CHARM Therapeutics Raises $80 Million to Advance Next-Generation Menin Inhibitors for AML
CHARM Therapeutics has secured $80 million in Series B financing to push its AI-designed menin inhibitor program for acute myeloid leukemia (AML) into clinical development, with trials expected to begin in early 2026. The round was co-led by New Enterprise Associates (NEA) and SR One, with participation from OrbiMed, F-Prime, Khosla Ventures and NVIDIA.
Founded by Laksh Aithani and David Baker, CHARM Therapeutics operates between Cambridge and London. The company has raised more than $150 million to date and is focused on advancing AI-enabled small molecule discovery, with its menin inhibitor program as the lead asset.
AML is an aggressive blood and bone marrow cancer driven by the uncontrolled growth of abnormal myeloid cells that fail to mature. A key driver in some AML subtypes is the interaction between menin and KMT2A, a gene that encodes lysine methyltransferase 2A (also known as MLL, or mixed-lineage leukemia protein). Under normal conditions, KMT2A regulates transcription and cell differentiation. But when bound to menin, it triggers upregulation of “stemness” genes that sustain leukemic growth.
Menin inhibitors are designed to block this interaction, restoring normal gene regulation and enabling malignant cells to differentiate or undergo apoptosis. Unlike conventional chemotherapies that aim to kill leukemic cells directly, menin inhibition acts by reprogramming them. The therapeutic class has been clinically validated, but first-generation molecules face challenges including resistance mutations that weaken drug binding, as well as safety liabilities such as QTc prolongation and drug–drug interactions.
CHARM reports that its inhibitors are structurally optimized to bind within the KMT2A–menin interface. This means that mutations preventing drug binding would also disrupt leukemogenic signaling, reducing the risk of treatment escape. Using its DragonFold protein–ligand co-folding platform, the company has developed a candidate that retains potency against all publicly described resistance mutations and demonstrated tumor regression in preclinical models. The drug candidate is predicted to work at low doses, with reduced risks of drug–drug interactions and QTc prolongation.
DragonFold is CHARM’s proprietary AI-driven structure prediction system designed for small-molecule discovery. Unlike earlier protein–ligand cofolding models, which reportedly often struggle to generalize beyond training data, DragonFold has shown benchmark performance comparable to AlphaFold-3 in predicting ligand binding poses. It incorporates features such as program-specific fine-tuning, ligand-based templating at inference, and a web-based interface for direct scientist use.
Image credit: How DragonFold Stacks Up
The platform has been integrated with Free Energy Perturbation (FEP) methods, enabling more accurate and scalable predictions of protein–ligand binding affinities. This combined DragonFold–FEP approach bypasses labor-intensive setup steps, traditionally reliant on crystallography, and has demonstrated accuracy on par with X-ray-based FEP. CHARM is already applying the method in internal discovery programs to accelerate lead optimization and potency prediction.
The company has also expanded its leadership as it prepares for clinical transition. Former Syndax CEO Briggs Morrison, who previously oversaw development of the first FDA-approved menin inhibitor, and oncology specialist Kim Blackwell join the board as non-executive directors. New investor representatives Matthew McAviney (NEA) and Mahesh Kudari (SR One) will also take board seats.
Topics: Startups & Deals
