Most compounds in this catalogue have a dose figure whose problem is provenance — a monkey number quoted as a human one, a trial protocol read as a recommendation. Dihexa's problem is different and simpler.
There is no human dose, because there has never been a human trial. And the mechanism the compound is sold on rests in large part on three papers that the journal that published them withdrew in April 2025.
The registry record
A ClinicalTrials.gov search on 20 September 2026:
| Search term | Studies returned |
|---|---|
dihexa |
0 |
PNB-0408 (development code) |
0 |
N-hexanoic-Tyr-Ile-(6)-aminohexanoic amide |
0 |
This site has learned to search development codes as well as market names — adipotide's only human trial is filed under Prohibitin Targeting Peptide 1 and is invisible to a search for its market name, which is documented on that compound's page. Dihexa was searched the same three ways and the answer does not change.
There is no approved product in any jurisdiction, so there is no label either. The result is that no source class this site accepts contains a human dose figure for this compound, and none ever has.
The retractions
In April 2025 the Journal of Pharmacology and Experimental Therapeutics retracted three papers from the Washington State University laboratory that developed the compound. The notices appeared in the same issue, in consecutive numbers:
| Retracted paper | Original publication | Retraction notice |
|---|---|---|
| Mimics of the dimerization domain of hepatocyte growth factor exhibit anti-Met and anticancer activity | JPET 2011;339(2):509–18 | JPET 2025;392(4):103568 |
| Development of angiotensin IV analogs as hepatocyte growth factor/Met modifiers | JPET 2012;340(3):539–48 | JPET 2025;392(4):103566 |
| The procognitive and synaptogenic effects of angiotensin IV-derived peptides are dependent on activation of the hepatocyte growth factor/c-Met system | JPET 2014;351(2):390–402 | JPET 2025;392(4):103567 |
These are three of the four papers the laboratory published in that journal between 2011 and 2014. All four received Notices of Concern in September 2021, so the withdrawal followed about three and a half years of formal, public doubt; the fourth, from 2013, still carries only its notice. Each retracted record is now flagged in PubMed with the publication type Retracted Publication.
The 2014 paper is the one that matters commercially. It established dihexa's mechanism of action: that the compound binds hepatocyte growth factor with high affinity, induces c-Met phosphorylation, produces hippocampal spinogenesis and synaptogenesis comparable to HGF itself, and that its cognitive effect in the Morris water maze was abolished by an HGF antagonist delivered into the brain.
That is the mechanism dihexa is marketed on, and it is the paper that is no longer in the literature.
The full text of the notices, the Washington State University investigation that produced them — which found falsified and/or fabricated data and named two individuals as solely responsible — and the subsequent False Claims Act settlement are set out on this site's cognitive peptides page, and are not restated here.
What a retraction does and does not establish
It is easy to overread this, and worth being precise.
A retraction withdraws specific papers. It does not run a new experiment, and it does not demonstrate that a compound is inert. Unretracted animal work on dihexa exists, and it is listed below.
What a retraction does remove is the published evidentiary basis for the claims those papers made. When a retail page says dihexa works through the HGF/c-Met system and cites the synaptogenesis result, it is citing a withdrawn paper — usually without knowing it, because the pages were written before 2025 and nobody revisits them.
There is also one place where the hypothesis was tested rather than argued about. Fosgonimeton, the clinical successor built from the same laboratory's work, ran a Phase 2/3 trial in mild-to-moderate Alzheimer's disease; topline results reported on 3 September 2024 state that in 312 patients the primary endpoint was not met (p=0.70), with neither key secondary endpoint reaching significance. That is the closest thing to a clinical test this mechanism has had, and it is a null result — which is a stronger statement about the compound than any retraction, because it is a measurement rather than a withdrawal.
The second retracted paper also contained the only circulating blood half-life figure for this chemical family — 80 minutes for a stabilised 6-AH analogue against under 5 minutes for the parent compound. Because that figure now sits inside a retracted paper, it is recorded here as withdrawn rather than reported as a fact, and it does not appear on this site's half-life chart.
What the remaining literature holds
A PubMed search for dihexa on 20 September 2026 returns 17 records. That count overstates the file in two directions.
Several records are substring artefacts — a 1992 study of surfactant clearance in rabbit lung cells, a porphycene subcellular localisation study, a 2014 paper on phase-segregated polymerizable lipids — matched on chemistry that has nothing to do with this compound. This is the same counting trap documented on the adipotide page, where only four of seven records were about the compound.
Three more are the retracted papers.
What remains is a small, real set of animal studies:
| Study | Species and model | What it used |
|---|---|---|
| Brain Sci, 2021;11(11):1487 | APP/PS1 transgenic mice (Alzheimer's model) | Dihexa administered orally, reporting improved cognition via PI3K/AKT signalling |
| Ann Med Surg, 2021;71:102917 | Rat sciatic nerve damage-repair model | Dihexa alongside stem cells and G-CSF |
| J Huntingtons Dis, 2024;13(1):55–66 | Rats given 3-nitropropionic acid | The compound under its PNB-0408 code |
| Cell Prolif-adjacent work, 2022 | Human pluripotent stem cells, in vitro | Hepatocyte differentiation |
Rodent efficacy in a disease model is where most compounds begin. It is also where the large majority of them stop, and dihexa has not moved past it in the fifteen years since the first of the retracted papers.
Why the dose figures circulating cannot be converted
Retail pages for dihexa quote oral figures in the single-digit milligram range. Whatever their arithmetic, the problem is upstream of it.
A human dose is not derived from a rodent dose by scaling. It is established by a dose-finding study: a Phase 1 design that administers ascending doses to people, measures what the compound does in blood, and identifies a range that is tolerated. That study is what the whole first tier of this site's dosage chart is built on, and it has never been run for this compound — not under the name dihexa, not under PNB-0408, not under its chemical name.
The honest description of every dihexa dose figure in circulation is that it is an inference from animal work, made by somebody selling the compound, about a mechanism established in papers that have since been retracted.
Method
The registry census was run against the ClinicalTrials.gov v2 API on 20 September 2026 with three search terms. The literature census was run against PubMed on the same date. Retraction and Expression of Concern notices were read from the PubMed records of the three original papers, each of which names its notice by volume, issue and article number.
Nothing on this page is a dosing instruction, and the absence of a dose figure here is the finding rather than a gap to be filled from elsewhere.