Every piece of data, however incomplete, helps you put together a picture of a problem that’s too complex to perform simple experiments on. Every model teaches you something. Sometimes it teaches you by making you figure out why it doesn’t work too well. Sometimes it works well in one place and not in another ,and you learn a lot by figuring out the boundary conditions. But every experiment has plusses and minuses.
As Clint Eastwood so famously said, you have to know your limitations.
In a speech at Digital Biota2, Douglas Adams (of Hitchhiker’s Guide to the Galaxy fame) described the historic difficulty (in the golden age of Physics between Newton and Einstein) of grappling with the science of life itself:
I can imagine Newton sitting down and working out his laws of motion and figuring out the way the Universe works and with him, a cat wandering around. The reason we had no idea how cats worked was because, since Newton, we had proceeded by the very simple principle that essentially, to see how things work, we took them apart. If you try and take a cat apart to see how it works, the first thing you have in your hands is a non-working cat.
Actually, even in Newton’s day you could open up a cat to see how at least part of it works, and then put it back together again into a functioning cat. It was just a tricky proposition, and you had (and still have) to open the cat up just right. Very few humans had that skill in Newton’s day, and most of them were crude animal doctors, not scientists.
Today biology is still tough to study. Of course, your modern vet takes female cats and dogs at least partially apart and puts them back together with amazing regularity. But you still can’t dig into an animal’s brain willy-nilly and expect a fully functioning animal afterwards. That’s why they invented microelectrodes that can be inserted into a small hole in a rat skull. Biology is a weird space to experiment in (speaking as a Chemist).
Even a “simple” cell is a mass of interconnected systems, and it’s hard to study only one part of it without changing how other parts function, which them in turn come back and change the function of whatever it was you were studying in the first place. This is the very definition of a feedback loop, and a simple cell is a morass of hundreds to thousands of feedback loops. Isolating a variable (i.e. keeping everything else constant and changing only ONE thing, to study the function of that one thing) as we do in Chemistry and Physics, is close to impossible. So we triangulate.
Each experimental method we use has limitations, and we use multiple methods to compensate for the limitations of any given experiment. The major classes of experiment in biology are in silico, which I discussed ad nauseum in Part One, in vitro, which are experiments in petri dishes on cells or isolated products of cells (eg. enzymes and receptors), and in vivo, or experiments in living systems such as rats or people.
Let’s go back to in silico for just a moment (we’ll quickly move on, I promise). A good example of triangulation in biology is looking for the correct shape (or conformation) of protein (enzyme, receptor, what have you) and the molecule it interacts with (its “substrate”).
One could, as the initial quote in Part One (the one I spent the entirety of Part One beating up on) suggested, churn out a conformation by brute force calculations. The problem is, as the Air Force report I linked to indicated, this is almost certain to be wrong. Why? Well, how many water molecules are hiding inside that space in the coiled up carbon chain? Are you sure it’s five, or is it six? Do you feel lucky, punk? Well, do ya?
On top of this, there may be more than one solution to the equations. In fact, with a polymer, there is certain to be. We call the solutions energy minima – in other words, proteins are lazy. Proteins fold in such a way as to minimize the energy needed to keep them in that state. They never stand when they can sit, and never sit when they can lie down. But, to take the analogy a bit too far, they can lay on their side, on their back, or even on their front. Which one is the preferred one? You can’t just go picking the absolute lowest energy state that the computer spits out when the differences in energy between several minima are low, because proteins don’t sleep alone. Intracellular conditions and interactions with their substrates can change the conformation. Furthermore, did you remember that bit about hydrogen bonds that we’re just starting to learn about? Did you include that 1.2 kcal / mole term? No? That might tip the balance between one conformation an another. And there’s more than just that term lurking in the swamp we haven’t explored yet.
So what do we do with our current tools and state of knowledge? Let’s take a concrete example. The greatest tool for figuring out the geometric conformation of chemicals, organic or not, is the in vitro technique of X-ray crystallography. Looking at the pattern of X-rays bouncing off of the electrons in a molecule can tell us the exact shape of a molecule. If you had to learn those bond angles for carbon and other compounds in high school chemistry, well, X-ray crystallography is how those angles were determined.
But X-ray experiments have limitations. Those measurements have to be done in a vacuum, so you take a sample of your protein, maybe even a sample of the protein bound to its substrate, and freeze it. Then you stick it in the sample chamber of your X-ray machine and pump out all the air. Do you think that the conformation of the frozen protein is always the same minimum energy conformation as it assumes in body-temperature water?
The conditions of crystallisation are often assumed to be absolutely relevant to the conditions of the biological assay. However, changes in buffer constituents, pH and crystallisation conditions can have a profound effect on the conformation of both ligands and proteins. For instance, the severe acute respiratory syndrome (SARS) coronavirus main protease was crystallised at different pH values and in complex with a specific inhibitor. The structures revealed substantial pH-dependent conformational changes and an unexpected mode of binding for the substrate-analogue inhibitor . At a pH value of 6 the structure of the monomers in the homodimer differs (one being in the active and the other in the inactive conformation) and the inhibitor binds in a different mode to each monomer.
That doesn’t mean that X-ray structures are useless, just that, as with mathematical models, they have to be used judiciously. Derek again:
when your X-ray data and your structure-activity data seem to diverge, it’s often a sign that you don’t understand some key points about the thermodynamics of binding. (An X-ray is a static picture, and says nothing about what energetic tradeoffs were made along the way). Instead of an irritating disconnect or distraction, it should be looked at as a chance to find out what’s really going on. . .
This is a key point. We scientists coming at the problems of biology from different angles are pretty collegial as scientists go. We work together to use our varied expertise to get a whole panorama from our individual snapshots.
If you read that paper I linked to (in the word “no”) from Drug Discovery Today, what you find is an X-ray crystallographer honestly and candidly discussing the limitations of his field and asking modelers to help us all get to goal, that goal being an accurate understanding of protein – ligand biding. You’ll find interspersed throughout that paper, advice for or requests for help from, modelers:
As a matter of fact, modellers can make an important contribution themselves to the structure determination of protein–ligand complexes. First, their knowledge of organic chemistry and stereochemistry of small molecules is often better than that of a protein crystallographer, so the modellers could help formulate appropriate refinement dictionaries with proper restraints and target values for bond lengths, etc. Second, their knowledge of, and eye for, judging protein–ligand interactions could help the crystallographer, by proposing ligand poses that both fit the electron density and make good sense in terms of protein–ligand interactions.
This is a great example of the back-and-forth between in vitro experimenters and in silico modelers, of how true professional relations occur in real life.
So we come to another important point for your bozo filter. Real scientists work together. Claiming to be a modern day Galileo is worth 40 points on John Baez's crank test. When someone is antagonistic to the entire scientific establishment, that's a major red flag.
In biology, techniques also work together. As I stated earlier, no single experiemnt can isolate all the variables in a biological system, so we use incomplete data from lots of different experiments conducted by different means to get at the truth slowly and circumspectly.
For example, beyond the basic level of protein structure, we come to the other major piece of biological triangulation, the correlation between in vitro and in vivo studies. A good example is that, for the designer of drugs in pill form, the first question one asks is “how will this get from the gut to the bloodstream?”.
In order to answer this question, the in vitro modelers came up with a line of human colon cancer cells that one could make a membrane out of in a petri dish. The theory was that these CACO-2 membranes would simulate the gauntlet that a small molecule drug has to run in order to end up in your bloodstream.
Unfortunately pretty much the only thing that the CACO-2 experiment tells you is how well the target compound passes through a membrane of CACO-2 cells. It has low predictive power for actual animal and human gut absorption. Which is unfortunate, because it’s cheap to do, and lots of managers still like to see the data, even if it is suspect (bad data is often worse than no data, a fact that non-scientists seem to have a hard time grasping).
Over at Org Prep Daily a few years ago, the chemist going by the moniker Milkshake pointed out rather forcefully that the best experiment to conduct is still the one where you feed a living rat the compound and see how fast the drug gets into, and gets washed out of, the bloodstream (this experiment is called pharmacokinetics, or PK, by biologists):
10. Ignore Caco-2 and do rodent PK tests instead, use human plasma and whole blood
Caco-2 permeability model is useless. Oral absorbtion/brain penetration tests in rodent should be done early in the project.
And this demonstrates a key point in medical research. That in vivo data trumps all. And data in the system you are interested in trumps data in another animal. So, for instance, human data trumps rat data if you are looking for a human medicine. And data in rats trumps crappy CACO-2 experiments. But in vivo experiments such as CACO-2 that are totally useless are few and far between. Most of the time the experiments in the major arenas of biological triangulation work together, despite their flaws, each giving up a little clue. We scientists working in those major arenas learn from each other. I tend to sit on the modeling side of things (A P-Chemist on the modeling side? Who’d a thunk it?). But, remember: GIGO. A model’s only as good as the sequence, structure and other data that goes into it.
So, for example, when you’re trying to figure out whether something you’re going to give humans might cause cancer, there are a few techniques to use to give you a clue. The in vitro techniques include the Ames and micronucleus tests, in which you bung some compound into some cells and see if their DNA gets damaged.
There was some back and forth between in vivo and in vitro experiments for these tests however, and you don’t simply throw compound and cells together, you need another piece of the puzzle:
Most of the cell lines and bacteria that are used for routine testing lack certain enzymes that metabolise foreign substances. A rat liver extract is therefore added to such test systems. This fraction, from which larger cell fragments, the nucleus, and the mitochondria have been removed, contains metabolic enzymes such as cytochrome P450-dependent monooxygenases that stimulate in-vivo metabolic processes.
“Simulate in vivo processes”. Because we look at the discrepancy between in vivo and in vitro data and try to reconcile it. What we don’t do is scrap in vitro or in silico completely because they don’t agree with in vivo. There’s a lot of stuff going on in a living organism, and in vivo and in silico techniques can go a long way towards isolating the extraneous stuff so you can look at the effects of a single variable. Reductionism is still a goal, even if it’s impossible to entirely achieve in biology.
In carcinogenicity studies, in vitro results are coupled with in vivo. A positive Ames or micronucleus test is cause to take some caution in experimentation, but you have take lots of drugs in your life that come up positive on those tests. The gold standard for genotoxicology is the multi-year rat carcinogenicity test, and a compound that passes that one but flunks the in vivo tests will be allowed in humans, but the FDA will require that the in vivo tests be noted in the drug’s label.
This whole area of research, looking at in vitro and even in silico evidence and trying to reconcile that data with in vivo data, is called translational medicine.
Translational medicine is the only reliable road to the future we have right now, but it is a long and winding one:
in essence, a lot of “translational” research takes close to two decades to bear fruit, and it’s fairly uncommon for it to take less than a decade. Moreover, as Dr. Ionnidis points out, less than 5% of promising claims based on basic science ever come to fruition as actual therapies. In other words, translational research is hard. Few promising ideas make it to therapies, and it takes a long time for those that do.
This is how it works in real life, folks. We’re all spelunkers wandering in natures caves, and almost everyone with a flashlight can help. Anyone claiming to be a lone genius with answers no one else has should be trapped in your bozo filter until you evaluate the claims more carefully. Anyone trashing an entire area of science outside of their field should stay trapped in your bozo filter until you can speak with an expert.