Drug Development · global
Targeting Drug Resistance in Leukemia: AI-Designed CHM-029 Enters First-in-Human Trial
CHM-029 aims to mimic the binding mode of a natural protein, making it harder for leukemia cells to evade treatment through mutations. The first patients have begun receiving this menin inhibitor, but whether it can deliver durable efficacy remains a question for human trials.
For patients with relapsed or refractory acute myeloid leukemia (AML), whether targeted drugs can remain effective is a major treatment challenge. CHARM Therapeutics announced on October 7 that the first patient had been dosed with CHM-029, designed to address drug resistance, in a Phase I/II trial, and that enrollment in the first cohort had also been completed. This new drug, designed with AI assistance, is beginning the transition from laboratory models to testing in humans.
CHM-029 targets the menin protein. In AML with certain genetic abnormalities, the binding of menin to the KMT2A protein maintains gene activity that promotes leukemia cell growth. Inhibiting this interaction could allow malignant cells to resume differentiation and proceed toward death; however, resistance mutations in menin may weaken drug binding, allowing the disease to progress again.
CHARM used its DragonFold platform to design CHM-029, aiming to make the drug more closely mimic the natural binding mode between KMT2A and menin. The idea behind this design is that if cancer cells must retain this natural interaction, a drug that mimics its points of contact might be less readily excluded by mutations. This strategy addresses a specific resistance mechanism, and its clinical effectiveness has not yet been established.
The company said that in preclinical studies, CHM-029 retained activity against all relevant resistance mutations described publicly to date and showed increasingly pronounced tumor shrinkage as doses increased in models both with and without resistance mutations. However, the announcement did not provide complete potency data or comparative results; performance in models also cannot be directly extrapolated to patient remission rates or duration of efficacy.
The trial, numbered NCT07751991, enrolls patients with relapsed or refractory AML harboring NPM1 mutations, KMT2A rearrangements, or NUP98 rearrangements. According to trial information compiled by Peptide and Protein News, the study plans to enroll 40 participants and uses oral dose escalation with additional enrollment cohorts. It primarily evaluates dose-limiting toxicities, adverse events, and dose adjustments to identify the maximum tolerated dose or optimal biological dose. These questions will first establish whether the drug can be used safely, then provide a basis for interpreting efficacy.
CHARM also said that fewer than 24 months elapsed between the first synthesis of CHM-029 and dosing of the first patient, and described it as the first molecule designed using a protein–ligand co-folding model to enter clinical studies. This “first” claim represents the company’s assertion of a technological milestone; the company website and GlobeNewswire carry the same announcement, which cannot be considered two independent validations and is insufficient to establish that AI methods outperform other approaches to drug design.
Initial clinical data are expected in 2027. Safety, drug exposure in humans, and treatment responses across genetic subgroups will then gradually become clearer. Whether CHM-029 can maintain activity in patients with resistance mutations and prolong disease control remains the central question this trial needs to answer.