Hormona

An AML patient starts on revumenib, the first FDA-approved menin inhibitor. For six months, her leukemia has responded, simultaneously predicting how. Then it does not. A resistance mutation in the menin protein has changed the binding pocket just enough to make the drug ineffective. This is the clinical problem that CHARM Therapeutics set out to solve on the day it was founded, and it is the precision of that problem statement, matched by a platform architecture built specifically to address it, that earned CHARM Therapeutics a 2026 Global Recognition Award. The company generated a next-generation menin inhibitor designed to maintain potency against the exact mutations that defeat first-generation therapy, in two years from founding, using DragonFold, its proprietary protein-ligand co-folding platform, now protected by US Patent No. 12,211,588 and validated as state-of-the-art in independent binding pose prediction benchmarks.

Technical Innovation and Architecture

DragonFold is the first published deep learning system that simultaneously predicts how both a protein and a drug candidate fold together in three dimensions, using only the protein’s amino acid sequence and the ligand’s molecular structure as inputs. Standard drug discovery platforms predict protein structure and drug structure separately, then attempt to fit them together as rigid bodies, missing the conformational dynamics that occur when a drug actually binds. DragonFold’s 3D-invariant, bond-aware transformer models both molecules simultaneously, integrating dedicated protein and ligand embedders through an architecture that captures induced fit at the moment of interaction, producing binding pose predictions that independent benchmarks in July 2025 confirmed as state-of-the-art performance.

CHARM has extended DragonFold’s structural prediction capability into lead optimization by integrating it with Free Energy Perturbation, the gold-standard method for calculating binding affinity, which has traditionally been too computationally intensive for large-scale use. By automatically generating accurate co-folded protein-ligand structures, DragonFold eliminates the manual setup bottleneck in FEP workflows, making high-accuracy binding affinity prediction scalable across hundreds of candidates in parallel rather than in sequential single-candidate workflows. US Patent No. 12,211,588 now protects the core computational framework, covering the deep learning method that predicts protein-ligand co-folding from primary sequence and molecular structure alone.

Market Strategy and Leadership

CEO Laksh Aithani co-founded CHARM at 25 with direct AI drug discovery machine learning experience from Exscientia and prior Y Combinator founder credentials. Professor David Baker, a 2024 Nobel Prize laureate in Chemistry for computational protein design, co-founded CHARM, the scientific architecture underlying DragonFold’s co-folding approach. Chairman Gary Glick previously founded IFM Therapeutics, Scorpion Therapeutics, and Odyssey Therapeutics, providing three-time drug-discovery company-building experience, guiding CHARM’s clinical strategy. The September 2025 Series B brought Briggs Morrison, former CEO of Syndax, who developed revumenib (the first FDA-approved menin inhibitor), onto the board as clinical development architect for the next generation of the same drug class.

CHARM’s clinical strategy builds on a precedent that repeats reliably in oncology: second-generation inhibitors that overcome first-generation resistance mechanisms consistently capture larger market positions than the drugs they replace. In BCR-ABL, EGFR, and CDK4/6 inhibition, second-generation programs built around known resistance mutations achieved blockbuster commercial outcomes. CHARM’s lead candidate is designed against the M327I/V and A204V menin resistance mutations that cause revumenib failure, following the same template as DragonFold’s dynamic co-folding, enabling resistance-aware molecular design that traditional methods cannot achieve.

Industry Impact and Future Vision

CHARM’s two-year candidate generation timeline from founding provides the most direct industry benchmark for what co-folding-native drug discovery delivers compared to conventional methods. In the menin inhibitor space, CHARM is the only company with a next-generation candidate specifically engineered against published resistance mutations, meaning its clinical differentiation thesis is built into the drug’s molecular structure, rather than asserted as a positioning claim. The Series B capital positions CHARM to file the IND for its lead AML candidate and enter first-in-human trials, the milestone market analysts have identified as the dividing line between AI drug discovery companies commanding premium valuations and preclinical-only platforms.

Beyond AML, DragonFold’s co-folding architecture is applicable to any drug discovery program where resistance mutation dynamics, induced fit modeling, or cryptic binding site design represents a barrier to conventional structure-based discovery, a description covering the majority of the most commercially important oncology targets in the industry’s pipeline. The 2026 Global Recognition Award captures a company that assembled the most credible founding team in AI drug discovery, built the most architecturally differentiated platform in the field, generated a clinical candidate against one of oncology’s most important resistance problems in two years, and is now advancing that candidate toward the clinical proof that will confirm whether DragonFold changes what is possible in cancer treatment.