of parasite resistance. There are probably many more out there. I’m not saying that I don’t make mistakes sometimes, too, but it’s important for scientists to discuss their findings with others before publication, to increase the chance that someone will catch this kind of mistake.