What is it: REC-4881 is an oral targeted therapy being studied for advanced solid tumors with AXIN1 or APC mutations. This Phase 2 trial was to evaluate its safety, efficacy, and pharmacokinetics in patients with unresectable locally advanced or metastatic disease.

What is it: JUV-161 is a muscle-targeting drug designed to promote muscle growth and prevent atrophy during periods of disuse. In this 96-day, placebo-controlled trial, healthy adults undergo temporary leg immobilization and receive weekly injections to test whether the drug reduces muscle loss.

For more info on all AI Drugs in Clinical Trials. NOTE: It’s still under development!

AlphaFold changed how scientists look at proteins. It gave us a reliable way to predict 3D structures from sequence and became a foundational tool across biology. But structure alone does not make a drug.

Designing real medicines requires predicting how molecules bind, how strongly they bind, and where entirely new binding sites might exist.

This week, Isomorphic Labs introduced its Drug Design Engine, IsoDDE, a unified AI system that aims to move from structure prediction to full in silico drug design.

What’s new: IsoDDE focuses on generalisation. The company reports that on the “Runs N’ Poses” benchmark, designed to test performance on novel protein-ligand systems, IsoDDE more than doubles AlphaFold 3’s accuracy on the most difficult category, where targets are highly dissimilar from training data.

For complex biologics, the system also improves antibody-antigen modelling.

On a challenging test of 334 unfamiliar antibody structures, IsoDDE produced far more high-accuracy predictions than existing AI models, including AlphaFold 3. It was especially strong at modeling the most variable and difficult part of antibodies, a region that often determines how well a therapy works.

It also predicts how tightly a drug candidate will bind to its target. Across several public benchmarks, IsoDDE outperformed other AI systems and even matched or exceeded traditional physics-based simulations, while running much faster and without requiring detailed experimental structures as input.

Why this matters for drug discovery: Perhaps most interesting is pocket discovery. IsoDDE can identify binding pockets from sequence alone.

If these results hold in real programs, this would be a major breakthrough for drug research.

Accurate structure, affinity prediction, and blind pocket discovery in one system could compress early-stage timelines and expand the druggable proteome.

Sources: [Press Release]

Select the right answer! (See explanation below and source)

Japanese giant Takeda Pharmaceutical and U.S. startup Iambic Therapeutics announced a multi‑year partnership to co‑develop small‑molecule drugs using generative AI.

Reuters reported that the deal could deliver more than $1.7 billion in development and commercial milestones and includes royalties.

Takeda gains access to Iambic’s NeuralPLexer generative models to design molecules across oncology and inflammatory diseases, while Iambic receives upfront research funding and can earn success‑based payments.

Iambic’s CEO said the technology can compress discovery timelines from six years to less than two, and Takeda’s CSO described the partnership as a way to maintain high molecular quality while accelerating research.

The announcement is widely viewed as the largest generative‑AI drug discovery deal yet and signals that big pharma is now willing to commit billions to AI platforms.

Sources: [Press Release]

Prions are misfolded proteins that trigger a chain reaction of misfolding in the brain, leading to neurodegeneration.