Most investors read the story. The factory reads the lab data behind the story. When the data doesn't match the story, you know before the deal — not after.
There is a real natural protein called FSTL1 that the heart produces after injury. Peer-reviewed science going back years — and earlier materials Sean Edwards shared in 2026 — shows this protein can help heart cells recover after a heart attack. That science is solid.
Regencor's drug, REG101, was engineered to deliver the same benefit as a pharmaceutical. In April 2026, Regencor provided a 213-page lab package — the first detailed, independent test of whether REG101 actually works the way the natural protein does. This document is built from that package. It is the most current, compound-specific data available. The question it answers is simple: does the drug work the way they say? Three independent labs gave the same answer.
The natural FSTL1 protein works by blocking a specific protein called BMP4 that causes damage to the heart after a heart attack. Regencor engineered REG101 to do the same thing in drug form — block BMP4, reduce damage, protect the heart.
The story is internally consistent. There's published science behind the natural protein. The mechanism has a name, a target, and a logical chain from lab to patient.
Three separate labs — all paid by the company — each ran a different type of test to check whether REG101 actually blocks BMP4 under conditions that match a real human body. All three found it doesn't.
Not borderline. Not inconclusive. Each lab used a different testing method. Each arrived at the same result: no meaningful blocking activity at normal body conditions. The natural FSTL1 protein they engineered this from? It works fine in those same conditions. REG101 does not replicate it.
This isn't a matter of interpretation. Three independent labs, three different test types, one consistent answer.
This test measures how tightly REG101 latches onto BMP4 — the protein it's supposed to block. A drug that can't physically grab its target can't stop it from doing damage.
At normal body conditions: the grip was 100 to 300 times too weak to be functional. Published science on natural FSTL1 shows it binds roughly 300 times more tightly. REG101 isn't in the same ballpark.
Full analysis — Kill Condition KC-02 Lab report pp. 10–20This test goes further — rather than measuring grip in isolation, it puts REG101 on actual living human cells and watches whether it interacts with BMP4 on the cell surface. More realistic conditions.
Result: no meaningful interaction on living cells. The document's own words: "No significant interaction was observed between rhFSTL1 and BMP4."
Full analysis — KC-02 / Sycophancy flag Table 3 Lab report p. 68The most direct test: forget binding and interaction — does REG101 actually prevent BMP4 from triggering the cellular chain reaction it's supposed to stop? At every dose tested, the cellular signal did not change at all. The lab couldn't even calculate a working dose because there was no effect to measure.
Full analysis — New claim: BMP4 SMAD null (Table 2) Lab report p. 168, Table 3.2.15The mechanism the company is selling doesn't hold up in their own lab data.
Every investor reading the narrative is evaluating a story built on a mechanism that three independent labs failed to confirm. You're evaluating the data. That's a fundamentally different position to be in before any money moves.
A 213-page lab package covering four distinct types of testing: how tightly the drug binds to proteins, how it behaves on live human cells, an industry-standard test across 12 different human cell types, and three rounds of functional testing spanning four months.
On the surface, it looks thorough. Multiple labs. Multiple methods. A progression of work over time.
Not a single one of the 12 cell types tested in the broad profiling section is a heart cell. The 12 types covered: blood vessel cells, skin cells, lung cells, and immune cells. For a drug being positioned as a cardiac therapy, no heart muscle cells, no heart tissue cells, no cardiac tissue of any kind was included in the largest section of the package.
That's not a minor gap. It means the entire profiling section — the most expensive and comprehensive part of the package — tells you nothing about what this drug does to the organ it's supposed to treat.
And when the profiling system searched its reference database for the drug that most closely resembles REG101's behavior pattern, the closest match was a cancer drug used for blood cancers — not any heart drug.
Cell types tested: Blood vessel cells (5 types), skin cells (2 types), lung cells (2 types), immune cells (3 types).
Heart cells included: zero. No heart muscle cells. No heart tissue cells. No cardiac fibroblasts. No myocardial tissue of any kind.
When the profiling system identified the closest drug in its reference library to REG101's behavior pattern: Pirtobrutinib — a drug used to treat B-cell blood cancers, not heart disease. Pearson correlation r=0.729, above the 0.7 threshold the test uses to call something a mechanistic match.
Full analysis — KC-06 + Sycophancy flag: "comprehensive profiling" (Table 3) Lab report pp. 73–98, p. 80A comprehensive lab package that never tested a heart cell is not comprehensive for a cardiac drug.
This isn't something a follow-up study patches. It means you're being asked to evaluate a heart therapy based on a package that has no information about what it does to the heart.
Three rounds of functional testing from February through April 2026. Multiple studies from multiple points in time. The presentation implies the science is building — that each study adds another layer of support.
The third study is a hidden failure. It is presented as a third data point, but it isn't one.
In February, a benchmark compound — a known-working drug used to validate the entire test — showed strong, measurable results at a very low dose. In April, that same benchmark compound was retested at a dose 200 times higher. It showed zero activity. Completely indistinguishable from giving nothing at all.
When a known-working drug shows no activity at 200 times its working dose, the test itself has failed — or the compounds degraded and became inactive between February and April. Either way, Study 3 is a broken replication attempt, not a third data point. The package includes it without flagging the failure anywhere.
The benchmark compound used to verify the test is working showed strong, measurable blocking activity at a very low dose. This is the positive control — the standard every valid study must pass to show the test itself is functioning.
Lab report — Study 1 dataZero activity. The document notes: "No IC50 values were calculated in this assay" — meaning there was no detectable dose at which anything happened. A compound that worked strongly in February at a tiny dose showed no effect in April at 200 times that dose.
Study 3 data is included in the package as if it contributes to the evidence. The benchmark failure is not flagged or explained.
Full analysis — Replication failure (Table 2) + Sycophancy flag (Table 3) Lab report pp. 172–213, esp. p. 179The clean science ended in February. What looks like three rounds of accumulating evidence is actually one solid study and a failed attempt to reproduce it — with the failure undisclosed.
Presenting three studies as accumulating support when one is a broken replication is a material misrepresentation of the data package.
Some tests did show REG101 crossing the minimum detection threshold for BMP4 interaction. These results are presented as meaningful positive findings that support the mechanism story.
Every positive result required either chemically preserved dead cells — not living cells — or extreme chemical conditions more acidic than anything in a real human heart, including a damaged one.
When the exact same experiments were run on living cells under normal body chemistry, every single positive result disappeared. The drug is also weaker than the natural protein it was engineered to improve upon — even under the most favorable test conditions, the natural protein outperformed REG101 by nearly twice the effect. And even at those extreme favorable conditions, REG101's grip on BMP4 was still 100 times weaker than published science says is needed for it to be functional.
The positive signals exist. They just don't survive contact with conditions that match a real human body.
When cells are chemically preserved — essentially dead and fixed in place, which opens up binding sites that don't exist on living cells — REG101 just barely crossed the 2.0 significance threshold for BMP4 interaction (fold: 2.3).
Same test on living cells: fold dropped to 1.2–1.3 — not significant. The interaction disappeared when the cells were alive.
Additional note: in the preserved-cell test, the natural FSTL1 protein showed fold 4.1 — nearly twice REG101's result. The drug is weaker than the natural protein it was designed to improve on, even in the most favorable possible conditions.
Full analysis — Sycophancy flag: BMP4 fold framing (Table 3) Lab report p. 66 (preserved), p. 68 (living)At a very high acidity level — more acidic than resting heart tissue, and more acidic than even acutely injured heart tissue in most conditions — REG101 showed some binding activity.
At normal body chemistry: binding strength drops 100–300 times below the functional threshold. Even at the extreme acidity where the drug performs best, it is still 100 times weaker than the published standard for what a functional FSTL1 interaction looks like.
Full analysis — Sycophancy flag: SPR binding at pH 6.0 (Table 3) Lab report pp. 20–30, 59, 70The positive results are artifacts of the test conditions — they vanish when you run the same test on living cells at normal body chemistry.
Any investor citing "positive binding findings" is citing results that don't survive physiological conditions. You know to ask: under what conditions? That question changes the entire conversation.
You evaluate a 213-page lab package that looks like a serious science program. You see a plausible story — the natural protein is real, the target is real, there's IP, Japanese government grant money, multiple labs, multiple studies. The risk looks like clinical translation: can it survive trials?
That's where most diligence stops.
You see that the drug doesn't do what they say it does, confirmed by their own labs. The risk isn't clinical translation — the mechanism hasn't been established at the bench.
The question isn't "can this get through trials?" The question is: "what does this drug actually do, and is there any cardiac indication that can be salvaged from what the data does show?"
The lab data does show partial blocking of two other proteins — Activin and GDF-8/Myostatin — at high doses. Both of those proteins are independently linked to scarring and muscle wasting in heart failure. That's a narrower but real target.
Whether REG101 can block those proteins at doses that are safe and achievable in a human being is still an open question — the current data used doses far above what is normally therapeutic, and the effect was partial even then. This needs its own study at standard doses before any cardiac claim can be constructed around it.
This is a fundable conversation — but only after the company acknowledges that the BMP4 story is not supported by their own data, and redirects to what the data does actually suggest. Knowing that before the deal gives you the leverage to have the right conversation at the right moment.