Lies, damn lies, and academics on GWAS

A colleague of mine back at Cambridge (my good ol’ alma mater) studies the history and philosophy of science, and they recently told me about a lecture they had on GWAS. Or perhaps the lecture was on the problems with GWAS, because the impression of the lecture they transmitted would certainly support that interpretation.

One criticism that stuck out to me in particular was the claim that GWAS cannot be trusted because current and future GWAS are conducted upon the basis of databases compiled from results of previously published GWA studies, which can allow for false positives and mistakenly identified variants to proliferate as the research area continues to develop. This is of course why we should be very careful about relying on GWAS, especially for studies on intelligence differences – something the lecturer took pains to emphasize.

This situation interested me, because two things are going on at different levels. On the one level, the lecturer is completely correct – even a small margin of error will inevitably result in a small probability of false-positives that are compounded within future research, allowing for small mistakes to accumulate into big mistakes over time. In truth, this is a feature of human knowledge systems (to which the scientific method belongs) in a broader sense, because the validity of every next step in science depends upon the validity of previous steps, and there is always a non-zero chance that previous steps were wrong-footed. As Gregory (Role of Probability Theory in Science) states:

“Of course, any theory makes certain assumptions about nature which are assumed to be true and these assumptions form the axioms of the deductive inference process… For example, Einstein’s Special Theory of Relativity rests on two important assumptions; namely, that the vacuum speed of light is a constant in all inertial reference frames and that the laws of nature have the same form in all inertial frames”

This is a useful example because Einstein’s assumptions are not a priori; they are instead assumed based on other developments in physics or natural philosophy which preceded them. As any coder will know all-too well, this cascade of ‘potential errors’ means not only that you have a reasonable expectation of encountering an ‘actual error’ within the system of knowledge you’re dealing with, but that these could also be concealed within your current and future research projects conducted in alignment with that method.

So again, since GWAS is a highly technical application of the scientific method for specific purposes, and this cumulative error probability is indeed a common feature to all science, it could be described as a ‘pitfall’ of GWAS. But in accepting that GWAS is ‘sketchy’ or ‘unreliable’ because of this, we’re also forced to accept that all science – the science that drives our cars or powers our lights or cools our refrigerators – is equivalently imperiled. Have you ever been cautioned about the untrustworthiness of lightbulbs or refrigerators due to human epistemic constants? You can probably guess by now that something else is going on here.

Let’s try a thought experiment. Imagine two men are having a conversation about a cute girl they both know. MAN-A is curious about the girl, and asks MAN-B for his opinion. Imagine that MAN-B excoriates the girl for the following reasons: “Oh hell no dude, that girl is disgusting. Do you realize she shits? Like, she actually goes to the bathroom? And what’s worse, I heard she gets periods constantly, and even sneezes in the springtime. I wouldn’t be caught dead with her.”

If MAN-B in our thought experiment seems stupid, misogynistic, or even totally detached from reality, it’s because he is. As unattractive as human bodily functions might be, they’re described in relation to the ‘human body’ for a reason; they’re universal. They are features, not flaws, of human physiology that all of us still alive are equally guilty of. If you ever find yourself interested in or curious about a nice girl you meet and you hear similar remarks reflected in a mutual friend’s opinion of her, he really shouldn’t be your friend.

The illogicality encapsulated within this highly vulgar and contemptible example, where MAN-B reveals his nasty attitudes through ridiculously discriminatory standards, is really the same thing you’re seeing in the GWAS example given above. It makes no sense to criticize a girl for having features common to all people or women, unless you’re an utter jerk with a grudge against them. In the very same way, it makes no sense to argue that GWAS are invalid or untrustworthy because of a feature common to literally all science. Yet if you are a hypocritical person who simply dislikes GWAS specifically, it makes perfect sense for you to use flawed reasoning and discriminatory standards to support your equally unjustifiable hostility.

To clarify, I am not saying (nor have I ever said) that GWAS is or should be free of criticism. One of the most important criticisms of GWAS is the distinctly WWEIRD (White, Western, Educated, Industrialized, Rich, Democratic) shade of GWA study participants, which precludes people lacking these characteristics in other parts of the world from sharing in the undeniable benefits that GWAS have brought to healthcare and family planning. Consider how fundamentally different this is to previous eras, where new innovations (e.g. the railway, 3D printing) could bring benefit on a global scale regardless of where or by what demic group (whites and the Japanese, respectively) they were pioneered. Today, cutting edge GWAS results in educational attainment, disease risk, and psychopathology continue to expand the scope of informed options available to white prospective parents or healthcare recipients, for whom these results are applicable. But if by chance you happen to be Desi, or Asian, or Sub-Saharan African, or of any other genetically distinct grouping, these benefits are often unavailable. Razib Khan wrote about this recently:

Because most GWAS are performed in European populations, PRS values for individuals not of European ancestry are far less accurate. This phenomenon is caused by several factors. One of the major ones is that each population has genetic variations that cause diseases special and unique to a given population (“private alleles” in the jargon). Studies which use only Europeans cannot detect unique variation in non-European populations by definition. Those variants are not found in Europeans! Additionally, sometimes genetic variants even give different risks in Europeans than non-Europeans because of interactions of genes. The predictions in one population do not transfer to another.

This is a serious problem with current GWAS research that I would expect any decent person to express concern about. It becomes especially relevant for family planning, because in a number of non-Western societies infanticide remains a common method for dealing with children with unwanted genetic diseases – something that embryo selection, for instance, would render obsolete.

Clearly, this is not what the lecturer said. The lecturer attacked GWAS on the basis of ‘muh cumulative error probability’ despite this being a feature inherent to the scientific method itself. All before an audience of philosophy of science students, no less.

I am not in favor of arbitrarily attributing ulterior motives, particularly bad faith, to others with whom I disagree. But in cases with fallacies so obvious, and discrimination so blatant, it’s almost unreasonable to think that underlying hostility isn’t at play. While it would be completely fine to give a lecture on epistemic issues in science in which you discuss the example of cumulative error probability in relevant GWAS cases, to selectively apply this principle to undermine GWAS as a whole is a logically invalid weaponization that says more about you as a lecturer than it does about GWAS itself. Perhaps if you had actually done your research into the topic you claim to know so much about, you would be lecturing instead on real issues specific to GWAS, such as the lack of diversity mentioned earlier. By neglecting to do so, you actually reveal yourself to be ignorant of the discussions within relevant disciplinary communities (e.g. BehavGen) regarding the limitations or flaws of GWAS that aren’t simply universal features of the scientific method.

Again: if a friend badmouths a girl because she has a feature literally everyone else has, he probably shouldn’t be your friend. If a lecturer badmouths GWAS because it shares features common to all applied scientific fields, well…

…You’re probably in college.

Leave a Reply

Your email address will not be published. Required fields are marked *