Start Here: How Your Body Works, What Genes Are, and Why I Look at Health the Way I Do
2026-11-03Same page, two depths. Switch any time. Written from a research and education perspective: a model of how bodies work and why they go wrong, not medical advice.
Welcome. If you found this site because you are tired, in pain, reacting to foods, dizzy when you stand, or just confused about a gene test somebody ran, this is the page to read first. Everything else on the site assumes you know the things on this page. It is long, because it has to be. You do not have to read it all at once. Each section stands on its own, and there is a glossary at the bottom.
Here is the big idea of the whole site in three sentences. Your body is a factory with many assembly lines. Your genes are the instruction manual for the workers on those lines. When you get sick, it is almost always because an assembly line is backed up, and the way to fix it is to find the backup and clear it, instead of just quieting the alarm bell that is ringing because of it. I call this way of thinking metabolic pathway medicine, and the last part of this page explains why it works.
One more thing before we start. I am a scientist, not a doctor. I do research. I study how bodies work, I build models of what might be going wrong in a particular person, and I explain what the science says would help if the model is right. I do not diagnose, treat, or cure anything. The point of this site is to make you smart enough to have a better conversation with the people who do.
Part 1: Your body is a factory
Think of your body as a factory that never closes. It takes in raw materials (food, water, air), turns them into energy and parts, builds and repairs itself, and takes out the trash. Every department has a job. When one department falls behind, the others have to cover for it, and that is usually where symptoms come from.
Here are the departments and what they actually do.
The gut: receiving and sorting. Your stomach and intestines break food into pieces small enough to absorb: proteins into amino acids, carbs into sugars, fats into fatty acids, plus vitamins and minerals. The gut is also a wall. It is supposed to let nutrients in and keep bacteria and junk out. And it is home to trillions of bacteria that make vitamins, train your immune system, and produce chemicals (including histamine) that affect the whole body. A lot of problems start here, because if the receiving department is not sorting properly, every other department is short on parts.
The liver: the chemical plant. The liver does more different jobs than any other organ. It stores sugar and releases it between meals. It makes cholesterol and packages fat to ship around the body. It makes most of the proteins in your blood. And it is the main trash-processing plant: it takes hormones you are done with, medications, chemicals from food and the environment, and the junk from your own metabolism, and it changes them step by step into forms your body can get rid of. This takes a lot of specific nutrients, and when the liver runs short, the trash backs up.
The kidneys: the filters. They clean the blood, keep your salt and water in balance, control your blood pressure, and turn on vitamin D. They also make sugar during a long fast and help control how acidic your blood is.
The pancreas: the thermostat for blood sugar. It makes insulin, which tells cells to take in sugar, and glucagon, which tells the liver to release it. It also makes the enzymes that digest your food.
Muscle: the biggest sugar sponge you own. Muscle takes sugar out of your blood when you move, even without insulin. It is also your protein bank: when you are sick or not eating enough, your body breaks muscle down for parts. More muscle means better blood sugar and a bigger reserve.
Fat: the warehouse, and also a gland. Fat stores energy, which is its honest job. It also stores chemicals your liver has not had time to process, which is why your body builds more of it when the trash is piling up. And fat tissue sends out hormone signals that affect appetite, inflammation, and how well insulin works. Belly fat sends out more of the inflammatory ones.
The heart and blood vessels: the delivery trucks and the roads. Every cell needs oxygen and nutrients delivered and waste picked up. The lining of your blood vessels is a living tissue that decides when to open and close the roads. When it is irritated, deliveries slow down everywhere.
The lungs: oxygen in, carbon dioxide out. Oxygen is what your power plants burn fuel with. Carbon dioxide is the exhaust.
The brain and nerves: the control room and the wiring. The brain runs on sugar and ketones and uses a quarter of your energy. The nerves carry signals both ways, and the smallest nerves run things you never think about: heart rate, blood vessels, gut movement, sweat. These tiny nerves are expensive to maintain and are the first to go when the factory is short on supplies.
The thyroid and adrenals: the thermostat and the alarm system. The thyroid sets how fast every department runs. The adrenals make cortisol, which handles stress, keeps inflammation in check, and keeps your blood sugar up overnight, and adrenaline, which is your emergency button.
The immune system: security. It patrols for infection and cleans up damage. Mast cells are one of its branches: they live in your skin, gut, and around your blood vessels and release histamine and other chemicals when they sense a threat. In a lot of people they are jumpy and fire when they should not.
Skin, bones, and connective tissue: the walls and the scaffolding. Connective tissue (collagen and its friends) holds everything in place. Building it uses a surprising amount of raw material, and some people are born with a stretchier version that needs constant repair.
And inside every cell: the power plants. Mitochondria are tiny structures inside almost every cell that burn sugar and fat with oxygen to make energy. They also throw off sparks (free radicals) when they work, and the body has a cleanup crew for those sparks. When mitochondria are damaged, you feel it as fatigue, brain fog, and crashing after exertion, because every department just lost power.
The one thing to take from this section: no department works alone. The gut feeds the liver. The liver feeds the mitochondria. The mitochondria power the nerves. The nerves run the gut. When one is in trouble, the trouble travels.
Part 2: What genes are
Every cell in your body contains a complete instruction manual called DNA. It is written in a four-letter alphabet (A, T, C, G), and it is about three billion letters long. The manual is split into chapters called chromosomes. Within the chapters, the parts that actually say how to build something are called genes. You have about 20,000 of them.
Each gene is a recipe for one protein. Proteins are the workers of the factory. Some are structural (collagen is a protein). Some carry things (hemoglobin carries oxygen). And a huge number are enzymes: workers whose only job is to take one chemical and turn it into another chemical. Almost every step of metabolism is an enzyme doing one specific job, over and over, millions of times a second.
Here is the chain: a gene is a recipe, the cell reads the recipe and builds the protein, the protein does its job. If the recipe has a typo, the protein might be built slightly wrong, or not at all.
You have two copies of almost every gene, one from each parent. That matters a lot, as you will see.
One more thing people get wrong. Your genes do not change over your life. But how much a gene gets used does change, depending on what you eat, how you sleep, what you are exposed to, and what else is going on in the body. Think of the recipe as fixed, but the kitchen deciding how many times a day to cook it.
Part 3: Why some gene changes do not matter and some do
A gene change is called a variant. Sometimes people say mutation, which sounds scarier than it usually is. Everyone has millions of variants. Most do nothing.
Think of the recipe for a cake. Here are the kinds of typos it could have:
Typo that does not change anything. "Add 2 cups of flower" instead of "flour." You still know what to do. Most variants are like this. The protein comes out the same, or close enough.
Typo that changes one word a little. "Bake at 325" instead of "350." The cake still comes out, but it takes longer. This is the most important kind for this site. The enzyme still works, just more slowly, or it needs more of its helper to work at full speed. We call these slow variants, and they are common. The famous one, MTHFR C677T, makes an enzyme that works at about two-thirds speed if you have one copy and about one-third speed if you have two. That is not a broken gene. It is a slow one.
Typo that deletes a whole line. "Add 2 cups of" and then nothing. The protein cannot be built at all. These are rare, they are the ones that cause serious inherited diseases, and they are what most genetic testing in hospitals is looking for.
Now, why the same typo can be a big deal in one person and nothing in another:
You have two copies. If one copy has a bad typo and the other is fine, the good copy often covers. Many serious conditions only show up when both copies are broken. This is why a lot of people are carriers of something and never know it.
It depends on what the gene does. A slow enzyme in a pathway with a detour is a shrug. A slow enzyme at a step with no detour, in a tissue that needs that product constantly, is a problem.
It depends on what else is going on. This is the part most people miss, and it is the whole reason this site exists. A slow variant is a dial turned down a little. Most of the time your body has enough spare capacity that nothing happens. But add a diet short on the vitamin that enzyme needs, plus a stretch of stress or infection that raises demand, plus a second slow variant on the same pathway, and now the dial matters. The gene did not change. The load did. This is why the same MTHFR variant is "nothing" in one person and part of the story in another.
Common versus rare. If a variant is common (lots of people have it), it almost certainly is not a disease-causing broken switch, because evolution would have gotten rid of it. Common variants are dials. Rare variants are more likely to be switches. Consumer DNA tests (23andMe, AncestryDNA) mostly look at common variants, which means they are mostly looking at dials. That is useful, but it is a different thing from the hospital test looking for switches.
And a warning about consumer DNA files. The raw data from those kits has errors. Studies that double-checked rare "disease" variants found in consumer raw data found that a large share, sometimes around 40 percent, were simply wrong. If a consumer file says you have a rare serious variant, it has to be confirmed by a real clinical lab before anyone acts on it. The common variants are much more reliable.
Part 4: ClinVar, and why "pathogenic" does not mean what you think
When you run your DNA through a tool, many of them show you a label next to each variant: benign, likely benign, uncertain, likely pathogenic, pathogenic. Most of those labels come from a public database called ClinVar.
ClinVar is a shared notebook. Laboratories and researchers around the world write in what they think a variant means and why. It is enormously useful. It is also just a notebook, and it has limits you need to know about.
It is opinions, not facts. Each label is one lab's judgment. Different labs can and do write different opinions about the same variant. ClinVar shows this as "conflicting interpretations," and it is common. There is a star rating next to each entry that tells you how much agreement there is; one star means one lab said so, and that is a lot of the database.
"Uncertain significance" means exactly that. Nobody knows. It does not mean "probably bad." Most rare variants that get looked at land here, because most variants have not been studied.
Labels get changed. A meaningful share of variants get reclassified over the years, usually from pathogenic or uncertain toward benign, as more people get sequenced and the variant turns out to be common and harmless somewhere.
It is built for rare disease. ClinVar answers the question "does this variant cause a genetic disease?" It was not built to answer "does this common slow variant make my enzyme run at two-thirds speed?" So the dials, the variants that matter most for the kind of thing this site is about, are usually labeled benign in ClinVar, or are not in it at all. "Benign" there means "does not cause an inherited disease on its own." It does not mean "does nothing."
It leans heavily on people of European ancestry. A variant that is rare in Europeans and common in West Africans can get labeled pathogenic by mistake, simply because the lab had not seen it in enough people. This is improving but it is not fixed.
Reading a label is not the same as reading a person. A label is the starting point of a question, not the end of one. The question is always: what does this gene do, is the variant a dial or a switch, do you have one copy or two, what else is going on in that pathway, and does anything in your blood work or symptoms match?
Part 5: In silico scores, the computer's best guess
For a lot of variants, no human has written an opinion. So tools use computer predictions instead. "In silico" just means "in a computer." You will see names like CADD, REVEL, PolyPhen, SIFT, and AlphaMissense.
What they do: they look at a variant and guess how likely it is to damage the protein. They use things like: is this spot in the protein the same in every animal from fish to humans (if so, changing it is probably bad), does the swap change the shape or charge of the protein, and does it look like variants already known to be harmful.
What they are good at: ranking. If you have ten unknown variants, the scores help you decide which ones to look at first.
What they are bad at: being right about any one variant. A high score means "this looks like the kind of change that is often damaging." It is a tilt, not a verdict. Clinical guidelines treat even a strong in silico score as weak evidence on its own. And the scores are built on the same European-heavy data as everything else.
What they cannot do: tell you whether a slow dial matters in you. They do not know your diet, your other variants, your labs, or your symptoms.
So when you see a scary-looking score on a consumer report, read it as "worth a closer look," not "something is wrong with me."
Part 6: What a biochemical pathway is
This is the most important section on the page, and it is simpler than it sounds.
A pathway is an assembly line. Raw material comes in one end, each worker (enzyme) does one change to it, and a finished product comes out the other end. The thing being worked on at each step is called a substrate, and what comes out is the product, which becomes the substrate for the next worker.
Most workers need tools. The tools are vitamins and minerals. We call them cofactors. The enzyme MTHFR, for example, cannot work without vitamin B2. Hundreds of enzymes need magnesium. Dozens need zinc. The enzyme that makes energy in the mitochondria needs iron, copper, B1, B2, B3, and more. A worker without a tool is a worker who stands there.
Here are the ideas you need to reason about pathways:
Bottlenecks. An assembly line only goes as fast as its slowest worker. If one enzyme is slow (a variant) or missing its tool (a nutrient shortage), everything behind it piles up and everything after it runs short. The pileup and the shortage are both symptoms. Example: if the enzyme that breaks down histamine is slow, histamine piles up (flushing, hives, racing heart), and the methyl groups it was supposed to pass along run short somewhere else.
Lines share workers and tools. The same magnesium is needed by three hundred different enzymes. The same methyl group can be spent clearing histamine, or clearing adrenaline, or building a nerve membrane, or making creatine for your muscles. When one line is working overtime, it steals tools from the others. This is why a body under stress runs short of many things at once.
Lines can be turned up and down. Your body senses when there is too much product and slows the line (feedback). It senses when there is too little and speeds it up, if it can. This is important: supplying the finished product of a line (eating creatine instead of making it, for instance) lets the body turn that line down and spend its tools elsewhere.
The lines live in different rooms. Some reactions happen in the mitochondria, some in the main part of the cell, some in the liver only, some in the gut lining. A nutrient that cannot get into the right room is not useful there.
Most lines have spare capacity. This is the piece that explains why a slow variant is usually harmless. Most enzymes run far below their maximum, so a two-thirds-speed enzyme is enough. The problem comes when demand rises past what the slow worker can keep up with.
Part 7: How re-routes work
Here is the good news about assembly lines: a well-designed factory has detours. When a worker is slow, there is often another way to get the same product. Finding and using those detours is most of what this site is about.
The kinds of detour:
Give the slow worker more raw material. A slow enzyme often works fine if you flood it with its substrate or its tool. A two-thirds-speed enzyme with twice as much B2 around often gets close to normal. This is the oldest trick in nutritional biochemistry, and it is the reason "give extra of the vitamin that enzyme needs" is not woo; it is enzyme kinetics.
Use the back road. In the methylation cycle, there are two ways to re-make the key molecule (methionine): the main road uses the folate enzymes and B12, and the back road uses a nutrient called betaine (also called TMG) that comes from beets, spinach, and your own choline. If the main road is slow, the back road can carry more traffic. That is why betaine and choline show up so often in my writing.
Enter the line further down. If the slow step is early in a line, you can sometimes supply the thing it was supposed to make, and the rest of the line runs normally. Folinic acid enters the folate cycle one step past where folic acid has to be processed. Creatine, eaten, lets you skip making it. Acetyl-L-carnitine supplies carnitine without needing the five-step process that builds it. Hydroxocobalamin supplies B12 in a form the cell can convert to whichever version it needs.
Take the load off. If the problem is that a line is overwhelmed, the fastest fix is often to stop sending so much to it. A four-week low-histamine diet does not fix a slow histamine enzyme. It gives the slow enzyme less to do. Cutting the chemicals coming into your house does the same thing for the liver's trash line.
Turn a competing line down. Because lines share tools, you can free up tools by supplying the product of one line so it stops hogging them. Creatine and phosphatidylcholine are my favorite example: making those two uses up more than half of your methyl groups. Eat them, and the methyl groups go to clearing histamine and repairing nerves instead.
Clear the exits. A lot of lines end in "and then it leaves the body," through bile, stool, urine, sweat, or breath. If the exits are blocked (constipation, a sluggish liver, no sweating), the product backs up and the line stalls, no matter how well the workers are doing. This is why so much of the practical advice on this site is boringly about fiber, water, bowel movements, and sweating.
Re-routes are not magic. They work when the detour actually exists in human biology, when the thing you are supplying can get to the right room, and when you are supplying it in a form the body can use. That is what the research part of my work is: checking, for a particular person, which detours are real.
Part 8: Why metabolic pathway medicine works
I call the way I think about health metabolic pathway medicine, a term I coined because nothing else described it. Here is what it means and why I think it works better than treating symptoms one at a time.
Symptoms are alarm bells, not the fire. Fatigue, pain, flushing, dizziness, high cholesterol, high blood sugar, anxiety: each one is the body telling you an assembly line is backed up. The standard approach is to quiet the bell. An antihistamine quiets the histamine bell. A statin quiets the cholesterol bell. Sometimes quieting the bell is exactly right, especially as a bridge. But the fire is still burning, and usually another bell starts ringing somewhere else, because the lines are connected.
The body is adapting, not failing. This is the belief underneath everything I write. When your body stores toxins in fat, raises cholesterol to run on fat because sugar stopped working, becomes insulin resistant to protect overstuffed cells, or races the heart because the leg vessels will not squeeze, it is doing something smart with the options it has. Disease is what it looks like when the body runs out of good options. That means you can usually turn it around by giving the body better options, and it means the symptom is a clue to which line is stuck, not an enemy to suppress.
Genes tell you where to look. Labs tell you if you found it. A slow variant tells me which line to check. It does not tell me that line is actually backed up in you. For that I look at the blood work that measures the pileup directly: methylmalonic acid for B12, homocysteine for methylation, ferritin for iron, a prostaglandin metabolite for mast cells, and so on. Genes are the map. Labs are the traffic report. Symptoms are the destination. You need all three to find the bottleneck, and acting on genes alone is how people end up taking twenty supplements and feeling worse.
Most of what matters is dials, not switches. The hospital model of genetics looks for broken switches: one gene, one rare disease. That model is correct and important, and it describes a small number of people. Most chronic illness is three or four dials turned down a little, in a person whose diet, stress, infections, or chemical exposure turned the demand up. That picture does not show up on a hospital gene test, and it does not show up on a standard blood panel either, because every single value is "within range." It shows up when you look at the pattern.
It works because enzymes obey chemistry. A slow enzyme really does speed up with more cofactor. A back road really does carry more traffic when you supply its fuel. A line really does turn down when you supply its product. These are not beliefs; they are measurable, and the best part is that they are checkable in you. If my model says your B12 is functionally low and your homocysteine is high, and we fix the inputs, your homocysteine should fall. If it does not, the model was wrong and we go back to the map. A good model makes predictions you can test.
Order matters. When several lines are backed up at once, the sequence you clear them in is most of the skill. Calm the jumpy immune cells and take the load off before you ask the factory to rebuild. Rebuild the gut wall before you push nutrients through it. Spare the methylation budget before you add methyl donors. Fix sleep before you add exercise. Go after deep drivers (old infections, biofilms, stored toxins) last, and gently, because stirring them up in an unstable body makes everything worse. Nearly every "I tried that and it made me worse" story I hear is a right ingredient added in the wrong order.
Food first, one thing at a time. The lowest-risk way to supply a tool is to eat it. Supplements come after, at the lowest dose that works, added one at a time so you know what did what. This is slower than a handful of pills. It is also how you learn what your body actually responds to.
What this is not. It is not a replacement for medical care. It is not diagnosis. It is a way of reasoning about why a body might be doing what it is doing, and what the published science says would address that, so that you and your doctor can make better decisions than "your labs are normal, try to exercise." I am a researcher. I build the model and show my work. You and your healthcare providers decide what to do with it.
How to use the rest of this site
Every post follows the same shape: what is happening, why (the model), what would test the model (labs), and what the model predicts would help (diet, supplements in order, lifestyle), with a plain version and a research version you can switch between at the top. If you are new, read the POTS post here or the how people get sick post here next; they show the method on a real picture. If you have a consumer DNA file, the gene reports on this site apply this way of thinking to one body system at a time.
The reference pages
Four pages on this site are built to be looked things up in rather than read through: the Nutrient Guide here (every vitamin, mineral, fat, and amino acid, what running low looks like, and how to test for it), the Lab Guide here (the same ground organized by test, with the ranges and the prep), the Medications table here (what common drugs drain and what to watch), and the Food-Sources table here (where each nutrient actually comes from, with amounts). Alongside them: the Genes and the Nutrients table here (which variants change the need for which nutrients), the Before Your Appointment checklist here (the Lab Guide on one printable page), the three trial diets (low-histamine trial here, low-FODMAP trial here, substrate separation here), the Small Fish and Mercury page here, and the full site Glossary here. Between them they answer most of the "but what do I actually do" questions the mechanism posts raise.
Glossary
The short version. The full site glossary, with an anchor for every term, is here.
- Amino acid: the building blocks of protein. There are twenty.
- Cofactor: a vitamin or mineral an enzyme needs to work. The tool.
- DNA: the instruction manual in every cell.
- Enzyme: a protein that changes one chemical into another. The worker.
- Gene: a recipe for one protein.
- Heterozygous / homozygous: one copy of a variant / two copies.
- In silico score: a computer's guess at how damaging a variant is.
- Metabolism: all the chemical reactions that keep you alive, taken together.
- Methylation: the process of handing out small chemical tags (methyl groups) for hundreds of jobs. Runs on folate, B12, B2, B6, and magnesium.
- Mitochondria: the power plants inside cells.
- Pathway: an assembly line of enzymes.
- Pathogenic / benign / uncertain: labels from ClinVar meaning "causes a genetic disease," "does not," and "nobody knows."
- Product / substrate: what comes out of an enzyme / what goes in.
- ROS (reactive oxygen species): the sparks thrown off by metabolism. Free radicals.
- SNP: a single-letter variant. Most consumer DNA data is SNPs.
- Variant: any change in a gene's spelling. A dial or a switch.
Educational content written from a research perspective, not medical advice. I am a PhD scientist, not a physician, and I do not diagnose, treat, cure, or prevent disease. Please take anything you want to try to your own healthcare provider.
What this page is for
Most of the posts on this site do the same thing: they take a presentation you see in clinic (orthostatic intolerance, widespread pain, fatigue with normal labs, a patient who reacts to everything), propose a pathophysiology for it, name the labs that would test that proposal, and lay out what the published literature suggests would address each piece if the proposal is right. To read those posts comfortably, you need a handful of ideas from metabolism and genetics that medical training covers quickly in the preclinical years and then rarely asks you to use again.
This page walks through those ideas slowly, and it does it through conditions you already manage. You know pernicious anemia, scurvy, phenylketonuria, hemochromatosis, warfarin dosing, and metformin-induced B12 deficiency. Every concept here is one of those, generalized. By the end, the rest of the site should read as an extension of medicine you already practice rather than a separate discipline.
A note on my role. I am a PhD in cellular and molecular nutrition with a background in clinical genomics. I do research consulting: for a given person I study their symptoms and genotype, build a model of what may be going on, work out what would address it if the model is right, and explain the reasoning. I do not diagnose, treat, cure, or prevent disease. The clinician who does is you, and this page is meant to make your job easier, not to second-guess it.
Part 1: Organs as metabolic workers
You already think of organs by what they do clinically. This section adds one layer: what each one does chemically, and which of those jobs tends to fail in the presentations this site covers. Each paragraph starts from something familiar.
The gut, and why PPIs cause more than one deficiency. You know that long-term proton pump inhibitors lower B12. The reason is that protein-bound B12 in food needs acid and pepsin to be released before intrinsic factor can bind it. The same acid is what keeps non-heme iron soluble enough to absorb, and the magnesium channel in the small bowel (TRPM6) works best in an acidic environment, which is why the FDA added a hypomagnesemia warning to PPIs in 2011. So one intervention, acid suppression, quietly lowers three nutrients that turn out to matter a great deal later on this page. The gut has two other jobs worth naming. It is a barrier: a single layer of cells held together by tight junctions, keeping bacterial endotoxin out of the bloodstream. When that barrier is leaky, the liver and fat tissue see low-grade endotoxin all day, which you recognize clinically as the inflammatory tone of metabolic syndrome. And the gut bacteria are themselves a metabolic organ: they make vitamin K and short-chain fatty acids, they make histamine, and they make an enzyme (beta-glucuronidase) that unwraps the estrogen and drug conjugates the liver sent out for disposal and sends them back into circulation. Enterohepatic recirculation is the reason some drugs have a second peak; the same mechanism recycles estrogen.
The liver, and why it is more than a lab value. You order LFTs to see if the liver is injured. Here we care about what the liver is doing when it is not injured. It stores and releases glucose (which is why fasting glucose is a liver number before it is a pancreas number). It packages fat into VLDL to ship it out, and that packaging requires phosphatidylcholine; a diet short on choline produces fatty liver even in the absence of alcohol or obesity, which is a result you may remember from TPN patients before choline was added. It makes cholesterol, and the enzyme statins block (HMG-CoA reductase) sits at the same branch point that makes coenzyme Q10, which is why statin myalgia is a CoQ10 story. And it is the body's chemistry plant for clearing things: drugs, hormones, and environmental chemicals go through phase I (the cytochrome P450 enzymes you know from drug interactions), phase II (where the molecule is tagged with glucuronic acid, sulfate, glutathione, a methyl group, an acetyl group, or glycine so it becomes water soluble), and phase III (pumped into bile). Acetaminophen toxicity is a phase II story you know well: glutathione runs out, the toxic intermediate accumulates, and N-acetylcysteine works because it rebuilds glutathione. Hold onto that example, because the same logic applies to a lot of what follows. Every one of those phase II tags has to come from somewhere, and the supply is shared with other jobs.
The kidneys, beyond creatinine. They activate vitamin D (the 1-alpha-hydroxylation step you know from CKD), they make about a third of fasting glucose, and they decide how much magnesium you keep. Loop diuretics, thiazides, and alcohol all waste magnesium at the kidney. Low magnesium matters more than it looks, because magnesium is required by every enzyme that uses ATP, which is most of them.
The pancreas, and the ten years before the diagnosis. You know that beta-cell function is already about half gone when type 2 diabetes is diagnosed. The point for this site is that the process was measurable for a decade before that, as a rising fasting insulin, while glucose was still normal. Insulin itself is stored as a zinc crystal, and the channel that releases it is the one sulfonylureas act on.
Muscle, the glucose sink. When a patient walks after a meal and their postprandial glucose drops, that is muscle taking up glucose without needing insulin, through contraction. Muscle is also where most mitochondria live and where creatine buffers energy. Sarcopenia in an older patient is a loss of glucose disposal capacity, which is one reason resistance training lowers A1c.
Fat, which is also a gland and a warehouse. You know adipose makes leptin and adiponectin, and that visceral fat makes inflammatory cytokines. One more job: fat stores the fat-soluble chemicals the liver has not cleared yet. Organochlorine pesticides, PFAS, and plasticizers partition into adipose, which protects the brain and liver from them. This is why rapid weight loss sometimes makes people feel worse: the warehouse is being emptied back into circulation.
Blood vessels, and the enzyme that needs everything. The vessel lining makes nitric oxide to open vessels, through an enzyme (eNOS) that needs an unusually long list of cofactors at the same time, including a folate-related one called tetrahydrobiopterin (BH4). When BH4 is depleted by inflammation, the enzyme does something you do not want: it makes superoxide instead of nitric oxide. Oxidized vessels, less dilation. This is a piece of why folate status shows up in cardiovascular conversations.
Nerves, especially the small ones. Nerve conduction studies test large myelinated fibers. The small unmyelinated fibers that carry pain and temperature and that run the autonomic nervous system (heart rate, vessel tone, gut motility, sweating) are invisible to that test; you need a skin biopsy to count them. These fibers are up to a meter long, have to be maintained by transport from a cell body that far away, and have the least reserve of any neuron. They are the first casualty when nutrients, energy, or an inflammatory environment go wrong, and when they are lost the presentation is dysautonomia, burning pain, and GI dysmotility with a normal EMG. You already know the extreme versions: B12 deficiency neuropathy, diabetic autonomic neuropathy, and the autonomic failure of amyloidosis. The posts on this site argue that milder, multifactorial versions of the same lesion are common.
Thyroid and adrenals. You know sick euthyroid syndrome: a hospitalized patient with low T3, normal TSH, and high reverse T3, because illness suppresses the conversion of T4 to T3. The enzyme that does that conversion needs selenium, and it is suppressed by IL-6 and cortisol. The posts here argue that a chronic, outpatient version of sick euthyroid is common and under-recognized because TSH is normal. Cortisol you know as a stress hormone; its other job is keeping mast cells and inflammation quiet, and it is at its lowest between 2 and 4 am, which will matter in the sleep post.
Immune cells, particularly mast cells. You know mast cells from anaphylaxis and urticaria. They also sit around every blood vessel and next to every small nerve, and they release histamine, tryptase, prostaglandin D2, heparin, and nerve growth factor. Inside cells, histamine is cleared by an enzyme that spends a methyl group to do it, which connects the allergy world to the folate world in a way that will come up repeatedly.
Connective tissue. Scurvy is the clinical anchor here: without vitamin C, the enzyme that stabilizes collagen's triple helix fails, wounds reopen, and vessels bleed. Collagen also needs iron for that same enzyme, copper for the enzyme that cross-links the fibers, and an enormous amount of glycine (every third amino acid). Patients with lax connective tissue (hypermobile Ehlers-Danlos and its milder relatives) are repairing tissue constantly, and that repair competes for the same raw materials as everything else.
Mitochondria. You know them as the site of the Krebs cycle and the electron transport chain, and you know thiamine deficiency (Wernicke's, wet beriberi) as a mitochondrial disease at the pyruvate dehydrogenase step. The chain also needs riboflavin, niacin, iron, copper, and CoQ10. Statins lower CoQ10. Metformin acts on complex I. Mitochondria leak superoxide as a normal byproduct, and the body's cleanup for that (glutathione, superoxide dismutase, the selenium-dependent peroxidases) is the same system that acetaminophen overwhelms. When mitochondria underperform, the patient reports fatigue, exercise intolerance, and feeling flattened for days after exertion, years before any lab abnormality.
The reason to lay these out together is the chain between them. Stomach acid determines B12 and magnesium absorption. B12 and magnesium are needed to run the methylation cycle. The methylation cycle makes creatine and carnitine and phosphatidylcholine. Those run the mitochondria and the liver's fat export. The mitochondria power collagen synthesis and the small nerves. The small nerves run gut motility and acid secretion. It is a ring, and a lesion anywhere on it propagates.
Part 2: Genes, in the terms you already use
You use genetics every day without calling it that. Here is the vocabulary, built from the cases.
A gene is a recipe for one protein. Cystic fibrosis is one gene, CFTR, making one protein, a chloride channel. Sickle cell is one gene, HBB, making one protein, the beta chain of hemoglobin. Most of the genes that matter for this site encode enzymes: proteins whose only job is to convert one molecule into another. Phenylalanine hydroxylase converts phenylalanine to tyrosine; when it is missing, you have PKU.
You have two copies of each gene, one from each parent. For most enzymes, one working copy is enough, which is why PKU and CF are recessive: a carrier with one bad copy and one good copy has half the enzyme and no disease. Some genes are different. For structural proteins like collagen, a bad copy can poison the good one (dominant-negative), which is why osteogenesis imperfecta and most Ehlers-Danlos types are dominant. And for a few genes, half the protein is simply not enough (haploinsufficiency), which is why familial hypercholesterolemia from one bad LDL receptor copy produces disease.
A variant is any spelling difference in a gene. The word "mutation" is used for the same thing but carries more weight than it should. Everyone carries millions of variants. The important distinction is between a variant that breaks the protein and one that merely slows it.
Broken versus slowed: the distinction this whole site rests on. The CF variant F508del produces a channel that is misfolded and destroyed. That is a broken protein, what geneticists call a null or loss-of-function allele. Now consider the common variant in MTHFR called C677T. It changes one amino acid, and the enzyme it produces works, but it holds onto its riboflavin cofactor less tightly and falls apart more easily when warm. A person with two copies has roughly a third of normal enzyme activity; one copy, roughly two-thirds. That is not a broken enzyme. It is a slow one, what geneticists call a hypomorph. You already know this pattern from pharmacology: CYP2C9 poor metabolizers are not missing the enzyme, they have a slow version, and you lower their warfarin dose accordingly. TPMT intermediate metabolizers have a slow enzyme and need a lower azathioprine dose. Slow enzymes are the normal state of affairs; everyone carries dozens.
Why the same slow variant is nothing in one patient and matters in another. Think about G6PD deficiency. Most carriers are asymptomatic their whole lives. Give them primaquine, fava beans, or a bad infection, and they hemolyze. The enzyme was always slow; what changed was the load. This is the single most important idea on this page, so here it is stated generally: most enzymes run with spare capacity, a slow variant uses up some of that spare capacity, and symptoms appear only when demand rises past what the slow enzyme can supply. The demand can come from diet (not enough of the cofactor the enzyme needs), from illness or stress (more work for that pathway), from a drug (metformin lowering B12, PPIs lowering magnesium), or from a second slow variant on the same pathway. The gene did not change. The load did.
Common variants are dials; rare variants are switches. Here is a useful rule from population genetics: if a variant is carried by 10 or 30 percent of the population, it cannot be a highly penetrant cause of serious early disease, because selection would have removed it. So common variants, by their nature, are the slowed-enzyme kind. Rare variants are where the broken-protein kind live. This matters because consumer DNA tests (23andMe, AncestryDNA) read a few hundred thousand common positions, so they are almost entirely reporting dials. The clinical sequencing you order for a suspected Mendelian disorder is looking for switches. The two tests are doing different jobs, and much of the confusion patients bring you about "my MTHFR mutation" comes from reading a dial as if it were a switch.
A caution about consumer raw data. When a consumer file reports a rare disease-causing variant, treat it as unconfirmed. A 2018 study that sent consumer raw-data findings for clinical confirmation found that about 40 percent of the rare pathogenic calls were false positives, an artifact of the array technology. The common variants on those arrays are generally accurate. The rule is simple: common variants from consumer data are reasonable to think with; rare variants from consumer data get confirmed by a clinical lab before anyone acts.
Genes do not change; gene expression does. A patient's genotype is fixed. How much of a given enzyme their cells make changes with diet, inflammation, hormones, drugs, and sleep. Rifampin inducing CYP3A4 is the example you know. The same is true of most metabolic enzymes.
Part 3: ClinVar, and what "pathogenic" actually means
When a patient brings you a printout with variants labeled benign, pathogenic, or uncertain, those labels almost always trace back to ClinVar, the public database run by the NIH where laboratories deposit their interpretations. It is the right tool for a specific question, and it is worth knowing both what the question is and what it is not.
ClinVar is a collection of lab opinions, with a confidence rating. Each entry is one laboratory's interpretation and the evidence they used. The same variant can have several entries that disagree, and ClinVar labels this honestly as "conflicting interpretations." Next to each variant is a star rating. One star means one lab submitted with stated criteria; two means multiple labs agree; three means an expert panel reviewed it; four means it is in a practice guideline. A large share of the database is at one star.
The question it answers is "does this variant cause an inherited disease?" The classification framework (the 2015 ACMG/AMP guidelines) weighs population frequency, computational predictions, functional studies, family segregation, and whether the variant appeared de novo, and sorts variants into pathogenic, likely pathogenic, uncertain significance, likely benign, and benign. It was built for Mendelian disease, where the question is binary. It works well for that.
"Benign" means "does not cause the Mendelian disease," and nothing more. MTHFR C677T is classified benign in ClinVar. That is correct: it does not cause an inherited disorder. It also runs at a third of normal speed in a homozygote. Both statements are true, because the label is answering a different question than the one the patient is asking. The common slow variants that this site is mostly about are, almost without exception, labeled benign or are absent from ClinVar entirely. The label is not wrong; it is just not about enzyme speed.
"Uncertain significance" means uncertain. It is not a mild pathogenic. Clinical practice is not to act on a VUS alone, and most rare variants that get examined end up here, because most variants have never been studied in enough people.
Labels change. Between 10 and 25 percent of variants get reclassified over a span of years, most of them moving toward benign, because as more people from more backgrounds are sequenced, a variant that looked rare and alarming turns out to be common and harmless in a population the original lab had not sampled. This is also the ancestry problem: the reference data has been heavily European, and variants common in other populations have been misclassified as pathogenic as a result.
Pharmacogenomics lives elsewhere. For CYP2D6, CYP2C19, SLCO1B1, TPMT, DPYD, and the rest, the curated source is CPIC and PharmGKB, with their own nomenclature and dosing guidelines. ClinVar is not where to look for those. And for the common nutrient-related slow variants there is no curated clinical database at all, which is why this site grounds every claim in the enzyme's biochemistry and in a lab value you can order, rather than in a label.
Part 4: In silico scores, the computer's guess
For most rare variants no lab has written an opinion, so software fills the gap. The tools (SIFT, PolyPhen, CADD, REVEL, AlphaMissense, SpliceAI) estimate how likely a variant is to damage the protein by asking: is this amino acid the same in every species from fish to human (if so, changing it is probably bad)? Does the substitution change the size or charge of the residue? Does it sit in the active site? Does it resemble variants already known to be harmful?
Think of them the way you think of a D-dimer: useful for ranking and for ruling things in or out of further consideration, poor as a standalone verdict. The formal guidance (the ClinGen calibration published in 2022) allows even the strongest computational score to count as no more than supporting-to-moderate evidence. A high score means "this looks like the kind of change that is often damaging," not "this patient's enzyme is broken." And the tools are especially weak for the slow-variant class, because a 30 percent reduction in enzyme speed often does not look like anything to a conservation-based predictor. When a patient shows you a frightening score from a consumer report, the right reading is "worth a closer look," not "something is wrong."
Part 5: What a pathway is, built from PKU
You already know a pathway; you just know it as a disease. In PKU, phenylalanine should be converted to tyrosine by one enzyme. The enzyme is missing. Phenylalanine piles up behind the block (that is what the newborn screen detects) and tyrosine runs short after it. The treatment is to restrict the input (a low-phenylalanine diet) and supply the output (tyrosine). That one disease contains every concept you need for the rest of this site.
A pathway is a series of enzymes, each converting the product of the last. The molecule going into an enzyme is its substrate; what comes out is its product, which is the next enzyme's substrate. Pathways are usually drawn as chains or cycles.
Enzymes need cofactors, and the cofactors are vitamins and minerals. This is the connection between the vitamin aisle and biochemistry, and it is exact, not metaphorical. Thiamine is the cofactor for pyruvate dehydrogenase (hence Wernicke's). B6 is the cofactor for the enzyme that makes GABA and the enzymes that make serotonin and dopamine (hence the seizures of pyridoxine-dependent epilepsy). B12 and folate are cofactors for methionine synthase (hence the macrocytic anemia and neuropathy of their deficiency). Vitamin C is the cofactor for collagen's stabilizing enzyme (scurvy). Iron, copper, zinc, magnesium, selenium, and manganese each sit in the active sites of dozens to hundreds of enzymes. A cofactor deficiency looks exactly like a slow enzyme, because functionally it is one.
Pathways have bottlenecks, and bottlenecks are measurable. When an enzyme is slow or short of its cofactor, its substrate accumulates and its product falls, and the accumulated substrate is often something you can order. Methylmalonic acid rising is the enzyme that needs B12 falling behind; that is why MMA is a better B12 test than serum B12. Homocysteine rising is methionine synthase or its downstream partner falling behind; that is why homocysteine is the functional methylation test. Lactate rising at low workloads is pyruvate dehydrogenase falling behind. An acylcarnitine profile shows which step of fat oxidation is backed up. These are the traffic reports for the pathways, and they are what this site relies on instead of genotype alone.
Pathways share cofactors and raw materials. Magnesium serves over 300 enzymes. The methyl groups carried on S-adenosylmethionine (SAMe) are spent by over 200 different enzymes, including the one that clears histamine, the one that clears adrenaline and estrogen (COMT), the one that makes creatine (which alone uses about 40 percent of the daily supply), the one that makes phosphatidylcholine, and the ones that regulate gene expression. Glycine goes into collagen, glutathione, creatine, heme, and the glycine conjugation step in the liver. When one of those consumers runs hot, the others are shorted. This is the mechanism behind patients with five low-normal values and nothing out of range: the pools are shared, and under load, everything runs a little short at once.
Pathways regulate themselves by product. Cholesterol inhibits HMG-CoA reductase. Glutathione inhibits the enzyme that makes it. Creatine inhibits the first step of its own synthesis. The clinical consequence is useful: if you supply the end product of a pathway, the pathway slows down and stops consuming its raw materials, which are then available for other jobs. You use this logic when you give thyroid hormone and watch TSH fall.
Most pathways run with reserve. This is why most slow variants are silent. Pyruvate dehydrogenase at 70 percent capacity is fine at rest. It is not fine in sepsis, in thiamine deficiency, or during a marathon. Demand reveals the bottleneck.
Part 6: How re-routing works, built from PKU, hemochromatosis, and homocystinuria
The PKU diet is a re-route: restrict the input, supply the output, and the missing enzyme stops mattering. Most pathways have more than one route to the same product, and most of what this site proposes is finding the working route. The types, each with a clinical anchor:
Give the slow enzyme more of its cofactor. Some inborn errors are "vitamin responsive": a slow enzyme that binds its cofactor poorly works better when the cofactor is supplied in excess. Pyridoxine-responsive homocystinuria is the textbook case; some CBS variants work acceptably at high B6 doses. The common version of this is riboflavin for MTHFR C677T: the variant's defect is in riboflavin binding, and riboflavin supplementation partly rescues it. This is enzyme kinetics, not supplement folklore, and it has been catalogued for dozens of enzyme-cofactor pairs.
Use the parallel road. Homocysteine can be converted back to methionine two ways: through methionine synthase (needs folate and B12) or through an enzyme called BHMT (needs betaine, which comes from beets and spinach and from choline). When the folate route is limited, supplying betaine lets the second route carry more traffic. Betaine is in fact an FDA-approved drug for homocystinuria for exactly this reason. NAD can be made from tryptophan, from niacin, from nicotinamide, or from nicotinamide riboside, each through a different enzyme, so a block in one route can be bypassed through another; pellagra in carcinoid syndrome (tryptophan diverted to serotonin) is treated with niacin, the parallel road.
Enter the pathway downstream of the block. If the slow step is early, supply what it was supposed to make. Leucovorin (folinic acid) rescue after methotrexate is this: methotrexate blocks dihydrofolate reductase, and folinic acid enters the folate pool past that enzyme. The same logic is why folinic acid is useful in people whose folate processing is slow, and why hydroxocobalamin (which the cell converts to whichever active B12 it needs) is a sensible default form. Levocarnitine for carnitine deficiency skips the five-step synthesis that needs SAMe, iron, vitamin C, and B6. Creatine supplied in the diet lets the body stop making it. Medium-chain triglycerides bypass the carnitine shuttle, which is why they are used in long-chain fatty acid oxidation disorders (and contraindicated in medium-chain ones).
Reduce the load. When the problem is demand rather than capacity, lowering the input is the fastest lever. The PKU diet is the archetype. A low-histamine diet does not fix a slow histamine-clearing enzyme; it gives it less to do. Phlebotomy in hemochromatosis does not fix HFE; it removes the load. Reducing chemical exposure takes work off phase II conjugation.
Supply one product to free the pool for another. Because the methyl groups on SAMe are shared, supplying creatine and phosphatidylcholine (the two biggest consumers) frees methyl groups for histamine and catecholamine clearance and for nerve membrane repair, without adding a methyl donor that an overloaded system may not be able to dispose of. This is why the posts here sequence those two before any methylated B vitamin.
Clear the exit. Many pathways end in excretion: bile and stool, urine, sweat, breath. If the exit is blocked, the pathway stalls no matter how well the enzymes work. Cholestyramine works by binding bile acids so they leave in stool instead of recirculating; fiber does a gentler version of the same thing for conjugated estrogens and xenobiotics. Constipation is not a trivial finding in this framework; it is a closed exit.
Before proposing any re-route, the questions are whether the alternative route actually exists in the relevant human tissue, whether the supplied compound reaches the right compartment in an active form, and what else the supplied compound does. Zinc, for instance, induces a protein that blocks copper absorption, so isolated zinc supplementation in a patient whose collagen cross-linking depends on copper is a re-route that creates a new bottleneck.
Part 7: Why metabolic pathway medicine works
I call the way of reasoning on this site metabolic pathway medicine. It is a term I coined because nothing existing described it. Here is what it means, built from cases, and why I think it earns its place next to the symptom-based management you were trained in.
Start with a case you already handle this way. A patient on long-term metformin presents with fatigue, paresthesias, and a borderline-normal serum B12. You do not treat the paresthesias. You check MMA, find it elevated, replace B12, and the paresthesias resolve. You reasoned from symptom to pathway (methylmalonyl-CoA mutase is B12-dependent) to bottleneck marker (MMA) to cause (metformin impairs B12 absorption) to the lever that addresses the cause. That is the entire method. The claim of this site is that the same reasoning applies to a much larger set of presentations than the ones where medicine already uses it.
Symptoms are readouts of a pathway state. Fatigue is energy supply falling short of demand. Flushing is mast cell mediator release. Orthostatic tachycardia is a compensation for blood that is pooling instead of returning. A rising LDL is, in a fraction of patients, the liver shipping fat out because glucose oxidation has stalled. High glucose is cells declining fuel they cannot burn. Each symptom is information about which pathway is stuck. Suppressing the readout (an antihistamine, a beta blocker, a statin) is often the right thing to do as a bridge and sometimes as a destination. It does not change the pathway state, and because the pools are shared, the readout tends to reappear somewhere else.
The body is adapting, not malfunctioning. This is the assumption underneath every post. Sequestering chemicals in fat protects the brain and liver. Insulin resistance protects a cell that is already overfilled with fuel. A racing heart on standing preserves cerebral perfusion. Low T3 in illness conserves energy. These are reasonable responses to the state of the system, and disease is what they look like once the system has run out of reasonable responses. The practical consequence is that intervening on the upstream constraint can reverse the sequence, and that the symptom tells you which constraint to look for.
Genes tell you where to look; labs tell you whether you found it. A slow variant raises the probability that a particular step is a bottleneck. It does not prove that it is, because of spare capacity. The bottleneck marker (MMA, homocysteine, ferritin and iron saturation, red-cell magnesium, the mast cell mediators, organic acids, acylcarnitines, GGT) is the test of the hypothesis. The symptoms are what the whole model has to explain. When genotype, marker, and presentation line up, the model is worth acting on. Acting on genotype alone is how patients end up on twenty supplements and feeling worse, and you have probably seen that patient.
Most of chronic illness is dials, not switches. Clinical genetics was built to find the rare broken-protein variant that explains a Mendelian disease, and it does that well. Most chronic, multi-system, "all your labs are normal" illness is instead three or four slow enzymes across connected pathways, in a patient whose diet, infections, stress, drugs, and exposures have pushed demand past the reduced capacity. That configuration does not produce a genetic diagnosis, and it does not produce a flagged value, because each value is individually low-normal. It produces a pattern, visible only when the panel is read with the pathway in mind. This is why the posts here ask for specific markers that are not on the standard panel.
The predictions can be checked. If the model says methionine synthase is functionally limited and the proposed levers are applied, homocysteine should fall and MMA should normalize within about twelve weeks. If it says mast cell mediators are driving a patient's orthostatic intolerance, stabilization and histamine restriction should lower the mediator markers and the standing heart rate increment within about eight weeks. If they do not, the model is wrong and gets revised. That is the property that separates this from a supplement list: every step comes with a marker that would show it failing, which is the same standard you apply when you recheck a TSH or an INR.
Order of operations is most of the skill. When several pathways are limited at once, the order you address them in decides whether the patient stabilizes or crashes. The sequence the posts here follow: quiet the mast cells and remove triggers first, because an activated immune state turns every subsequent input into a stressor; restore absorption and the gut barrier second, because nothing reaches the pathways otherwise; replete cofactors and spare the methylation budget third, using non-methylated forms before methyl donors (the methyl trapping post explains the reaction you may have seen when a patient started methylfolate and became anxious and sleepless); support the mitochondria fourth; add movement only once post-exertional crashes have stopped; and go after deep drivers (viral reactivation, biofilms, mobilizing stored chemicals) last and under supervision, because provoking them in an unstable patient produces the crash that makes people give up. The patient who feels terrible for days after any new intervention is telling you to respect this order.
Food first, one variable at a time, lowest dose that moves the marker. Getting a cofactor from food is the lowest-risk route and it arrives with its companions. Supplements come after, added one at a time a week or more apart, so each response is attributable, the same logic you use when you change one drug at a time.
What it is not. It is not a substitute for diagnosis or for your management, and it is not a claim that nutrients replace drugs. Prescribed medications are frequently the right bridge while a pathway is being restored, and adjusting them is yours to do. It is a method for building a mechanistic model of a presentation, naming the measurements that would test it, and reading the literature for what would address each node if the model holds, so that you and your patient can make decisions from mechanism rather than from a normal panel and a shrug.
The reference pages
Four pages are built for lookup rather than reading: the Nutrient Guide here (biochemical role, deficiency gradient, assay selection, forms, and traps for each nutrient), the Lab Guide here (the same material indexed by assay, with reference and functional intervals and pre-analytic rules), the Medications table here (drug-induced depletion by class, with mechanism and monitoring), and the Food-Sources table here (nutrient density per serving with bioavailability notes). Alongside them: the nutrigenomic variant table here, the appointment checklist with a clinician-facing rationale page here, the three structured dietary trials (low-histamine here, low-FODMAP here, substrate separation here), the fish and methylmercury page here, and the site glossary here. They carry the detail the mechanism posts assume.
Reading the rest of this site
Every post is a worked example: the presentation, the proposed pathophysiology, the amplifiers, the markers that would test the model, and the diet, sequenced supplements, and lifestyle changes the literature supports, with a patient-facing version and a research version you can switch between at the top. The POTS post here and the metabolic disease post here are the best first reads; they run the method on a full presentation. The connective tissue post here explains the shared-budget idea that most of the others lean on.
Educational and research content, not medical advice. The author holds a PhD in cellular and molecular nutrition, is not a licensed physician, and does not diagnose, treat, cure, or prevent disease. The frameworks described here are offered for discussion between patients and their qualified clinicians.
