The Short Hypothesis · Hypothesis, not yet tested

Autism at the Cellular Level: Where Short Peptides Might Fit

Autism research and short-peptide studies touch some of the same cellular processes. What does that overlap mean, and what would it take to test it?

Disclosure: Julia Shulman co-founded Pure Fire Nutritional, which sells peptide bioregulator supplements, including products related to several peptides discussed here. This article does not recommend any product for purchase or use. How we work

The Question

What does it mean for a body system to work well?

Before asking whether anything could help a cell, it's worth asking what "working well" even means.

A healthy body isn't a machine running at one correct setting. Its systems keep adjusting to each other, and there are always trade-offsGains in one area that come with costs in another. A system tuned for one challenge is often more vulnerable to a different one.: a system tuned for one challenge is often more fragile to another.

The stress response is a good example. It protects us in the short term, but kept switched on for too long it becomes costly.

That's why difference is not dysfunction. A biological adaptationA change that helps a cell or body cope with its circumstances. Something that is an adaptation in one setting can carry costs in another. can be beneficial in one context and costly in another. Many differences found in autistic people are group averages, often found only in subgroups, and often consequences rather than causes.

What does a “group average” difference mean?

When a study says autistic people have higher or lower levels of something on average, it is comparing two large groups.

The two groups usually overlap a lot, so an average difference says little about any one person. A difference found only in a subgroup applies to some people, not everyone.

So "working well" can't mean "matching the average". It has to mean working well for a particular person, in the life they actually live. That's one reason good autism research starts from what autistic people themselves prioritise (Kapp 2013; Pellicano 2014; Pellicano & den Houting 2022).

See the evidence: robustness, stress and difference

A healthy body isn't a machine running at one correct setting. It's a set of systems that keep adjusting to each other: energy use, repair, immune defence, sleep and the stress response all shift with what the day demands. Biologists call the ability to keep working through disturbance robustness, and it always comes with trade-offs. A system tuned to cope well with one kind of challenge is often more fragile to another (Kitano 2004).

The stress response is a good example. Stress hormonesChemicals such as cortisol and adrenaline that the body releases when it faces a challenge. In the short term they raise alertness and energy. and nervous-system arousalHow revved up the body's automatic systems are: heart rate, breathing, alertness. High arousal readies you for action; low arousal lets you rest. protect us in the short term. Kept switched on for too long, the same responses become costly (McEwen 1998). Whether a response is helpful or harmful depends on timing and context, not just on its level.

That's why difference is not dysfunction. A biological adaptation can be beneficial in one context and costly in another. Autism is usually described as a difference in how the brain develops, and its genetics point there (Satterstrom 2020). But differences turn up throughout the body too, and most are group averages, often found only in subgroups, and often consequences rather than causes: restricted diets go with a less diverse gut microbiomeThe trillions of bacteria and other microbes living in the gut. Their mix varies a lot from person to person and shifts with diet. (Yap 2021), and disturbed sleep goes with higher inflammatory markersSubstances in the blood that rise when the immune system is active, for example during an infection. They are used as rough signs of inflammation. (Irwin 2016).

Three questions for every finding

This piece asks the same three questions of every finding, whether it comes from autism research or from peptide research:

1. Cause. Does the measured difference actually contribute to a difficulty the person cares about?
2. Benefit. Would changing it make life better by the person's own measures, not just move a biomarkerA measurable sign in the body, such as a level in the blood, used to track a biological process. A biomarker changing doesn't always mean a person feels or functions differently.?
3. Safety. Is the change safe in the people, route and duration tested?

Most of what follows can't yet answer any of the three. Saying so plainly is the point.

Why does each question need its own proof?

Finding a difference doesn't show it causes a problem: it could be a side effect of diet, sleep or stress.

Even if it does matter, changing the marker might not make anyone feel or function better. And a change can be helpful in one way while being unsafe in another. So each step has to be tested on its own.

Seven places where the research meets

Each card gives the autism side; the evidence panels add the peptide studies.

See the evidence: the links behind the loop

A proposed loop, not an established one. Each arrow stands for a link reported in some studies: for example, lower heart-rate variabilityThe small, natural changes in timing from one heartbeat to the next. More variation usually means the calming branch of the nervous system is more active. under social stress (Cheng 2020), situation-dependent cortisol responses (Taylor & Corbett 2014), and anxiety and sensory over-responsivityReacting more strongly than most people to everyday sounds, lights, textures or smells, so that they feel intense or overwhelming. going with sleep problems (Mazurek & Petroski 2015). The loop as a whole hasn't been tested in autistic people, and whether it applies, and where someone enters it, will differ from person to person. Easing load at any point, for example through a quieter environment, may matter as much as anything aimed at the biology.

The figure is why this section matters: body systems don't act alone. A difference in one, such as arousal, can ripple into others, such as sleep, and back again.

What is a feedback loop?

A feedback loop is a chain where each step feeds the next, and the last step feeds back into the first.

Here the idea is that a heavy day could raise arousal, unsettle stress-hormone timing and disturb sleep, and a poor night could leave less capacity for the next day. It is a proposed picture pieced together from separate findings; no one has tested the whole loop in autistic people.

Enter the short peptides

A St. Petersburg group led by V. Khavinson developed Epithalamin, a pineal extract, and then synthetic peptidesPeptides are short chains of amino acids, the building blocks of proteins. Synthetic ones are made in a lab rather than extracted from animal tissue. two to four amino acidsSmall molecules that link together like beads to make peptides and proteins. Each is often written as one letter, so AEDG is four amino acids in a row. long, including Epitalon (AEDG) (Jain 2026). A separate Moscow group developed Semax and Selank.

In ageing, stroke, stress and Alzheimer's modelsIn research, a model is a stand-in for a human condition, such as animals or cells treated to mimic part of it. Results in a model may not carry over to people., these groups report effects on processes autism research keeps returning to: oxidative balanceThe balance between reactive oxygen molecules, which cells make all the time, and the antioxidants that keep them in check., immune signalling, gene activity, nerve-cell connections and melatonin. That overlap is the interesting part, and where the problems start.

Almost all the work is in cells or animals, much of it from the developers' own groups, with little independent replication (Jain 2026). We found no published clinical studies testing the peptides reviewed here on their own in autistic participants, in searches of PubMed, ClinicalTrials.gov and the Russian-language database CyberLeninka (search log). Three Russian clinical reports describe Semax given to autistic children alongside several other drugs. None tested it on its own or had a comparison group that could isolate its effect, so they can't show whether Semax helped or was safe. The only study we found that tested one of these peptides on its own in an autism model used rats exposed to valproate before birth: an animal model, not a clinical study (Malyshev 2013).

In the two most direct lab tests, the peptides showed no effect on the specific pineal and mitochondrial outcomes measured.

So the real question isn't whether peptides "work for autism". It's this:

Can we identify specific, measurable physiological vulnerabilities in subsets of autistic people, and investigate, in properly designed studies, ways to support those systems without trying to alter the neurotype itself?

That opens researchable questions, especially around circadian regulationHow the body's internal 24-hour clock times sleep, hormones, body temperature and other daily rhythms. and sleep, metabolic health, and cellular responses to stress.

The crucial distinction: regulating a measurable biological process is not the same as improving someone's wellbeing. Suppose a peptide altered oxidative-stress markers in cells derived from autistic adults. That would be mechanistically interestingInteresting for what it reveals about how a process works inside cells, rather than as evidence that something helps people.. It wouldn't establish that oxidative stress caused their difficulties, that the change was beneficial, or that the peptide was safe. Each of those has to be established independently.

More detail: an open research direction

Framed that way, there is an open therapeutic research direction here: studying whether specific physiological difficulties, such as disrupted sleep or a stress system that stays switched on, can be eased in adults who want help with them, without aiming at autism itself.

See the evidence: the two direct lab tests

In the two most direct lab tests, the peptides didn't change melatonin release from isolated pineal glandsPineal glands removed from animals and kept alive in a lab chamber. The pineal gland is the pea-sized gland in the brain that makes melatonin. or mitochondrial activity in cultured neuronsNerve cells grown in a dish, here made from donors' skin cells. They let researchers test effects directly, but without the rest of the body around them. (Djeridane 2003; Kraskovskaya 2024). Those are specific outcomes in specific systems, not proof the peptides have no pineal or mitochondrial effects.

Gene regulation, cellular stress and adaptation

This is where the two lines of research overlap most, and where the peptide claims are boldest. It comes down to two linked questions: how cells decide which genes to use, and how they respond to stress.

The genetic evidence is strong: large studies keep implicating autism-linked genes that regulate chromatinDNA in our cells is wrapped around proteins into a material called chromatin. Tightly packed chromatin tends to keep genes switched off; looser chromatin lets the cell read them., the packaging that decides which stretches of DNA a cell can read. The chemical marks on DNA are much less settled.

The peptide claims rest mostly on cell studies and computer models from the developers, with no independent confirmation and none in autism-relevant cells.

A cell isn't a developmental process. Changing gene expressionWhether, and how much, a cell is using a particular gene to make its product. Cells adjust their gene expression constantly in response to their surroundings. in a dish isn't changing development, and a developing brain is exactly where broad effects on gene switching or nerve connections could carry risks no one has studied.

What's the difference between a cell in a dish and a developing brain?

A cell in a dish is studied on its own, for hours or days, under controlled conditions.

A developing brain is built step by step over years, and the timing of each gene switch shapes what comes next. A change that looks harmless in a dish could alter that sequence in ways no one has measured, which is why the two are kept apart here.

The core claim can be tested directly. If these peptides act on specific stretches of DNA, a scrambled version with the same amino acids in a different order shouldn't work. If it works just as well, the "gene switch" idea is in trouble.

Why test a scrambled peptide?

A scrambled peptide has the same building blocks as the original, in a different order.

If only the original order works, that supports the idea that cells recognise the peptide's exact sequence. If the scrambled one works just as well, the effect isn't coming from the sequence, and the sequence is the heart of the gene-switch claim.

See the evidence: genes, marks, stress responses and peptide–chromatin claims

Explain this to me

Why does the way DNA is packaged matter?

Every cell carries the same DNA but uses only some of its genes. DNA is wound around proteins, and how tightly a stretch is wound helps decide whether the cell can read it. Many genes linked to autism help control that packaging.

The genes: strong. Large sequencing studiesStudies that read people's DNA letter by letter, often in thousands of people, to find genetic changes that turn up more often in one group. consistently implicate genes that regulate chromatinDNA in our cells is wrapped around proteins into a material called chromatin. Tightly packed chromatin tends to keep genes switched off; looser chromatin lets the cell read them., the packaging that decides which stretches of DNA a cell can read (De Rubeis 2014; Satterstrom 2020; Fu 2022). That's strong evidence that chromatin regulation matters in how some brains develop. It isn't evidence that autistic people's chromatin is "faulty" now.

The marks: much less settled. Studies of the marks themselves, such as DNA methylationOne of the most studied epigenetic marks: a small chemical tag on DNA that usually helps turn nearby genes down. It can shift with age, medication and the type of cell sampled., are smaller and harder to read. A post-mortem brain pilot found a handful of replicated differences (Ladd-Acosta 2014); blood studies find only small, uncertain ones (Andrews 2018). Marks can reflect medication, age or the cells sampled as easily as a cause.

Stress responses are partly adaptations. The oxidative and energy differences in the cards above are often read as damage. But cells also use oxidative signalsSmall, controlled amounts of reactive oxygen that cells use as alarm messages to sense trouble and adjust. and shifts in energy use to sense trouble and adjust. A lower antioxidantA substance that neutralises reactive oxygen molecules before they can damage the cell. reading could be a cost, an adaptation to diet, sleep or load, or simply a consequence of them. The same marker can mean different things in different bodies.

The peptide claims: mostly cells and computer models. The developers report that AEDG loosens tightly packed chromatin in older people's blood cells, that labelled peptides enter the nucleus and, from computer models, that they bind specific DNA sequences or histonesThe spool-like proteins DNA wraps around to form chromatin. Chemical tags on histones help control which genes are read. (Khavinson 2003; Fedoreyeva 2011; Khavinson 2020). We found no independent genome-wideLooking across the whole genome, all of a cell's DNA, rather than at a few chosen spots. confirmation, and none of it used autism-relevant cells.

A cell isn't a developmental process. Changing gene expressionWhether, and how much, a cell is using a particular gene to make its product. Cells adjust their gene expression constantly in response to their surroundings. in a dish isn't changing development. Development is timed and cumulative: which genes switch on, in which cells, at which stage, shapes what gets built. So cell-culture results shouldn't be read as claims about development, and a developing brain is exactly where broad effects on gene switching or nerve connections could carry risks no one has studied. Changing a process in a developing brain is very different from investigating a specific physiological problem in an adult.

The experiment that would settle the core claim. If these peptides really act on specific stretches of DNA, the effect should depend on the exact order of their amino acids. That can be tested directly. Take nerve cells grown from consenting autistic and non-autistic adult donors, plus matched cell lines that differ only in a chromatin gene such as CHD8A gene that helps control chromatin. Disruptive changes in it are found in a small, recognisable subgroup of autistic people.. Expose them, blindThe researchers testing the cells don't know which sample got which treatment, so their expectations can't influence the results., to the peptide, to a scrambled version with the same amino acids in a different order, and to versions with single amino acids swapped. Then map which parts of the genome open up, which genes change activity, and how the cells handle a controlled stress. A real sequence-specific effect would show up with the original peptide and fade with the altered ones. If the scrambled peptide works just as well, the "gene switch" idea is in trouble.

Finding	Study type	Relevance to autism
Genes: exome sequencingReading just the protein-coding parts of a person's DNA, about 1 to 2 percent of it, where many disruptive genetic changes are found. implicates chromatin-remodelling and nerve-communication genes (De Rubeis 2014; Satterstrom 2020; Fu 2022); high-confidence genes are catalogued in SFARI GeneA public online database, run by an autism research foundation, that lists genes linked to autism and rates the strength of the evidence for each. (Abrahams 2013)	Human genetic data	Direct
Genes: disruptive CHD8 changes define a recognisable subgroup (Bernier 2014)	Human genetic data	Direct, subgroup only
Gene activity: post-mortem cortex shows lower activity in synaptic genesGenes for the parts of nerve cells that pass signals across synapses, the tiny gaps between one nerve cell and the next. and higher activity in immune and glialGlial cells are the brain's support cells, including microglia. They nourish, protect and clean up after nerve cells. genes (Voineagu 2011; Gandal 2022)	Human post-mortem tissue	Direct
Marks: four genome-wide significant methylation regions in post-mortem brain from 19 autistic people and 21 controls; three replicated in other samples. The authors call it "suggestive" (Ladd-Acosta 2014)	Human post-mortem tissue	Direct
Marks: methylation differences within identical-twin pairs who differ in autism or related traits (Wong 2014)	Human observational data	Direct
Marks: blood methylation meta-analysis found only small, uncertain differences (Andrews 2018)	Human observational data	Direct
AEDG loosened tightly packed chromatin in cultured blood cells from people aged 76 to 80 (Khavinson 2003); several peptides did the same in older adults' white blood cells (Khavinson 2004) (lab cells)	Cell evidence	Unrelated model
Labelled short peptides entered the nucleus of human cells in a dish (Fedoreyeva 2011; lab cells)	Cell evidence	Unrelated model
Claim that these peptides bind specific DNA sequences or histones (Khavinson 2005; Khavinson 2020; computer modelling)	Modeled	Unrelated model
AEDG raised nerve-development markers in gum-derived stem cells (Khavinson 2020; lab cells)	Cell evidence	Unrelated model
EDR and KED preserved dendritic spinesTiny bumps on a nerve cell's branches where it receives signals from other cells. in Alzheimer's models, in mouse neurons in a dish (Kraskovskaya 2017) and in an Alzheimer's mouse model (Khavinson 2021; KED injected into the abdomen; EDR route not stated)	Animal evidence	Unrelated model
Semax in rat offspring exposed to valproateAn epilepsy medicine. Rats exposed to it before birth are a widely used animal model in autism research. before birth, the only autism-relevant model we found (Malyshev 2013; no abstract; route not stated)	Animal evidence	Indirect (animal autism model)
Any peptide in CHD8 or other autism-gene models, or in autism-derived neurons (not studied)	Proposed	Not yet tested

Limits and open questions.

• Gene-activity and methylation differences in post-mortem tissue may reflect consequences of life experience, medication or cause of death, not causes.
• We found no independent chromatin-mappingLab methods, such as ChIP-seq, ATAC-seq and CUT&RUN, that show exactly where on the DNA a molecule sits or which stretches of chromatin are open. evidence (for example ChIP-seq, ATAC-seq or CUT&RUN) that any of these peptides acts at specific sites in the genome.
• Oxidative signals and shifts in energy use are also how cells sense and adapt to stress; short bursts of stress responses protect, while long-running ones cost (McEwen 1998). A single marker reading can't tell adaptation from damage.
• More dendritic spines or more "open" chromatin isn't automatically good. In a developing brain, both are tightly timed, and broad changes could carry risks no one has studied.
• A clean test of the chromatin claim would use neurons from autistic and non-autistic donors, plus matched cell lines with and without a CHD8 change, with scrambled-sequence control peptides.

Biological activity is not biological optimization

A substance that changes something in a cell is biologically activeIt causes a measurable change in a cell or body. That alone says nothing about whether the change is helpful, harmful or neither.. That alone says little about whether it makes the system work better.

Pushing one variable can cost something elsewhere, so "more" isn't automatically better.

That's why the peptide findings in this piece, even the positive ones, only answer a fourth, easier question: did something change? The three questions that matter (cause, benefit and safety) come after that, and none has been answered for autistic people.

More detail: why “more” isn't always better

Body systems are built on trade-offsGains in one area that come with costs in another. A system tuned for one challenge is often more vulnerable to a different one. (Kitano 2004). Pushing one variable can cost something elsewhere. More "open" chromatin, more nerve-cell branches or higher melatonin each sounds like an improvement, but none is automatically good. In a system that's already balanced for the life it's living, "more" can simply mean "different", or worse.

One illustrative hypothesis: melatonin and sleep

Sleep shows what a testable version of this idea could look like. It's one illustration, not the main event.

Sleep difficulties in autistic people have many possible causes, including anxiety and sensory over-responsivityReacting more strongly than most people to everyday sounds, lights, textures or smells, so that they feel intense or overwhelming.. Low melatoninA hormone the body makes mainly at night, largely in the pineal gland, that helps tell the body it's time to sleep. applies only to a subgroup, so a melatonin-based idea could only ever apply to some people.

Pineal-peptidePeptides extracted from, or designed from, the pineal gland, the pea-sized gland in the brain that makes melatonin. researchers report melatonin rising mainly when levels start low. But no peptide study has measured sleep, and one lab study found Epitalon had no effect on melatonin release from rat pineal glands.

See the evidence: melatonin and sleep

Sleep difficulties in autistic people have many possible causes. Many autistic adults sleep worse (Morgan 2020). In children, anxiety and sensory over-responsivityReacting more strongly than most people to everyday sounds, lights, textures or smells, so that they feel intense or overwhelming. are both linked to sleep problems (Mazurek & Petroski 2015), and the stress loop above may play a part. In a sizeable subgroup, melatoninA hormone the body makes mainly at night, largely in the pineal gland, that helps tell the body it's time to sleep. is low, and in one large study low melatonin was linked to insomniaOngoing trouble falling asleep, staying asleep or waking too early, enough to affect daytime life. (Pagan 2014). So a melatonin-based idea could only ever apply to some people.

Pineal-peptidePeptides extracted from, or designed from, the pineal gland, the pea-sized gland in the brain that makes melatonin. researchers report melatonin rising mainly when levels start low: in older adults given Epithalamin (Korkushko 2004), and in one small placebo-controlledSome participants get the real thing and others get a look-alike with no active ingredient, called a placebo. This shows whether the real thing beats expectation and chance. study of under-the-tongue Epitalon that didn't measure sleep (FDA 2026). Some links already push back. In a lab study of young and old rat pineal glands, Epitalon had no effect on melatonin release (Djeridane 2003). One lab study isn't the last word, but it suggests Epitalon doesn't act on the gland directly. If low melatonin comes from low ASMT activity, the last step in making it (Melke 2008), pushing the gland harder might not help anyway.

The Experiment

What would convince us?

Every result, in a dish or in people, would have to answer the three questions from the start of this piece: cause, benefit and safety. A peptide could pass the first and fail the other two. Each has to be established independently.

A fair test would have two steps, both independent and registered in advance: a blinded lab screen in cells from consenting autistic and non-autistic adult donors, with scrambled-peptide controls; then a randomised, double-blind sleep study in consenting autistic adults (not children) with insomnia and low melatonin, comparing peptide, melatonin (which already has evidence in autistic children) and placebo. Any positive result would need to be replicated by a lab with no ties to the developers or sellers.

What would prove it wrong: no effect in cells from autistic donors; a scrambled peptideA version with exactly the same amino acids as the original, rearranged in a different order. It tests whether the order itself matters. of the same amino acids working just as well; no intact peptide in the bloodShort peptides can be broken down quickly in the gut and blood. Checking for the whole, unbroken peptide in blood shows whether any of it survives to reach the body.; melatonin rising without better sleep; failing to beat plain melatonin; or safety signals.

The Experiment: full design of both steps

Step 1: a lab screen. With ethics approval and donor consent, cells from autistic and non-autistic adult donors, about 10 to 12 per group, testing each peptide blind against a scrambled version and controls for energy use, oxidative stress and gene activity.

Step 2: a sleep study.

• Who: consenting autistic adults (not children) who have insomnia and low night-time melatonin, which directly tests the "works only when levels are low" claim.
• Step zero: confirm the product contains what it says and reaches the blood by the route tested.
• Design: randomised, double-blind and pre-registeredThe study's plan and main outcome are published before it starts, so the results can't be reshaped afterwards to look better., comparing peptide, prolonged-release melatoninMelatonin in a form that releases slowly over several hours, rather than all at once. and placeboA look-alike with no active ingredient. Comparing against it shows whether the real thing does better than expectation and chance..
• Measures: wrist-sensor trackingA watch-like device that records movement through the night to estimate when and how well someone sleeps., a sleep diary, melatonin markers, ASMT variantsNatural differences in the gene for ASMT, the enzyme that carries out the last step in making melatonin. (genetic results handled with consent and privacy protections), side effects and well-being, with sensory accommodations.
• Size: a pilot of about 30 to 40 people first. A full three-group trial would need a formal sample-size calculationA calculation done before a trial to work out how many people it needs to reliably detect a real effect, if there is one., likely dozens of people per group.
• Governance: ethics approval, FDA clearance to study an unapproved compound in people (or FDA confirmation that it isn't needed), an independent lead researcher and statistician, an autistic adult advisory board, full disclosure of any funder's commercial interests, and a commitment to publish null resultsResults showing no effect. Publishing them matters because otherwise only positive findings get seen, which skews the overall picture..

"Not studied" isn't "studied and found not to work."

Why compare against melatonin, not just placebo? Beating placebo only shows something beats nothing. Melatonin already has evidence for sleep in autistic children, so it is the fair benchmark. Comparing against it stops a "better than placebo" headline from being oversold and shows whether a peptide adds anything.

Why adults, and why cells first? A developing brain is where broad effects on gene switching or nerve connections could carry the most risk, and no reliable safety data exist in children. Cells from consenting adult donors let the mechanism be tested before anyone is exposed.

Supporting cells, not correcting people

If autism involves differences in cellular regulation, what does it actually mean to support cellular function without assuming those differences need correcting?

Two easy answers fail. One assumes every biological difference associated with autism is pathological, to be pulled back toward the average. The other assumes that accepting neurodiversityThe idea that natural variation in how human brains work, including autism, is part of normal human diversity. means looking away from physiological difficulties that deserve attention: sleepless nights, exhaustion, gut pain.

Systems biologyA way of studying living things that looks at how many parts interact as a whole, rather than at one gene or molecule at a time. offers a better frame: living systems don't run at one correct setting; they hold many variables in balance, and a different balance can be stable and workable.

So the question for any cellular finding isn't "how far is this from typical?" It's "is this causing a problem for this person, and would changing it help?" A melatonin level below the typical range matters if someone can't sleep. It may not matter at all if they sleep well.

Supporting physiological resilience and changing the environment aren't competing approaches. Reducing unnecessary demands matters, but so does understanding whether the body can be better supported in responding to demands that can't always be avoided. Neither requires treating autistic differences themselves as problems to be corrected.

The most interesting possibility is understanding what helps those cells function well in the person and environment they're actually part of.

More detail: fit between a person and their world

A trait that costs something in one environment can be an asset in another. Psychology says something similar about people: wellbeing depends on the fit between a person and their world, not on how closely they match a norm.

The most interesting possibility isn't making autistic cells behave like non-autistic cells. It's understanding what helps those cells function well in the person and environment they're actually part of.

Safety, and what this isn't

None of the peptides discussed here is an FDA-approved drug. Semax and Selank are registered medicines in Russia but are not approved in the U.S. Nothing here supports giving any of these peptides to anyone, including children.

For Epitalon, FDA found no human safety data and flagged immune-reaction and product-quality concerns for injected forms (FDA 2026). FDA also raised a theoretical concern: because Epitalon is reported to activate telomerase, long-term use could in principle raise cancer risk. This has not been tested. An independent 2026 review called Epitalon "biologically active but clinically unproven" (Jain 2026).

If sleep, energy or gut comfort is a struggle, talk to a clinician.

I started this series because the most interesting peptide ideas are often untested, and the honest thing is to say so. The cell biology of autism is real, but the peptide research was done in other models (ageing cells, stroke, stress and Alzheimer's), not in studies designed to test these peptides in autism. Autistic adults deserve research built around their priorities. If this holds up, careful studies will show it. If not, that's worth knowing too.

— Julia

FAQ

Is this saying peptides treat autism?

No. It asks whether specific physiological difficulties that some autistic people experience, such as poor sleep or a stress system that stays switched on, could be studied properly, and whether short peptides belong in that research at all. None of the peptides here has been tested that way.

Have these peptides been studied in autistic people?

We found no published clinical studies testing the peptides reviewed here on their own in autistic participants, in searches of PubMed, ClinicalTrials.gov and the Russian-language database CyberLeninka on 18 October 2026. Three Russian clinical reports describe Semax given to autistic children alongside several other drugs, without testing it on its own. The only study we found that tested one of these peptides on its own in an autism model used rats exposed to valproate before birth: an animal model, not a clinical study (Malyshev 2013). Details are in the search log.

Should I try these peptides for sleep?

We don't recommend it. No study has shown these peptides improve sleep in anyone, there are no established human safety data, and Epitalon isn't approved for any use in the U.S. If sleep is a struggle, talk to a clinician. Melatonin and behavioural approaches have real research behind them.

Should I give these peptides to my autistic child?

No. We found no published clinical studies testing these peptides on their own in autistic children or adults, there are no reliable safety data in children, and a developing brain is exactly where effects on gene switching or nerve connections could carry risk. Talk to your child's doctor about sleep, gut comfort or any other concern.

What's the difference between Epitalon and Epithalamin?

Epithalamin is a peptide mix extracted from animal pineal glands; Epitalon is one synthetic four-amino-acid peptide designed from it. They're chemically different, so results for one don't automatically apply to the other (Jain 2026).

Full evidence and sources

How to read the evidence badges

Badges describe the type of study, not a single ladder of strength. A human observational study isn't necessarily stronger causal evidence than a well-controlled animal experiment, and a small human study can be weaker than a careful cell experiment on a narrow question.

Relevance to autism is a separate dimension, shown in its own column:

• Direct: autistic participants or tissue.
• Indirect: an autism-relevant animal model or mechanism.
• Unrelated model: ageing, stroke, stress or Alzheimer's research.
• Not yet tested: a gap, not a finding.

Study types used: Proposed · Modeled · Cell evidence · Animal evidence · Small or uncontrolled human data · Small human trial · Human observational data · Human post-mortem tissue · Human genetic data · Randomised human trial(s).

"Studied, not supported" marks a test that was run and didn't find the effect on that specific outcome.

Search log: clinical studies of these peptides in autistic people

English-language searches: Sunday 18 October 2026, 12:03–12:10 EDT. Russian-language searches: same day, 12:14–12:20 EDT. Autism terms for PubMed: (autism OR autistic OR "autism spectrum" OR ASD).

Database	Query	Hits
PubMed	(epitalon OR epithalon OR "Ala-Glu-Asp-Gly" OR AEDG) AND autism terms	0
PubMed	epithalamin AND autism terms	0
PubMed	("Glu-Asp-Arg" OR "EDR peptide" OR its trade name) AND autism terms	0
PubMed	("Lys-Glu" OR "KE peptide" OR its trade name) AND autism terms	0
PubMed	("Lys-Glu-Asp" OR "KED peptide") AND autism terms	0
PubMed	semax AND autism terms	0
PubMed	selank AND autism terms	0
PubMed	semax AND (valproic OR valproate)	2: the rat valproate study (Malyshev 2013) and an unrelated 1989 anticonvulsant study
PubMed	selank AND (valproic OR valproate); all other peptides AND (valproic OR valproate OR neurodevelopmental)	0
ClinicalTrials.gov (API v2)	condition "autism" with each term: epitalon, epithalon, AEDG, epithalamin, Glu-Asp-Arg, Lys-Glu, Lys-Glu-Asp, semax, selank, plus the trade names of EDR and KE	0 for every term

A broader PubMed search (semax AND neurodevelopmental terms) also returned one 2007 hypothesis paper proposing Semax for ADHD and Rett syndrome. It reports no data and isn't about autism (PMID 16996699).

Russian-language searches. Each peptide term was combined with аутизм, РАС and "расстройства аутистического спектра". CyberLeninka searches full text, so "hits" include any article mentioning both words anywhere. We screened the top 10 results of each search by title and, where relevant, full text. "РАС" also matches unrelated abbreviations (for example, academy names), which inflates counts.

Database	Peptide term	Hits (аутизм / РАС / "расстройства аутистического спектра")	Relevant to autistic participants
CyberLeninka	эпиталон (Epitalon)	0 / 1 / 0	None (the 1 hit is a gerontology overview)
CyberLeninka	эпиталамин (Epithalamin)	0 / 4 / 0	None (unrelated)
CyberLeninka	the EDR trade name, in Russian	0 / 1 / 0	None (an antioxidant review)
CyberLeninka	селанк (Selank)	1 / 3 / 1	None (rat stress study, conference report, unrelated reviews)
CyberLeninka	семакс (Semax)	29 / 851 / 7	No study testing Semax itself. See the four reports below.
eLibrary.ru	эпиталон аутизм; семакс аутизм	0; 0	Only these two searches completed. Later searches returned "403 Forbidden", so eLibrary.ru was not searched systematically.
Google Scholar	эпиталон аутизм; селанк аутизм; семакс аутизм; семакс "расстройства аутистического спектра"	0; 0; 3 (none about Semax in autistic people); 0	Reached only through a limited fetch; treat as a spot check, not a systematic search.

What the Semax hits actually are.

• Malyshev AV, et al., 2015, Вестник Московского университета. Серия 16. Биология, no. 3: a review of the rat valproate model by the same first author as Malyshev 2013. The Semax data come from rats, not people. https://cyberleninka.ru/article/n/eksperimentalnaya-model-autisticheskogo-rasstroystva-indutsirovannyy-fetalnyy-valproatnyy-sindrom
• Simashkova NV, et al., 2012, Неврология, нейропсихиатрия, психосоматика, suppl. 2: a clinical description of children with autism and hyperactivity. It lists Semax among several neuroprotective drugs combined with antipsychotics. This is not a test of Semax. https://cyberleninka.ru/article/n/multidistsiplinarnaya-kliniko-biologicheskaya-harakteristika-giperkineticheskih-rasstroystv-pri-detskom-autizme
• Kalinina MA, et al., 2016, Современная терапия в психиатрии и неврологии, no. 1: a study of other drugs in autistic children. Semax appears only as part of earlier background treatment in the comparison group. https://cyberleninka.ru/article/n/kompleksnoe-lechenie-rasstroystv-autisticheskogo-spektra-malymi-dozami-risperidona-i-memantina-u-detey
• Maksimova AA, 2020, Universum: медицина и фармакология, no. 7(69): 71 autistic children at one clinic in Georgia. Semax was one of several components of a combined course, with improvement reported by parents. There was no control group and no way to separate Semax's contribution. https://cyberleninka.ru/article/n/svyaz-somaticheskih-patologicheskih-izmeneniy-s-negativnym-povedeniem-detey-s-rasstroystvom-autichnogo-spektra

None of these tests a reviewed peptide on its own in autistic participants. That is why the article says "on their own" and doesn't claim the peptides have never been given to autistic people.

eLibrary.ru was searched only partially (access blocked after two queries), so the Russian search is incomplete. Not searched: the KE trade name in Russian, and Russian clinical-trial registries.

Technical notes

• Absorption. No human absorption data have been identified for any route (Jain 2026; FDA 2026).
• Routes. Routes are given only as each source states them. "Route not stated" means the abstract doesn't say; we never infer one. Results from a dish, an injection and under-the-tongue use don't carry over to each other.
• Who did the work. Most peptide studies come from the developers' St. Petersburg network or the Moscow Semax/Selank group. Djeridane 2003 includes Khavinson and Anisimov as co-authors, and Kraskovskaya 2024 includes authors from the developers' institute, so neither null result is fully independent.
• Kozina 2007. Both peptides showed antioxidant activity; only Epithalamin raised antioxidant enzymes.
• Ivko 2020 and 2021. The 2020 paper (PMID 33342107) is a Russian-language cell study. The sublingual melatonin study is cited as the FDA summarised it (Ivko 2021, Adv Gerontol 11(1):53–58).
• Malyshev 2013 and Wan 2024. Malyshev has no abstract on PubMed, so we couldn't check design or route. Wan 2024 has no PubMed abstract; we checked it against the full text.
• Yasenyavskaya 2021. PubMed indexes the first author as "Leonidovna YA" (a patronymic).
• Pagan 2014. The 51% figure uses a 5th-percentile cut-off; insomnia was self-reported, and a children's sleep questionnaire showed no association.

Sources

Framing and autism biology

1. Kapp SK, et al. Deficit, difference, or both? Autism and neurodiversity. Dev Psychol. 2013. PMID 22545843

2. Pellicano E, et al. What should autism research focus upon? Community views and priorities from the United Kingdom. Autism. 2014. PMID 24789871

3. Pellicano E, den Houting J. Shifting from "normal science" to neurodiversity in autism science. J Child Psychol Psychiatry. 2022. PMID 34730840

4. Irwin MR, et al. Sleep disturbance, sleep duration, and inflammation: a systematic review and meta-analysis of cohort studies and experimental sleep deprivation. Biol Psychiatry. 2016. PMID 26140821

5. Rossignol DA, Frye RE. Mitochondrial dysfunction in autism spectrum disorders: a systematic review and meta-analysis. Mol Psychiatry. 2012. PMID 21263444

6. Weissman JR, et al. Mitochondrial disease in autism spectrum disorder patients: a cohort analysis. PLoS One. 2008. PMID 19043581

7. Frye RE, et al. Biomarkers of mitochondrial dysfunction in autism spectrum disorder: a systematic review and meta-analysis. Neurobiol Dis. 2024. PMID 38703861

8. Frustaci A, et al. Oxidative stress-related biomarkers in autism: systematic review and meta-analyses. Free Radic Biol Med. 2012. PMID 22542447

9. Chen L, et al. Oxidative stress marker aberrations in children with autism spectrum disorder: a systematic review and meta-analysis. Transl Psychiatry. 2021. PMID 33414386

10. Vargas DL, et al. Neuroglial activation and neuroinflammation in the brain of patients with autism. Ann Neurol. 2005. PMID 15546155

11. Morgan JT, et al. Microglial activation and increased microglial density observed in the dorsolateral prefrontal cortex in autism. Biol Psychiatry. 2010. PMID 20674603

12. Voineagu I, et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature. 2011. PMID 21614001

13. Masi A, et al. Cytokine aberrations in autism spectrum disorder: a systematic review and meta-analysis. Mol Psychiatry. 2015. PMID 24934179

14. Suzuki K, et al. Microglial activation in young adults with autism spectrum disorder. JAMA Psychiatry. 2013. PMID 23404112

15. Zürcher NR, et al. [11C]PBR28 MR-PET imaging reveals lower regional brain expression of translocator protein (TSPO) in young adult males with autism spectrum disorder. Mol Psychiatry. 2021. PMID 32076115

16. Jiang HY, et al. Maternal infection during pregnancy and risk of autism spectrum disorders: a systematic review and meta-analysis. Brain Behav Immun. 2016. PMID 27287966

17. Choi GB, et al. The maternal interleukin-17a pathway in mice promotes autism-like phenotypes in offspring. Science. 2016. PMID 26822608

18. Kitano H. Biological robustness. Nat Rev Genet. 2004. PMID 15520792

19. McEwen BS. Protective and damaging effects of stress mediators. N Engl J Med. 1998. PMID 9428819

Autonomic and stress physiology

20. Cheng YC, et al. Heart rate variability in individuals with autism spectrum disorders: a meta-analysis. Neurosci Biobehav Rev. 2020. PMID 32818581
21. Taylor JL, Corbett BA. A review of rhythm and responsiveness of cortisol in individuals with autism spectrum disorders. Psychoneuroendocrinology. 2014. PMID 25108163
22. Hadwin JA, et al. Cortisol awakening response in children and adolescents with autism spectrum disorder: a systematic review and meta-analysis. Evid Based Ment Health. 2019. PMID 31253603
23. Mazurek MO, Petroski GF. Sleep problems in children with autism spectrum disorder: examining the contributions of sensory over-responsivity and anxiety. Sleep Med. 2015. PMID 25600781
24. Lai MC, et al. Prevalence of co-occurring mental health diagnoses in the autism population: a systematic review and meta-analysis. Lancet Psychiatry. 2019. PMID 31447415

Genes and epigenetics

25. De Rubeis S, et al. Synaptic, transcriptional and chromatin genes disrupted in autism. Nature. 2014. PMID 25363760
26. Satterstrom FK, et al. Large-scale exome sequencing study implicates both developmental and functional changes in the neurobiology of autism. Cell. 2020. PMID 31981491
27. Fu JM, et al. Rare coding variation provides insight into the genetic architecture and phenotypic context of autism. Nat Genet. 2022. PMID 35982160
28. Abrahams BS, et al. SFARI Gene 2.0: a community-driven knowledgebase for the autism spectrum disorders (ASDs). Mol Autism. 2013. PMID 24090431
29. Bernier R, et al. Disruptive CHD8 mutations define a subtype of autism early in development. Cell. 2014. PMID 24998929
30. Gandal MJ, et al. Broad transcriptomic dysregulation occurs across the cerebral cortex in ASD. Nature. 2022. PMID 36323788
31. Ladd-Acosta C, et al. Common DNA methylation alterations in multiple brain regions in autism. Mol Psychiatry. 2014. PMID 23999529
32. Wong CC, et al. Methylomic analysis of monozygotic twins discordant for autism spectrum disorder and related behavioural traits. Mol Psychiatry. 2014. PMID 23608919
33. Andrews SV, et al. Case-control meta-analysis of blood DNA methylation and autism spectrum disorder. Mol Autism. 2018. PMID 29988321

Sleep and melatonin

34. Morgan B, et al. Sleep in adults with autism spectrum disorder: a systematic review and meta-analysis. Sleep Med. 2020. PMID 31739229
35. Pagan C, et al. The serotonin-N-acetylserotonin-melatonin pathway as a biomarker for autism spectrum disorders. Transl Psychiatry. 2014. PMID 25386956
36. Melke J, et al. Abnormal melatonin synthesis in autism spectrum disorders. Mol Psychiatry. 2008. PMID 17505466
37. Jonsson L, et al. Mutation screening of melatonin-related genes in patients with autism spectrum disorders. BMC Med Genomics. 2010. PMID 20377855
38. Rossignol DA, Frye RE. Melatonin in autism spectrum disorders: a systematic review and meta-analysis. Dev Med Child Neurol. 2011. PMID 21518346
39. Gringras P, et al. Efficacy and safety of pediatric prolonged-release melatonin for insomnia in children with autism spectrum disorder. J Am Acad Child Adolesc Psychiatry. 2017. PMID 29096777
40. Galli-Carminati G, et al. Melatonin in treatment of chronic sleep disorders in adults with autism: a retrospective study. Swiss Med Wkly. 2009. PMID 19452292
41. Keogh S, et al. Effectiveness of non-pharmacological interventions for insomnia in children with autism spectrum disorder: a systematic review and meta-analysis. PLoS One. 2019. PMID 31437211

Gut and brain

42. Yap CX, et al. Autism-related dietary preferences mediate autism-gut microbiome associations. Cell. 2021. PMID 34767757
43. Wan L, et al. Effect of oral faecal microbiota transplantation intervention for children with autism spectrum disorder: a randomised, double-blind, placebo-controlled trial. Clin Transl Med. 2024. PMID 39187939

Peptide research

44. Ivko OM, et al. The influence of AEDG and KE peptides on mitochondria staining and L7A ribosomal protein expression during human pineal gland and thymus cell senescence in vitro. Adv Gerontol. 2020 (in Russian). PMID 33342107
45. Kraskovskaya N, et al. Short peptides protect fibroblast-derived induced neurons from age-related changes. Int J Mol Sci. 2024. PMID 39518916
46. Kozina LS, et al. Antioxidant properties of geroprotective peptides of the pineal gland. Arch Gerontol Geriatr. 2007. PMID 17317455
47. Khavinson V, et al. Pinealon increases cell viability by suppression of free radical levels and activating proliferative processes. Rejuvenation Res. 2011. PMID 21978084
48. Arutjunyan A, et al. Pinealon protects the rat offspring from prenatal hyperhomocysteinemia. Int J Clin Exp Med. 2012. PMID 22567179
49. Safarova ER, et al. Effect of Semax peptide on survival of cultured rat pheochromocytoma cells during oxidative stress. Bull Exp Biol Med. 2003. PMID 12802399
50. Tomasello MF, et al. Semax, a copper chelator peptide, decreases the Cu(II)-catalyzed ROS production and cytotoxicity of aβ by metal ion stripping and redox silencing. Bioinorg Chem Appl. 2025. PMID 40496623
51. Kazakova TB, et al. In vitro effect of short peptides on expression of interleukin-2 gene in splenocytes. Bull Exp Biol Med. 2002. PMID 12447482
52. Kazakova TB, et al. Synthesis of IL-2 mRNA in cells of rat hypothalamic structures after injection of short peptides. Bull Exp Biol Med. 2005. PMID 16224591
53. Mendzheritskii AM, et al. Regulation of content of cytokines in blood serum and of caspase-3 activity in brains of old rats in model of sharp hypoxic hypoxia [title abridged]. Adv Gerontol. 2014 (in Russian). PMID 25051764
54. Yasenyavskaya AL, et al. The influence of Selank on the level of cytokines under the conditions of "social" stress. Curr Rev Clin Exp Pharmacol. 2021. PMID 32621722 (PubMed indexes the first author as "Leonidovna YA")
55. Medvedeva EV, et al. The peptide Semax affects the expression of genes related to the immune and vascular systems in rat brain focal ischemia: genome-wide transcriptional analysis. BMC Genomics. 2014. PMID 24661604
56. Dergunova LV, et al. Semax suppresses mRNA transcripts encoding proinflammatory mediators induced by reversible ischemia of the rat brain [title abridged]. Mol Biol (Mosk). 2021 (in Russian). PMID 34097675
57. Khavinson VKh, et al. Peptide Epitalon activates chromatin at the old age. Neuro Endocrinol Lett. 2003. PMID 14647006
58. Khavinson VKh, et al. Effects of short peptides on lymphocyte chromatin in senile subjects. Bull Exp Biol Med. 2004. PMID 15085253
59. Fedoreyeva LI, et al. Penetration of short fluorescence-labeled peptides into the nucleus in HeLa cells and in vitro specific interaction of the peptides with deoxyribooligonucleotides and DNA. Biochemistry (Mosc). 2011. PMID 22117547
60. Khavinson V, et al. DNA double-helix binds regulatory peptides similarly to transcription factors. Neuro Endocrinol Lett. 2005. PMID 15990728
61. Khavinson V, et al. AEDG peptide (Epitalon) stimulates gene expression and protein synthesis during neurogenesis: possible epigenetic mechanism. Molecules. 2020. PMID 32019204
62. Kraskovskaya NA, et al. Tripeptides restore the number of neuronal spines under conditions of in vitro modeled Alzheimer's disease. Bull Exp Biol Med. 2017. PMID 28853087
63. Khavinson V, et al. Neuroprotective effects of tripeptides-epigenetic regulators in mouse model of Alzheimer's disease. Pharmaceuticals (Basel). 2021. PMID 34071923
64. Malyshev AV, et al. Semax corrects brain dysfunction caused by prenatal introduction of valproic acid. Dokl Biol Sci. 2013. PMID 23821048
65. Khavinson VKh, et al. Molecular cellular mechanisms of peptide regulation of melatonin synthesis in pinealocyte culture. Bull Exp Biol Med. 2012. PMID 22816096
66. Djeridane Y, et al. Effect of a synthetic pineal tetrapeptide (Ala-Glu-Asp-Gly) on melatonin secretion by the pineal gland of young and old rats. J Endocrinol Invest. 2003. PMID 12809170
67. Korkushko OV, et al. Effect of peptide preparation epithalamin on circadian rhythm of epiphyseal melatonin-producing function in elderly people. Bull Exp Biol Med. 2004. PMID 15452611
68. Labunets IF, et al. Effect of epithalamin on circadian relationship between the endocrine function of the thymus and melatonin-producing function of the pineal gland in elderly people. Bull Exp Biol Med. 2004. PMID 15455130
69. Korkushko OV, et al. Normalizing effect of the pineal gland peptides on the daily melatonin rhythm in old monkeys and elderly people. Adv Gerontol. 2007 (in Russian). PMID 17969590
70. Goncharova ND, et al. Regulatory effect of Epithalon on production of melatonin and cortisol in old monkeys. Bull Exp Biol Med. 2001;131(4):394–6. PMID 11550036
71. Khavinson VKh, et al. Short peptides stimulate serotonin expression in cells of brain cortex. Bull Exp Biol Med. 2014. PMID 24909721
72. Khavinson VKh, et al. Effect of Vilon and Epithalon on glucose and glycine absorption in various regions of small intestine in aged rats. Bull Exp Biol Med. 2002. PMID 12420071
73. Mukhina AY, et al. State of colon microbiota in rats during chronic restraint stress and Selank treatment. Bull Exp Biol Med. 2019. PMID 31236882

Reviews and regulators

74. Jain A, et al. Epitalon and Epithalamin: a narrative review of mechanistic, preclinical, and clinical evidence. Int J Mol Sci. 2026 Oct 8;27(19):8922. doi:10.3390/ijms27198922. PMID 42884164
75. US Food and Drug Administration. FDA Briefing Document, Pharmacy Compounding Advisory Committee (PCAC) Meeting, July 23–24, 2026: Epitalon-Related Bulk Drug Substances (Epitalon (Free Base) and Epitalon Acetate). FDA evaluation dated 12 May 2026. https://www.fda.gov/media/193345/download (includes the summary of Ivko OM, et al. Adv Gerontol. 2021;11(1):53–58)

The Short Peptide publishes research explainers for general education. This is not medical advice. Short Hypothesis pieces describe untested hypotheses. It does not recommend any product.