In a groundbreaking study published in Nature Biotechnology, an international team of researchers has successfully harnessed the power of quantum computing and artificial intelligence (AI) to design potential inhibitors for KRAS, a protein long considered "undruggable" and implicated in numerous cancers.

Understanding the Challenge: Targeting KRAS

KRAS mutations are present in approximately 25% of all human cancers, including pancreatic, lung, and colorectal cancers . Despite its prevalence, KRAS has been notoriously difficult to target due to its structural complexity and high affinity for its natural ligands . Traditional drug discovery methods have struggled to develop effective inhibitors, making KRAS a prime candidate for innovative approaches.

The Quantum-Classical Hybrid Approach

The research team developed a hybrid quantum-classical generative model to design small molecules capable of inhibiting KRAS. This approach combined quantum algorithms running on a 16-qubit IBM quantum computer with classical machine learning techniques . The model was trained on a dataset of 1.1 million molecules, including 650 known KRAS inhibitors and over 250,000 molecules from the VirtualFlow platform.

The quantum component generated a "prior distribution" of potential molecules, which the classical component refined into viable candidates. A reward function guided the model to prioritize molecules with properties indicative of effective KRAS binding.

Promising Results: Two Novel Inhibitors

From the generated candidates, 15 molecules were synthesized and subjected to laboratory testing. Two molecules, ISM061-018-2 and ISM061-022, demonstrated significant potential:

• ISM061-018-2: Exhibited broad-spectrum inhibition across several KRAS mutants, including the prevalent G12D variant, with a binding affinity of 1.4 μM.

• ISM061-022: Showed selective activity against specific KRAS mutants such as G12R and Q61H.

Both molecules featured unique scaffolds distinct from existing KRAS inhibitors and demonstrated minimal off-target effects in preliminary tests.

Implications and Future Directions

This study marks the first instance of a quantum-generative model yielding experimentally confirmed biological hits, showcasing the practical potential of quantum-assisted drug discovery . While the researchers do not claim a definitive "quantum advantage," the hybrid approach outperformed purely classical models in generating viable drug candidates.

The team plans to enhance their model by increasing the number of qubits in the quantum component and exploring transformer-based generative algorithms to improve molecular diversity and quality . Such advancements could further reduce the time and cost associated with drug discovery, potentially compressing years of work into months.

Conclusion

The integration of quantum computing and AI in drug discovery represents a significant leap forward in the fight against cancer. By successfully targeting KRAS, a protein once deemed "undruggable," this study opens new avenues for developing effective cancer therapies and underscores the transformative potential of emerging technologies in medicine.