Perturbation Framework - Drug Identification Based on Response

Your Body Is a Sensor Array. Learn to Read It

#

Take a look at my interactive classifier tool for a quick assessment.

Legal disclaimer: This is not a diagnostic tool and should not be used to treat suspected overdoses or drug toxicities. Think of the tool as an advanced Field Sobriety Test (which is not admissible in court, by the way).

As a former drug addict, I’ve always been obsessed with a question: how much can you figure out about what a drug is doing without sending anything to a lab? No bloodwork, no mass spectrometer, no immunoassay. Just a heart rate monitor, a flashlight, your own observations, and a framework for interpreting what you see. Because you never truly know what you get on the streets.

I started developing this framework based on subjective data, how I feel on each individual drug, which effects I experience in different classes. I became fairly good at “feeling” the mechanism of action, and subsequently ID the drug. But I realize that building a framework based on my individual response wasn’t as useful as I thought, so I attempted to make it more general and applicable to people who don’t have a mental reference for each drug.

And I still think you can figure out quite a lot. More than most people would think. The important correction is that quite a lot is not the same as a validated identification.

Pharmacology has always used perturbation: apply a stimulus, measure the response, eliminate alternatives. Isolated tissue preparations are obviously not the same as a smartwatch and a flashlight, but the underlying logic is what got me interested. You’re running inference on a biological system by observing its outputs.

I’m calling it the Perturbation Framework. Because that’s what a drug is. It’s a perturbation. A signal injected into a system. And the system responds in ways that can be predictable, measurable, and if you know what to look for, interpretable.

This framework is now something I use as a reference for experiments on novel or unidentified compounds. I will repeat the important part: this is not a diagnostic tool. It is a thinking tool built from hard personal experience, not a replacement for toxicology, medical care, or common sense.

What This Is (and Isn’t)

#

Let me be honest upfront. This framework doesn’t tell you “you took drug X.” That’s a job for analytical chemistry, and pretending otherwise would be dishonest. It tries to rank the mechanisms or drug classes that best resemble the observations entered.

What it might let you say is: “whatever was ingested resembles adrenergic and dopaminergic activation more than serotonergic or GABAergic involvement, and the reported time course resembles an oral stimulant.”

That’s mechanism-level inference. And it’s surprisingly powerful (in my opinion). It is also unvalidated.

The framework treats drugs as system perturbations, not labels. You’re not asking “is this amphetamine?” You’re asking “which neurotransmitter systems appear to be pushed, and how hard?” That’s a better question. It is more honest, but it remains an inference from noisy observations.

The Core Insight: Weighted Evidence, Not Binary Labels

#

The old way: “Drug X causes Y.”

The better way: “This collection of signs gives +22 points toward sympathomimetic, +12 toward dopaminergic, -8 against opioid, and -6 against GABAergic.”

Every sign you observe is a piece of evidence. Some are strong: pinpoint pupils alongside respiratory depression strongly suggest opioid involvement. Some are weak: mild nausea could be anything. The framework assigns weights based on toxicology references and personal experience.

And here’s the part people miss: what’s absent can matter as much as what’s present. No sweating in someone who looks stimulated? That weighs against a classic sympathomimetic pattern and toward anticholinergic. Normal pupils in someone heavily sedated? That weighs against the classic opioid toxidrome and toward other sedatives.

It does not rule anything out. Mixed exposures, dose, route, timing, tolerance, lighting, illness, and individual variation can break a clean pattern.

Five Checks in Two Minutes

#

These five observations are high-signal differentiators in clinical toxicology. Outside a clinical setting they are much easier to perform badly, and none of them should delay emergency care. Think of this section as the intuition behind the model—not a home poisoning protocol.

1. Pupils

#

Pupils are one of the most informative observations you can make.

Pinpoint (tiny, even in dim light): Supports opioid or cholinergic involvement.

Blown wide (slow to react): Can fit sympathomimetic, anticholinergic, serotonergic, or psychedelic effects.

Normal: Can fit GABAergic, cannabinoid, low-dose, mixed, or simply badly timed observation.

Why this can work: the iris has two muscles controlled by opposing branches of the autonomic nervous system. The dilator is sympathetic (α1). The constrictor is parasympathetic (M3). Opioids can drive constriction via the Edinger-Westphal nucleus. But pupils are evidence, not a barcode.

2. Armpits

#

I’m serious. Axillary moisture is one physical finding used when separating similar toxidromes.

Wet armpits: Fits sympathomimetic, serotonergic, or cholinergic activity.

Bone dry armpits: Supports anticholinergic activity.

Here’s why this matters: sympathomimetics and anticholinergics can both cause dilated pupils, tachycardia, agitation, and high temperature. They can look almost identical. Except for sweat. Sympathomimetic states activate eccrine sweat glands. Anticholinergic states inhibit them. Useful clue; not ambiguity magically resolved.

3. Gut Sounds

#

Loud gurgling: Can support cholinergic or serotonergic activity.

Reduced or absent sounds: Can support anticholinergic or opioid activity.

Normal: Does not cleanly exclude any of them.

Most of the body’s serotonin is outside the brain, much of it in the gastrointestinal tract. Acetylcholine drives GI motility, while opioids and anticholinergics suppress it. The gut therefore carries information—but it is not broadcasting a receptor readout in Morse code.

4. Reflexes

#

Clonus + hyperreflexia (especially legs): Important in serotonin toxicity when the exposure history and other findings fit.

Diminished reflexes + ataxia: Can fit sedative-hypnotics.

Fasciculations (visible twitching under the skin): Can fit cholinergic excess.

Rigidity or catalepsy: Can occur with dissociatives and several other serious states.

Nystagmus while awake and agitated: Can support a dissociative pattern.

The Hunter Serotonin Toxicity Criteria are a clinical decision rule developed in a defined population and compared with toxicologist assessment. Copying one sign out of that context does not inherit the reported sensitivity and specificity.

5. Vital Signs Pattern

#

Not only the individual numbers—the pattern.

The pattern can narrow a hypothesis. It cannot identify the powder.

The Mechanism Profiles

#

This is the meat of the framework. Nine profiles, each representing a receptor system or drug class. For every sign you observe, you add or subtract points. The highest score is your primary hypothesis—not a result with known accuracy.

I’m not going to dump the full scoring tables here — I have a companion reference document and an interactive scoring tool for that. What I want to do instead is give you the intuition for each profile: what it looks like, what it feels like, and what separates it from things that look similar.

Sympathomimetic (Adrenergic)

#

In one sentence: Fight-or-flight, cranked up pharmacologically.

Catecholamines are flooding the system. Heart pumps harder, blood vessels constrict, pupils widen, sweat glands activate, appetite disappears, energy goes through the roof. The person is wired, wide-eyed, warm, and damp.

Hallmark signs: Mydriasis, sustained tachycardia, hypertension, diaphoresis (sweating), psychomotor activation, bruxism, suppressed appetite.

A useful negative: If someone looks stimulated but their skin is dry, the pattern becomes less classically sympathomimetic and more suggestive of anticholinergic activity.

Telling stimulants apart by duration: You can’t distinguish amphetamine from cocaine by pupils or heart rate. Time course can help, but dose, route, formulation, metabolism, redosing, and mixtures create enormous overlap. Duration narrows a story; it does not identify a substance.

HRV hypothesis: Sympathetic activation often coincides with lower short-term HRV. Whether a wearable can distinguish a drug mechanism from stress, posture, breathing, sleep loss, or its own signal processing is another question entirely.

Dopaminergic (The Reward Overlay)

#

This isn’t a separate toxidrome — it’s a layer on top of the adrenergic profile that tells you how much reward and motivation seem to be in the mix.

A drug that makes your heart race but doesn’t change your mood may be more adrenergic than dopaminergic. When euphoria, drive, focus, libido, and compulsive re-dosing urge are prominent, I suspect dopamine is more involved.

A pattern I notice: repetitive stereotyped behavior — picking at skin, compulsively reorganizing objects, starting the same task over and over. That is not a dopamine assay, but it is part of the subjective profile that inspired the framework.

Serotonergic (5-HT)

#

In one sentence: The neuromuscular one. When serotonin goes too high, the body can get twitchy.

The clinical triad is mental status changes + autonomic instability + neuromuscular hyperactivity. Clonus is especially important when the surrounding history and findings fit.

Why serotonin is weird: Most of your body’s serotonin is outside the brain. That’s one reason GI symptoms can be prominent. The serotonin system also sits at a crossroads of thermoregulation, neuromuscular control, and autonomic stability—which is why serotonin toxicity can look chaotic.

The critical differential: Serotonergic and sympathomimetic toxicity can both cause elevated vitals, sweating, and agitation. Clonus and hyperreflexia move the hypothesis toward serotonergic toxicity; they do not turn a non-clinician into a toxicologist.

Opioid

#

In one sentence: Tiny pupils, slow breathing, raised pain threshold. The classic triad.

Miosis + respiratory depression + reduced consciousness is a high-stakes opioid pattern. Respiratory depression is the important part. If breathing is slow, irregular, or inadequate, act on that emergency rather than waiting for every item in the triad.

Pinpoint pupils are strongly associated with opioids, but normal pupils do not safely exclude them. Mixed exposure, specific opioids, dose, hypoxia, and timing can all change the picture.

Sub-classifying: Different opioids vary in histamine release, additional serotonergic or adrenergic effects, onset, and duration. Those differences are interesting after the person is safe. They are not a dependable street-identification system.

GABAergic / Sedative-Hypnotic

#

In one sentence: The brakes are on. Coordination goes, speech slurs, memories don’t form.

GABA is the brain’s primary inhibitory neurotransmitter. Enhance it, and neural activity slows across systems involved in coordination, speech, memory, and judgment.

Hallmarks: Ataxia and slurred speech are common. Opioids can also sedate; pupils and respiratory pattern help form the differential, but polydrug exposure ruins this neat binary split very quickly.

GHB warning: GHB has a steep and unpredictable dose-response relationship, especially with other depressants. Sudden deep sedation, bradycardia, abnormal breathing, and myoclonic movements are reasons for emergency assessment—not reasons to admire that the classifier guessed correctly.

Anticholinergic

#

In one sentence: Everything dries up. Brain goes haywire.

The classic mnemonic from emergency medicine: “Hot as a hare, blind as a bat, dry as a bone, red as a beet, mad as a hatter, full as a flask.”

Translates to: hyperthermia, blurred near vision, reduced sweat and saliva, flushed skin, delirium, and urinary retention.

Dry skin or axillae can be a useful differentiator from a sweaty sympathomimetic presentation. The hallucinations may look mundane—cigarettes, insects, people, or objects that are not there—but that character is not exclusive enough to identify a mechanism on its own.

HRV hypothesis: Blocking cardiac parasympathetic signalling can reduce beat-to-beat variation and raise heart rate. My old version called a near-flat wearable trace at 95–105 bpm “diagnostic.” It is not. Device processing, rhythm disorders, movement, breathing, illness, and measurement error can all create a convincing fake.

Cholinergic

#

In one sentence: The opposite of anticholinergic. Everything is turned on and secreting.

SLUDGE: Salivation, Lacrimation, Urination, Defecation, GI distress, Emesis. Muscles may fasciculate, pupils may constrict, and bronchial secretions can become life-threatening.

If you see this after pesticide exposure, treat it as a medical emergency.

Dissociative (NMDA Antagonist)

#

In one sentence: Disconnected from reality, reduced response to pain, sometimes nystagmus in a person who is awake.

Dissociatives do not simply sedate; they can disconnect conscious experience from sensory input. The person may be eyes-open and moving but profoundly disconnected from the environment.

Nystagmus, hypertension despite apparent sedation, and analgesia can support the pattern. None is unique, and different dissociatives vary wildly in duration and risk.

Cannabinoid

#

In one sentence: Red eyes, munchies, altered time, sometimes dizzy when you stand up.

Conjunctival injection and tachycardia are common, while orthostatic symptoms can occur. They are recognizable, not specific. Synthetic cannabinoids are a different risk category: full agonism and unknown products can produce seizures, severe psychosis, rhabdomyolysis, cardiovascular toxicity, and death.

Psychedelic (5-HT2A)

#

In one sentence: The world looks completely different while basic physiology may remain surprisingly ordinary.

The subjective intensity of classical psychedelics can be disproportionate to vital-sign changes. A useful distinction from delirium is retained insight: someone may know they took a drug and remain oriented despite altered perception. But compounds sold as psychedelics can have very different pharmacology, and “physiologically benign” should never be inferred from a label or a visual effect.

HRV: The Hidden Goldmine—or My Favourite Overreach

#

If you own a wearable that tracks heart rate variability, you’re holding a useful non-invasive signal. HRV describes variation between heartbeats, with vagal activity, breathing, and measurement conditions playing major roles.

Quick primer:

• RMSSD — often used as a short-term marker influenced by parasympathetic activity.
• HF power (0.15–0.4 Hz) — strongly tied to respiration and vagal modulation.
• LF/HF ratio — historically described as “sympathovagal balance,” but that interpretation is heavily disputed and much less clean than the name suggests.

The patterns I originally expected were:

I still think this is worth investigating. I do not think these are established drug fingerprints. HRV changes with breathing, posture, sleep, fitness, age, stress, illness, time of day, recent exercise, and device processing. A review of HRV interpretation is a good antidote to turning one wearable number into a receptor panel.

Confounders: What Will Ruin Your Data

#

I’ll be blunt. If you don’t control these, you don’t have data. You have noise that looks like data, which is worse.

Caffeine, nicotine, sleep deprivation, exercise, stress, ambient temperature, lighting, body position, meals, medications, illness, and expectation can all change the observations in this framework. Stress alone can produce mydriasis, sweating, and elevated vitals that resemble a stimulant pattern.

Log these every session: caffeine and nicotine timing, sleep, exercise, stress, temperature, light, body position, last meal, medications, and the device used.

The placebo problem is especially bad in self-experimentation. When the observer is also the subject, you can’t easily blind yourself. Baseline days, pre-defined measurements, and randomized self-blinding can improve an experiment. A dose-response pattern can be interesting, but it is not automatic proof: expectation can scale, absorption can be nonlinear, and high doses can create new mechanisms and risks.

The Subjective-Objective Mismatch

#

This is one of the most interesting features in the framework, and one that doesn’t get enough attention.

Pay attention when what someone reports feeling doesn’t match what their body is doing.

Feels calm, but HR and BP are elevated. Something may be suppressing the experience of arousal while the body stays activated—or the person may be anxious, dehydrated, ill, or combining substances.

Feels stimulated and wired, but weak peripheral signs. The subjective effect may be more central, or the measurements may simply have missed the peak.

Feels sedated, but vital signs are normal. Mild sedative or cannabinoid effects are possibilities. It still does not rule out an opioid or mixture.

Reports analgesia without sedation. NMDA antagonism or partial μ-agonism are hypotheses, not conclusions.

These mismatches can be signal. They can also be confounding. The gap between subjective and objective is informative precisely because it forces another question.

The Honest Limitations

#

This framework cannot identify exact compounds with certainty. Closely related drugs can hit the same receptors, while route, dose, contaminants, and time course reshape the response. Definitive identification requires analytical chemistry.

It performs badly with polydrug combinations. Two substances can produce overlapping or conflicting patterns, and the dangerous component may not be the highest score.

It cannot account for all individual variation: metabolism, tolerance, baseline physiology, illness, medications, and genetic differences all move the profile.

The weights are mine. They were not learned from a representative dataset, the tool has no measured sensitivity or specificity, and it has not been compared against confirmed toxicology. A score tells you which of my assumptions the answers matched.

And it is not a substitute for medical testing. If there is a medical emergency, call for help. This is a research and harm-reduction framework, not a diagnostic tool.

For the clinical concept behind the broad categories, see the MSD Manual overview of common toxidromes.

Why I Think This Matters

#

Here’s the thing that got me excited about this in the first place.

We’re surrounded by tools that measure our physiology in real time—heart-rate monitors, HRV trackers, skin-temperature sensors, even pupilometry apps on phones. But we treat them as black boxes. Your watch tells you your heart rate is high. Okay, and? What does that mean in context?

This framework tries to turn those numbers into inference. It connects raw data to biology. It treats the body as a sensor array that is constantly broadcasting information about its internal state, if you know how to listen.

The correction is that a noisy sensor array does not become a mass spectrometer because I gave each checkbox a weight.

I still stand by the idea: structured observation is better than pure vibes, negative evidence matters, and the mismatch between what a person feels and what the body does can be fascinating. I also stand by leaving the limitations visible instead of replacing the article every time my thinking changes.

The full scoring tables, interactive classifier tool, and complete reference remain available as companion resources. Use it, break it, improve it. If you find something that doesn’t work, or a pattern I missed, I want to know.

— Henrik