$SLS SELLAS Life Sciences: the disease, the drugs in use today, and what REGAL is actually testing

01What acute myeloid leukemia actually is

Acute myeloid leukemia is a cancer of the blood-forming cells inside the bone marrow. In a
healthy marrow, immature cells called myeloblasts mature into the red cells that carry oxygen,
the platelets that stop bleeding, and several kinds of white cells that fight infection. In AML
that maturation stops. The blasts keep dividing but never finish growing up, they crowd out the
marrow, and the patient runs out of the mature blood cells that keep a body alive. That is why
the disease presents as anemia, bruising, bleeding and infection rather than as a lump.

The numbers set the frame for everything that follows. The National Cancer Institute’s SEER
program estimates 22,720 new cases and 11,500 deaths in the United States in 2026. Relative
five-year survival is 33.4% for cases diagnosed between 2016 and 2022. The median age at
diagnosis is 70 years, and roughly six patients in ten are 65 or older when the disease is
found. That last figure does more work than it looks like: it is the reason a large share of AML
patients cannot be offered the most aggressive treatments, and the reason a drug that can be given
gently, at home, to an older person in remission has commercial value out of proportion to the
size of the population.

Survival has improved, but slowly and from a very low base. The same SEER series shows relative five-year survival rising from 5.4% in 1975 to 33.2% in 2018. Four decades of research
moved one patient in three from dying within five years to being alive at five years. Against that
history, a drug that adds months rather than years is not a disappointment. It is the normal size
of progress in this disease.

02Remission, CR1, CR2: the words that decide everything

Three words decide everything that follows, and they are worth getting exactly right, because companies and press releases use them loosely and the difference between them is
the difference between a large market and a small one.

Complete remission, or CR. After induction chemotherapy, doctors check the marrow. A
complete remission means fewer than 5% blasts in the bone marrow, no evidence of disease elsewhere,
and recovery of the blood counts in the peripheral circulation, conventionally platelets above
100,000 per microliter and neutrophils above 1,000. That definition is not marketing language: it
is written into the FDA labels of approved AML drugs. Remission is not cure. It means the disease
has fallen below the threshold at which a microscope can find it.

CRp, or complete remission without platelet recovery. Same clearance of blasts, but the
platelets have not come back to normal. The REGAL protocol sets its bar at platelets above 20,000 per microliter with neutrophils above 1,000, no Auer rods,
no circulating blasts and no disease outside the marrow. A patient in CRp has beaten the leukemia
back but has not fully rebuilt the marrow.

CR1 and CR2. These are counters, not categories. CR1 is the remission achieved with
first-line treatment. If the disease comes back and a second, harder round of therapy pushes it
down again, that is CR2, a second remission after a relapse. The distinction is the whole subject
here, because every approved maintenance drug in AML is approved in CR1, and the trial that will
decide SELLAS Life Sciences’ near future is run in CR2.

The reason the distinction matters clinically is that relapse changes the disease. A leukemia
that has already survived one full course of chemotherapy has been selected for resistance. The
second remission is usually shorter and harder to hold than the first, and the patient reaching it
is a year or two older, more heavily pre-treated, and frequently in worse physical condition than
the person who walked in at diagnosis.

03What happens when the leukemia comes back

Relapse is the rule, not the exception. In a pooled analysis of 3,012 patients enrolled across
nine ECOG-ACRIN cooperative-group studies between 1984 and 2008, 1,779 patients, or 59.1%, reached
a first complete remission. Of those who reached CR1, 58.9% relapsed. Roughly six in ten
patients who get the good news get the bad news later.

What follows relapse is the harshest number in this disease. In that same analysis, median
survival measured from the moment of relapse was 0.5 years, and five-year overall survival was
10%, with a median follow-up approaching a decade. The authors found no improvement across
successive eras of the study. A separate European series of 667 younger relapsed patients, from
the HOVON and SAKK groups, splits the outcome by risk group: five-year survival after first
relapse was 46% in the favorable group, 18% in the intermediate group and 4% in the adverse
group.

Those two studies are why the second remission is such a strange and precarious place to stand.
The patient in CR2 has done something unusual. They relapsed, which most patients do, and then
responded again to salvage therapy, which many do not. They are, at that moment, free of
detectable disease. And they are sitting on top of a disease that has already proven twice that it
knows how to come back.

04The transplant: the one curative option, and who never gets it

There is one treatment in AML that is genuinely curative for some patients, and it is not a
drug. An allogeneic stem cell transplant replaces the patient’s diseased marrow with
blood-forming cells from a healthy donor. The therapeutic effect is only partly the high-dose
chemotherapy that clears the way. The larger part is immunological: the donor’s transplanted immune
system recognizes surviving leukemia cells as foreign and attacks them, an effect clinicians call
graft-versus-leukemia.

Transplant in second remission works, in the sense that it can produce long-term survivors.
A registry analysis of 4,682 transplant recipients found five-year survival of 39% for patients
transplanted in CR2, compared with 18% for those transplanted in first relapse and 21% for
those transplanted after primary induction failure. Nothing else in AML offers a number like that
to a patient who has already relapsed once.

The catch is access. A transplant is one of the most physically brutal procedures in medicine.
It requires a matched donor, an organ system able to survive the conditioning regimen, and a
patient able to withstand months of immune suppression, infection risk and graft-versus-host
disease. In a series of 1,073 patients aged 60 or older newly diagnosed at MD Anderson between
2012 and 2021, 18% received a transplant — 37% among those treated with intensive chemotherapy, 22% among those given low-intensity therapy with venetoclax, and 10% among those given low-intensity therapy alone. MD Anderson is one of the
largest transplant centers in the world, so that figure is closer to a ceiling than to an
average.

Recall the median age at diagnosis: 70. Put the two facts together and the shape of the problem
appears. The single most effective treatment for AML in second remission is unavailable to most of
the people who reach second remission. The REGAL protocol lists the three reasons a patient may
not be a candidate, and they are worth reading in the order the protocol gives them: intercurrent
medical conditions, the patient’s own preference, or the absence of an available donor.

That population — in remission, out of options, waiting — is the entire commercial premise of galinpepimut-S.

05The drugs a patient receives today

AML treatment has changed more in the last decade than in the thirty years before it. The
following are the products a patient may actually receive today, each with its FDA approval date
and the population its label covers. Read the indications closely: almost every one of them says
newly diagnosed or relapsed or refractory. Those are the two ends of the disease. The
middle — a patient in remission, with nothing visible to treat — is where the label language runs
out.

Intensive chemotherapy remains the backbone for patients fit enough to take it: seven
days of cytarabine with three days of an anthracycline, the regimen universally known as “7+3”,
in use since the 1970s. It is given to induce remission, not to maintain it.

Venclexta (venetoclax), from AbbVie and Genentech, changed frontline practice for older
and unfit patients. It blocks BCL-2, a protein leukemia cells use to avoid programmed cell death.
Approved on an accelerated basis in November 2018 and converted to regular approval in October
2020, its AML label covers newly diagnosed patients aged 75 or older, or younger patients
with conditions that rule out intensive chemotherapy, in combination with azacitidine, decitabine
or low-dose cytarabine. In the VIALE-A trial, venetoclax plus azacitidine produced median overall
survival of 14.7 months against 9.6 months for azacitidine alone, a hazard ratio of 0.66.
In VIALE-C, with low-dose cytarabine, the survival difference was 7.2 months against 4.1
and did not reach statistical significance.

Vyxeos (CPX-351), from Jazz Pharmaceuticals, is a liposomal formulation of daunorubicin
and cytarabine in a fixed ratio, approved in August 2017 for newly diagnosed therapy-related AML
and AML with myelodysplasia-related changes. Median survival was 9.6 months against 5.9.

Mylotarg (gemtuzumab ozogamicin), from Pfizer, is an antibody-drug conjugate aimed at
CD33. It has an unusual history: approved in 2000, withdrawn in 2010 after a confirmatory trial
failed, then re-approved at a lower fractionated dose in September 2017 for both newly diagnosed
and relapsed or refractory CD33-positive disease.

FLT3 inhibitors treat the roughly one third of patients whose leukemia carries a mutation
in the FLT3 gene. Rydapt (midostaurin), Novartis, April 2017, is given with induction and
consolidation in newly diagnosed disease. Xospata (gilteritinib), Astellas, November 2018,
treats relapsed or refractory FLT3-mutated AML, with median survival of 9.3 months against
5.6 for salvage chemotherapy. Vanflyta (quizartinib), Daiichi Sankyo, July 2023, is
approved for newly diagnosed FLT3-ITD-positive AML with induction, consolidation, and as
single-agent maintenance after consolidation — a detail worth holding on to.

IDH inhibitors address a different mutation. Tibsovo (ivosidenib), Servier, July
2018, for IDH1. Idhifa (enasidenib), Bristol Myers Squibb, August 2017, for IDH2 in relapsed
or refractory disease. Rezlidhia (olutasidenib), Rigel, December 2022, for relapsed or
refractory IDH1-mutated AML.

Menin inhibitors are the newest class, aimed at leukemias driven by KMT2A rearrangements
or NPM1 mutations. Revuforj (revumenib), from Syndax, was approved in November 2024 for
relapsed or refractory acute leukemia with a KMT2A translocation, with the label extended to
NPM1-mutated AML in October 2025. Komzifti (ziftomenib), from Kura Oncology, was approved in
November 2025 for relapsed or refractory NPM1-mutated AML. Both were approved on complete-response rates in the low twenties in patients with no satisfactory alternatives: 21.2% for revumenib in KMT2A-translocated disease and 23.1% in its NPM1 cohort, 21.4% for ziftomenib.

Hypomethylating agents — azacitidine, sold as Vidaza, and decitabine — sit underneath all
of this. Here is a fact that surprises people: neither is FDA-approved for AML as a single agent.
Vidaza’s label, unchanged in substance since its 2004 approval, covers myelodysplastic syndromes. Branded Dacogen is listed as discontinued, with generic decitabine
carrying an MDS-only indication. Both are used constantly in AML, and both are used off-label when
used that way.

06Maintenance therapy, and the gap nobody fills

Maintenance is the idea of treating a patient who has nothing visibly wrong with them. The
leukemia is below the detection threshold, the patient feels reasonably well, and the doctor gives
a drug anyway, gently and for a long time, to attack whatever is left before it can rebuild. The
concept is old in pediatric leukemia and was, for decades, a graveyard in adult AML. Trials of
maintenance chemotherapy improved the time to relapse without improving how long patients lived.

That changed once. In the QUAZAR AML-001 trial, 472 patients in first remission after
intensive induction, who were not going on to transplant, were randomized to oral azacitidine or
placebo. Median overall survival was 24.7 months against 14.8 months, a hazard ratio of 0.69
and a p-value of 0.0009. On the strength of that result the FDA approved Onureg, from
Bristol Myers Squibb, on 1 September 2020.

Onureg is the precedent that matters, and the wording of its indication is where the whole argument lives. The label covers “continued treatment of adult patients with acute myeloid
leukemia who achieved first complete remission or complete remission with incomplete blood
count recovery following intensive induction chemotherapy and are not able to complete intensive
curative therapy.” Every clause is a boundary. First remission, not second. After intensive
induction, which excludes the growing share of older patients who never received it. Unable to
complete curative therapy, which is the label’s way of saying not going to transplant.

One other approved product carries maintenance language. Vanflyta (quizartinib) is
approved as single-agent maintenance after consolidation — but only in newly diagnosed
FLT3-ITD-positive AML, which is to say in first remission, in a molecularly defined subgroup, and
its label carries an explicit limitation against use as maintenance after transplant.

So the map of approved maintenance in AML has exactly two entries, and both sit in first
remission. A 2024 review in the British Journal of Haematology puts it plainly: oral
azacitidine is the only non-targeted therapy approved by both the FDA and the EMA for maintenance
in transplant-ineligible AML patients in remission after induction chemotherapy. A 2023 review in
Haematologica is blunter about the field as a whole, noting that improvements in survival
from maintenance agents “have not been consistently demonstrated in clinical trials.”

There is no drug approved anywhere for maintenance in second remission. A patient who
relapses, responds to salvage, reaches CR2 and cannot be transplanted is standing in a gap in the
treatment paradigm. What they receive is decided by their physician’s judgment rather than by any
label: observation, an off-label hypomethylating agent, venetoclax, low-dose cytarabine, or
supportive care. The absence of a standard is not an oversight. It reflects the fact that nobody has ever run a successful randomized trial in that population.

People have tried to fill it. A phase 2 study at MD Anderson, published in Cancer in 2021, tested lenalidomide as maintenance in high-risk patients who had reached a first or second remission and were not candidates for
immediate transplant — the same clinical situation REGAL addresses. Twenty-eight patients were
enrolled; two-year overall survival was 63% and relapse-free survival 50%, with a median remission
duration of 18.7 months. The authors concluded the approach was safe and feasible.

That study is worth holding next to the SELLAS pilot data, because the two have the same shape:
small, single-arm, single-centre, encouraging. Neither can establish that a drug extends life,
which is the reason a randomised phase 3 with an overall survival endpoint is the only thing that
settles the question — and the reason no approval has followed either of them.

That is the gap SELLAS Life Sciences has been trying to fill since it licensed the technology in 2014.

07Where SELLAS fits: galinpepimut-S and the WT1 target

SELLAS Life Sciences Group describes itself in its filings as a late-stage clinical biopharmaceutical company. It is based in New York, has two clinical assets, no approved products and no revenue. Its lead candidate, galinpepimut-S — GPS — is a
therapeutic cancer vaccine, and understanding what that phrase does and does not mean is
essential to reading the story correctly.

A preventive vaccine teaches an immune system to recognize a pathogen it has not met. A
therapeutic cancer vaccine does something harder. It takes a patient whose immune system has
already failed to control a cancer, and tries to point that same immune system at a target it has
been ignoring. It is not chemotherapy: it does not kill cells directly. It is an instruction, and
whether the instruction is followed depends on the patient’s immune system.

The target is WT1, the Wilms tumor 1 protein. WT1 is heavily overexpressed in leukemic
blasts and largely absent from healthy blood-forming cells, which is what makes it attractive: an
immune response aimed at WT1 should hit leukemia and spare the marrow. In 2009 a National Cancer
Institute pilot project convened a panel to rank 75 cancer antigens by their suitability as
immunotherapy targets, weighting therapeutic function, immunogenicity, oncogenicity, specificity
and expression level. WT1 placed first, with an aggregate score of 0.81. SELLAS cites that
ranking constantly, and it is real. It is also worth reading what the authors themselves wrote:
none of the 75 antigens possessed all the characteristics of an ideal target. A first-place finish
in that field is a statement about relative promise, not about proven efficacy.

GPS itself is a mixture of four peptides derived from WT1. Two of the four are
deliberately altered at a single amino acid so the immune system reads them as foreign rather than
as self, which is intended to break the tolerance a body normally has toward its own proteins. The
peptides were designed by researchers at Memorial Sloan Kettering, and SELLAS licensed the
technology from MSK in September 2014, with milestone obligations of up to $17.4 million per
licensed product and a royalty in the mid-single digits.

Administration is unglamorous and, for this population, is the point. GPS is injected under the
skin in an emulsion with the adjuvant Montanide, which forms a depot that releases the peptides
slowly. Each dose is preceded by GM-CSF, a second adjuvant that recruits immune cells to the
injection site. In REGAL the schedule runs every two weeks for six doses, then every four weeks,
then every six, then every two months in year two and every three months thereafter. Compare that
with a transplant, and the appeal to a 72-year-old in remission is obvious.

GPS holds FDA orphan drug designation in AML, granted December 2015, plus orphan status in
mesothelioma and multiple myeloma, Fast Track designation in all three, EMA orphan opinions, and a
Rare Pediatric Disease designation for pediatric AML granted in October 2024. It does not
hold Breakthrough Therapy designation, and it does not hold RMAT. Those designations reflect
regulatory encouragement and unmet need. None of them is evidence that the drug works.

08The REGAL trial, and the count that triggers it

The REGAL trial is the reason SELLAS is discussed at all, and its design is unusually
easy to describe. Adults with AML in second complete remission or CRp2, after second-line salvage
therapy, who are not candidates for allogeneic transplant, are randomized one-to-one to GPS as
maintenance monotherapy or to best available therapy chosen by the investigator. The primary
endpoint is overall survival — not response rate, not progression-free survival, but how long
patients live. Randomization is stratified by CR2 versus CRp2, cytogenetic risk at diagnosis,
minimal residual disease, and whether the first remission lasted less than or more than a year.

The control arm deserves attention, because it tells you what SELLAS is really competing
against. Investigators may choose observation, with palliative hydroxyurea permitted; a
hypomethylating agent, azacitidine or decitabine; venetoclax; or low-dose cytarabine. Patients
whose remission can be held with a targeted FLT3 or IDH inhibitor are excluded from the trial
entirely. GPS is therefore not racing another experimental drug. It is racing current practice in
a setting where current practice has no approved option, and where one of the permitted comparators
is doing nothing at all.

REGAL is event-driven, which is the single most important structural fact about the
stock. The analysis does not happen on a date. It happens when a predetermined number of deaths has
accumulated among enrolled patients. ClinicalTrials.gov records 127 patients as the actual enrollment, and SELLAS cited the same figure in April 2024. The protocol had planned for 125 to 140 patients across roughly 95 sites in North America, Europe and Asia. In November 2022 the company disclosed that the required event count for the
final analysis had been reduced from 105 to 80 deaths, with the interim analysis moved from
80 events to 60, and the enrollment target raised. The same release gave the statistical
assumption: a hazard ratio of 0.636, corresponding to median overall survival of 12.6 months for GPS against 8 months for best available therapy.

There is a second set of numbers, and the two do not match. In an abstract presented by the REGAL
investigators in the Journal of Clinical Oncology in 2023, the study is described as enrolling
125 to 140 patients to give at least 90% power under an assumed hazard ratio of 0.52, based on
median overall survival of 8.0 months for best available therapy against 15.4 months for GPS.
The company’s own November 2022 release gives 0.636 and 12.6 against 8. Both figures come from
sources tied to the trial, and the difference is not trivial: one implies a survival gap of about
four and a half months, the other of more than seven. Neither the alpha level nor a reconciliation
of the two has been published.

The event clock has run slowly, and slowness has been read by investors as a good sign, since
events are deaths. The disclosed sequence: the 60-event interim threshold was reached in
December 2024. In January 2025 the independent data monitoring committee reported that
GPS had exceeded the predetermined futility criteria and recommended continuation without
modification, noting that fewer than half of enrolled patients had been confirmed deceased at a
median follow-up of 13.5 months, and that 80% of a random sample of GPS patients showed a specific
T-cell immune response. The committee reviewed again in August 2025 and recommended
continuing. SELLAS disclosed 72 events as of 26 December 2025 and 78 events as of 11 May
2026. In the second-quarter release on 11 August 2026 the company said only that it was
approaching the 80th event, without giving a number.

As of this writing the 80th event has not been announced. When it occurs, SELLAS has said
it will announce it, after which come database lock, blinded data review, statistical analysis,
unblinding and disclosure of topline results. The company has said it is preparing for a potential
Biologics License Application should the analysis succeed. No submission date has ever been
given.

Two caveats belong next to the encouraging interim language. First, passing a futility analysis
is a low bar by design: it means the trial was not obviously failing at 60 events, not that it was
succeeding. Second, an immune response is not a clinical outcome. The history of cancer
immunotherapy is full of agents that reliably provoked T-cell responses and did not extend
life.

09The evidence behind GPS, and how much of it there is

Every investor in SELLAS eventually meets one pair of numbers: 21.0 months against 5.4
months. That is the median overall survival reported for GPS versus best standard care in AML
patients in second remission, at a median follow-up of 30.8 months. An earlier readout of the same
study, at 19.3 months of follow-up, gave 16.3 months against 5.4. The comparison is the emotional
core of the investment case, and it is genuinely striking.

It is also, on the company’s own description in its annual report, built on ten patients.
The study was a phase 2 conducted at the Moffitt Cancer Center, not a SELLAS registrational trial.
Ten AML patients who had relapsed, received second-line chemotherapy and reached CR2 were treated
with GPS. The comparison group was fifteen patients managed with watchful waiting by the same
clinical team over the same period. SELLAS states explicitly that the groups were “not matched by
randomization” and describes them as broadly comparable.

Ten patients against fifteen, not randomized, at a single center. In a disease where outcome
depends heavily on cytogenetic risk, on how long the first remission lasted, on age and on
performance status, a non-randomized comparison of twenty-five people cannot separate the effect
of a drug from the effect of which patients happened to receive it. A physician deciding to
vaccinate a patient rather than watch them is making a judgment about that patient, and healthier
patients live longer whatever you give them. This is the mechanism that has produced more failed
phase 3 trials in oncology than any other.

There is a second issue worth flagging for anyone checking the sources. The result does not
appear to have been published in a peer-reviewed journal; the figures are traceable to SELLAS
filings and press releases. The GPS phase 2 that was published, in Blood Advances in
2018, is a different study in a different population: 22 patients in first remission, median
age 64, median disease-free survival of 16.9 months from CR1, with an immune response in 9 of 14
patients tested. That paper is real and reasonably encouraging. It is not the CR2 result, and the
two should not be blended.

None of this means GPS does not work. It means the evidence that it works, as of today, is a
small non-randomized series, and that REGAL exists precisely because that kind of evidence cannot
settle the question. A phase 3 with roughly 127 patients and a hard survival endpoint is the appropriate test. The reason to be careful is arithmetic: the trial is powered for a hazard ratio
of 0.636 and a survival gap of roughly four and a half months. The pilot data suggest a gap of
fifteen months. If the pilot were an accurate estimate of the drug’s effect, the trial would have
crossed its futility and interim thresholds far more emphatically than it has.

10The precedent the company does not cite

Company materials describe WT1 as the top-ranked immunotherapy target and GPS as a first-in-class
approach. What they do not mention is that another WT1 peptide vaccine has already been tested
against placebo in AML, and failed.

OCV-501 is a WT1 peptide presented to helper T cells. It was studied in a randomized
phase 2 in elderly AML patients in first remission. The five-year follow-up, published in
Cancer Immunology, Immunotherapy in 2023, reported five-year disease-free survival of 36.0% in the OCV-501 group against 33.7% on placebo, p=0.74. The trial randomized 134 patients; the five-year figures rest on the 105 still evaluable at that point, 52 and 53 per group. No difference. The
authors noted that patients who mounted a stronger antibody response did better — a finding that
cuts both ways, since it may mean the vaccine helps those able to respond, or simply that patients
with functioning immune systems live longer regardless.

OCV-501 is not GPS. It is a single peptide against a mixture of four, presented to a different
arm of the immune system, tested in first remission rather than second, in a Japanese elderly
population. A fair reading is that the two are cousins rather than twins. But it is the closest
completed randomized test of the same biological idea in the same disease, and it produced a
p-value of 0.74.

The wider history is sobering in the same direction. An analysis published in the Journal for ImmunoTherapy of Cancer in 2015 observed that of 175 oncology drugs approved by the FDA between 1996 and 2014, exactly one therapeutic cancer vaccine had been approved: sipuleucel-T. Hundreds of phase 2 and phase 3
vaccine trials, one product. The field has improved since — a 2025 meta-analysis of eleven
randomized trials in advanced non-small-cell lung cancer found cancer vaccines associated with an
overall survival hazard ratio of 0.85, statistically significant, while progression-free survival
showed no significant benefit. That pattern, survival moving without the disease measurably
slowing, is characteristic of immunotherapy and is part of why these agents are so hard to read
early. Interestingly, that same meta-analysis identified immune response as a predictive
biomarker of benefit, which is the argument SELLAS makes about its 80% T-cell response figure.

A more encouraging data point, published in Blood Advances in 2026, comes from a Japanese
phase 2 of WT1 peptide vaccines as post-transplant maintenance in 17 pediatric patients with
refractory acute leukemia: three-year overall survival of 70.6% against a historical benchmark of
30%, with survival strongly associated with the size of the WT1-specific T-cell response. Seventeen
children, a historical control, a different setting. Suggestive, not decisive.

Put together, the honest summary is this. WT1 is a serious target with a serious rationale. The
approach has produced one clear randomized failure, several small encouraging series, and no
definitive success. REGAL will be the most rigorous test the idea has ever received.

11The second asset: SLS009 and CDK9

SELLAS’ second asset gets less attention and carries a nearer-term readout. SLS009, also
called tambiciclib, is a small-molecule inhibitor of CDK9, licensed from GenFleet Therapeutics in
March 2022 for all territories outside Greater China.

CDK9 is not a cell-cycle kinase despite the name. It regulates transcription: it phosphorylates
RNA polymerase II and allows it to keep reading a gene rather than stalling. Blocking it shuts down
production of short-lived proteins first, and one of the shortest-lived is MCL-1, an anti-apoptotic
protein many leukemia cells depend on to stay alive. Inhibit CDK9, MCL-1 decays within hours, and
the cell’s survival machinery fails. The mechanism is well established; the difficulty with earlier
CDK9 inhibitors was toxicity, and SELLAS positions SLS009 as more selective and better tolerated.

The reported data are early and have been presented in fragments across successive releases,
with the denominators changing. The most complete set, released in July 2025 as final phase 2 data,
covered 54 evaluable patients, of whom 23 carried ASXL1 mutations. Overall response rate was 33%
across all cohorts and 40% at the 30 mg twice-weekly dose; in ASXL1-mutated patients the reported
response rate at that dose was 50%, or 9 of 18. Median overall survival was 8.8 months in
patients with a median of one prior line. At ASH in December 2025, in 35 evaluable patients with
myelodysplasia-related AML, overall response was 46% with complete responses in 29%, and response
rates of 48% in ASXL1-mutated and 57% in TP53-mutated patients — though the release did not give
the denominators behind those percentages.

ASXL1 matters because it is a marker of poor prognosis with no targeted therapy. A drug with
real activity there would have a clear place. The caution is that these are single-arm response
rates in small, shifting subgroups, of a kind that frequently shrink when tested against a
control.

The active question is the frontline study. A randomized phase 2 in newly diagnosed AML patients
identified as unlikely to benefit from azacitidine plus venetoclax is enrolling toward 80 patients,
with 28 enrolled as of 11 August 2026 and topline data expected in the fourth quarter of
2026. That gives SELLAS two potential catalysts within months of each other, which is unusual
for a company this size and cuts both ways.

12The balance sheet, the warrants and the valuation

SELLAS finished the second quarter of 2026 with $138.3 million in cash, against $71.8
million at the end of 2025. Operating expenses for the first half were $19.9 million — $11.4 million of R&D and $8.5 million of G&A, the latter including a one-off $1.0 million arbitration charge — for a net loss of $18.0 million. The company
carries no financial debt: total liabilities of $7.7 million consist of payables, accruals
and lease obligations. On the reported burn, the cash position is measured in years rather than
quarters, and the company states it has resources for at least the twelve months following the
filing.

How that cash arrived is the part most readers miss. It was not an offering. In the first half
of 2026 SELLAS received approximately $82.9 million from the exercise of 48.8 million
warrants at a weighted average price of about $1.70. Warrant exercises are dilution that has
already happened: shares outstanding went from 153,103,459 at 31 December 2025 to 201,918,874 at
30 June 2026, an increase of roughly 32% in six months. A further 9.7 million warrants remain
outstanding, alongside 2.6 million options and 3.3 million restricted stock units.

Those warrants were themselves the product of a chain of financings priced when the stock was
low. A registered direct offering in January 2025 raised about $23.1 million net at a combined $1.27 per share with warrants exercisable at $1.20. A September 2025 inducement brought about $22.0
million by having those warrants exercised at $1.20 in exchange for new warrants at $1.88. An
October 2025 inducement raised about $29.1 million the same way, issuing new warrants at $2.00.
Each round converted paper into cash and issued fresh paper behind it.

In March 2026 the company signed an at-the-market sales agreement with TD Cowen for up to
$150.0 million. As of the second-quarter filing, no shares had been sold under it.
The facility is authorised for up to $150 million and was undrawn as of the last filing.

Set against that balance sheet, the market’s valuation is the number that frames everything. At around $15 per share during the session of 27 August 2026, on Finviz Elite data, 201.9 million shares imply a market capitalization of roughly $3.1 billion. That is a company with no revenue, no approved
product, $138 million of cash and two clinical assets, valued above Syndax, Kura Oncology, Geron
and Immatics, all of which have either approved medicines or broader pipelines. The price therefore embeds expectations above those of companies that already sell an approved medicine.

One further contract shapes the picture. In December 2020 SELLAS licensed GPS in mainland China, Hong Kong, Macao and Taiwan to 3D Medicines for a $7.5 million upfront payment, milestones of up to $194.5 million and tiered royalties starting in the high single digits. In December 2023 SELLAS started arbitration against its own partner over the triggering and payment of those milestones and over what it described as 3D Medicines’ failure to use commercially reasonable efforts to develop GPS in China. On 24 July 2026 the sole arbitrator dismissed SELLAS’ claims and allocated roughly $0.7 million of 3D Medicines’ legal fees plus $0.3 million of arbitration costs to SELLAS, recognised as a $1.0 million charge within general and administrative expenses in the second quarter. The $194.5 million of milestones remains a contractual maximum, not money received.

13What to watch, and what each signal means

Everything above reduces to a small number of observable facts, and the useful thing
for a reader is knowing which ones to watch rather than being told what to conclude.

The 80th event. SELLAS has committed to announcing it. The last disclosed count was 78 as
of 11 May 2026. When the announcement comes it starts a defined sequence — database lock, blinded
review, statistical analysis, unblinding, topline — that takes weeks, not days. The announcement of
the event is not the result.

The comparison that will decide the trial. The threshold is a hazard ratio of 0.636,
median survival of 12.6 months against 8. The pilot suggested 21.0 against 5.4 on ten treated
patients. Those two expectations are far apart, and the trial will land somewhere.

Whether the control arm behaves as assumed. The protocol assumed 8 months for best
available therapy. If patients randomized to observation, hypomethylating agents or venetoclax do
better than that — and venetoclax-based regimens have improved considerably since the trial was
designed — the gap narrows regardless of how well GPS performs. This is the quiet risk in every
trial whose control arm is “whatever the doctor thinks best.”

The SLS009 frontline readout in the fourth quarter of 2026, on 80 planned patients with 28
enrolled in August. Unlike REGAL, this one has a randomized comparator built in.

The $150 million ATM. Undrawn as of the last filing. The share count is restated in every quarterly filing, which is where any use of the facility becomes visible; the current figure is carried on the $SLS stock hub.

What approval would actually mean. If REGAL succeeds and a BLA follows, GPS would enter a
setting with no approved competitor, in a population defined by three characteristics — second
remission, no transplant, no targetable mutation — that is real but narrower than “AML.” Sizing that opportunity starts from the 22,720 annual US cases, subtract those who never reach
a first remission, apply the roughly 59% who relapse, then the fraction who achieve a second
remission, then remove those who go to transplant and those whose disease can be held with an FLT3
or IDH inhibitor. The number that survives that sequence is the addressable population, and it is
several steps removed from the headline incidence figure.

The company itself has framed the situation in a phrase that is easy to skim past. In its
second-quarter release, SELLAS said it will announce the 80th event “when it occurs.” Everything
else — the valuation, the volume, the retail attention — is a market pricing a coin that has not
yet landed, in a disease where the last four decades moved five-year survival from 5% to 33%, one
difficult increment at a time.