By Daniel Tarade
If you should go skating
On the thin ice of modern life
Dragging behind you the silent reproach
Of a million tear-stained eyes
Don't be surprised when a crack in the ice
Appears under your feet
You slip out of your depth and out of your mind
With your fear flowing out behind you
As you claw the thin ice
Thin Ice- Pink Floyd
Practitioners of the scientific method march towards capital T-ruth. That is the ideal, according to science advocates. Through empirical observation, scientists arrive at the bedrock of the universe; they uncover the laws, unflinching and forever to be, that are used to deduce and predict any natural phenomena. Yet, debate on the presence of Truth continues. Certain philosophers of science espouse that Truth exists but can never be known. Only approximated. Others argue that Truth is a myth and does not exist. This article will not provide an exhaustive overview of these positions. Instead, I want to clamber down a rabbit hole hidden behind a simple scientific claim and imagine what Truth might look like in an all-knowing world. I previously critiqued one of my own manuscripts to highlight the gaps that exist between scientific claims and scientific observations. If Truth is to exist, that gap will have to shrink to nothingness. As I have recently published another article, I am up to the challenge of Truth and science. What limits on Truth exist, and what can humans can uncover?
As a quick recap, my research focuses on oxygen-sensing proteins found in animals. There are three proteins central to the story; the closely related HIF1α and HIF2α proteins and their negative regulator VHL. Hypoxia inducible factor (HIF) is a master regulator of the cellular milieu under low-oxygen conditions. In fact, adaptation to low oxygen conditions is contingent on the HIF family of proteins. It is important to understand that HIF comes in several flavours (AKA paralogs): HIF1α and HIF2α. Under conditions of sufficient oxygen, a group of enzymes stick molecular oxygen onto the HIFα proteins. The HIFα protein, now labeled with oxygen, are recognized by the VHL protein, which targets HIFα for degradation. Thus, under homeostatic conditions, HIFα proteins are inactive. When oxygen is limiting, enzymes no longer modify HIFα, which itself becomes invisible to the VHL machinery. As oxygen decreases, HIFα protein accumulates and enacts functional changes. Imagine this transition as the promotion of HIF to commander-in-chief. HIF is a transcription factor that activates dozens of other genes. The question that we tackled was simple; does VHL differentially regulate HIF1α and HIF2α?
To begin exploring this question, we utilized a simplified biochemical system. Rather than work with a model organism, we instead worked with model proteins. See, the interface sufficient for binding between VHL and HIFα is known. So, we performed interaction studies using a twenty amino acid HIFα peptide and full-length VHL protein. We studied how tightly VHL bound to the HIFα paralogs and observed that VHL binds approximately two times more tightly to HIF1α than HIF2α. By studying mutant peptides of HIF1α and HIF2α, we found that a single amino acid dictated this differential binding (remember that proteins are polymers of amino acids). At one particular 3-dimensional co-ordinate, HIF1α possesses a methionine and HIF2α possesses a threonine. Our reasoning as to why there was a differential binding took into account the corresponding VHL amino acid (in vertebrates, a phenylalanine) that was positioned next to either methionine or threonine. Previous computational chemistry experiments had posited that the methionine-phenylalanine interaction is stronger than the threonine-phenylalanine interaction. These observations helped us formulate a model that we then applied to the whole of animal evolution. If methionine is associated with increased VHL binding and other residues are associated with weaker binding, might we learn something about the role of HIF throughout evolutionary time? We looked at the sequence of HIFα and VHL in dozens of invertebrate and vertebrate species. We found that protostome species, particularly the arthropods (think insects), featured substitutions of both the HIFα methionine and the VHL phenylalanine. Conversely, in deuterostome species, which includes sea urchins and certain worms, we find a high prevalence of HIFα methionine and VHL phenylalanine. Importantly, in vertebrate species, HIF1α always has a methionine and HIF2α in all but one instance has a threonine. Our conclusion based on these observations is complicated and esoteric, even for our field, but we are not interested in animal oxygen-sensing, are we? No, we are here to look for Truth.
Let’s tackle the low-hanging fruit first. By performing real-time binding experiments, we reported that VHL dissociated 1.704379562 times more quickly from HIF2α than from HIF1α. Is this Truth? The first line of the evidence that says no is that we repeated the experiment three times and obtained different results each time; variability in scientific measurements is inherent to our instruments and abilities. Based on statistical modelling, the ‘Truth’ could be anywhere from 1.6573439 and 1.7511501 with 95% confidence. A common position on Truth is that it exists but is not accessible to us. Taking this argument further, as instrumentation becomes more precise so too will the measurements. The models will suggest finer and finer ranges of acceptable Truths. But, this all gets more complicated. We performed a similar binding experiment using a different type of instrument. With this machine, we observed values that were an order of magnitude larger than those we observed first. We attributed this to a higher temperature made necessary by the instrument’s setup (25 °C v. 4 °C). This is not surprising; increased kinetic energy speeds up the rate of association and dissociation events. Under these conditions, HIF2α dissociated 2.812324 times faster than HIF1α. Thus, in a quest for a never obtainable Truth, a scientist would have to perform these association experiments under all temperatures. The same goes for all other variables, such as salt concentration and pH. Under all these conditions, the rate at which HIF proteins bind to VHL will change. As there are an infinite number of temperatures and concentrations, it is impossible to complete all the necessary measurements. The best possible truth would require interpolation and resembles a phase diagram more than any single number. There are other practical considerations. We work with bits and pieces of these proteins because working with the full-length sequence is difficult in some cases. Our binding experiments involve immobilization of the HIFα peptide on a substrate and measuring changes in the local environment upon binding of VHL. However, HIFα would never be immobilized in a cellular environment. In theory, these could all be rectified as technologies advance. Still, we would be left with boring truth made subjective by randomness and experimental error. But, what if we spoke in the abstract? Forget quantitative Truth. Does qualitative Truth exist?
What do I mean by quantitative and qualitative Truth? Quantitative Truth is expressed as an absolute value. For example, the speed of light is 299 792 458 meters per second. Qualitative Truth is stated as a principle. Nothing can travel faster than the speed of light. As I explore scientific quantification, it seems clear that quantitative Truth cannot exist. One possible exception are the scientific truths expressed with discrete values. There are eight planets in our solar system. I will get back to this point later, but the example I choose intentionally highlights a role for subjectivity in scientific Truth. Let’s instead discuss these qualitative Truths. Even if I cannot nail down with absolute certainty how much more tightly HIF1α binds to VHL than HIF2α, would it be unTruthful to say that HIF1α does indeed bind more tightly? Just as before, many technical limitations arise. It is impossible to directly observe HIF1α in the cellular environment. Only by affixing fluorescent molecules can we visualize a protein with a microscope. At such a point, you are no longer studying native HIF1α. Just like our biochemical experiments stripped HIF1α of its native context, it is impossible to observe native HIF1α without manipulation. Humans observe a world that they too exist within. It follows that it is impossible to know qualitative Truth. Advances in technology might minimize the manipulation of a system required for observation, but it is inconceivable that observation without manipulation could ever exist.
There is one additional flaw in the conception of scientific Truth. It is evident that any communicable Truth requires operationalization. Let’s revisit a candidate qualitative Truth; HIF1α binds more tightly than HIF2α to VHL. That statement requires caveats a plenty. One, how do you define HIF1α and HIF2α? Which species does one study? What if HIF1α does not bind more tightly to VHL than HIF2α in lamprey but does in humans? We would have separate Truths for different organisms. If we restrict our line of inquiry to humans, which sequence do we study? Every gene exists in flux within a population. There exist humans with subtle changes in HIF1α and HIF2α. Other humans have serious, disease-causing mutations in HIF2α or VHL. Thus, we will have individuals Truths for wild-type HIF1α and HIF2α sequences (a wild-type sequence is the most common sequence found within a population) and for other rarer sequences. What about modifications of HIF1α and HIF2α? We already encountered a common modification that regulates interaction with VHL: hydroxylation (the sticking of oxygen to a protein). So, we must specify hydroxylated but otherwise unmodified HIF1α and HIF2α. How does one define tightness of binding? In general, scientists use affinity constants. An affinity constant is the concentration of one protein needed for half of another protein to be bound. It is a stochastic measure designed to represent a collection of transient events; proteins interactions are not static but breath in and out. Even if one protein binds tighter on average than another protein, any given ‘weaker’ protein might interact longer than any given ‘stronger’ protein. We just keep digging. The more soil that what toss up to the surface, the questions become more pedantic. At a certain point it becomes unescapable: any construction of scientific Truth, even qualitative in nature, is bootstrapped by human definitions. Just like the scientific Truth of planets orbiting the sun relies on the subjective definition of a planet, humans define HIF1α and HIF2α. They define humans and non-humans. They define binding strength. Humans are forced to operationalize the world that they are studying. It becomes impossible for humans to know Truth except in relationship to our own definitions. Any Truth that exists in relation to subjective definitions becomes truth.
Humans define Truth, and we can also kill it. I have defined Truth as an immutable law, something indelible and existing independently of humanity. First describing quantitative observations, such as the binding rate of two proteins, it becomes clear that humans cannot know with certainty any measured value. Even as machines become more precise, there will always be random variations in instrumental measurements. But, a discussion of qualitative Truths, those fancy principles like gravity and evolution, are in need of dissection. It is obvious that any absolute Truth requires an infinitely narrow definition. Operational parameters are necessary to define a context for the Truth. Yet, this is also makes it impossible for humans to know Truth.