Two Ex-GSK AI Leads Go Stealth with a New Oncology Drug Discovery Startup

Clockwork Bio has been launched as a new AI-native biotech focused on redesigning early-stage oncology drug discovery. The company is developing a platform centered on modeling and controlling cell state transitions, aiming to address limitations in current pipelines that struggle with complex, heterogeneous diseases such as cancer.

The company was founded after Shane Lewin and Nick Peterson both left GSK after working there for 6 years. Clockwork Bio is targeting what a growing number of groups now describe as cancer’s core data problem. Traditional drug discovery pipelines have worked best in diseases driven by clear genetic mutations, often surfaced through genome-wide association studies or large-scale screening. 

In oncology, however, a tumor is not a static lesion but an evolving population of cells that accumulates mutations, adapts to treatment, signals to neighboring cells, and interacts continuously with the immune system. Capturing that behavior requires integrating multiple data layers at once, including pathology, sequencing, spatial biology, treatment history, and clinical outcomes. 

That is also why standard pipelines often struggle in cancer, where disease states are frequently shaped less by a single molecular driver than by epigenetic regulation, signaling pathways, and cellular context.

This complexity has translated into low success rates and high costs. Early-stage target discovery programs in oncology are reported to take 4–7 years and cost around $30 million, while still progressing to clinical trials with approval probabilities near 3.6%.

Clockwork Bio’s approach centers on generating and learning from controlled transitions between healthy and diseased cell states in vitro. The platform is described as combining several recent technical developments:

• active learning systems that iteratively design and run experiments
• AI-based phenotyping to quantify differences between cellular states
• emerging therapeutic modalities capable of modulating cell behavior with higher precision

The goal is to build systems that can both induce disease-relevant states in cells and reverse them, while continuously learning the underlying biological rules. This shifts the focus from static target identification toward dynamic modeling of disease processes.

Broadly speaking, cancer AI has recently started moving beyond scan reading and pathology support toward systems that connect pathology, sequencing, spatial biology, treatment history, and clinical outcomes in one framework. 

For now, Clockwork Bio remains in stealth with formal product launch and additional updates expected later in 2026.

Topic: AI in Bio