insider brief
Researchers developed a hybrid quantum-classical model to design two small molecules that target the KRAS protein, a key player in cancer biology. This research combined quantum circuits and classical machine learning to screen millions of compounds and produced two promising drug candidates that were validated through lab testing. Although early results show promise, this approach is still limited by current quantum technologies and requires further optimization before clinical application.
An international team of researchers has successfully designed two promising small molecules that target proteins involved in cancer and presented a hybrid quantum-classical model for targeting proteins involved in cancer.
The study, published in Nature, combines quantum and classical computational tools to design potential drugs for KRAS, a protein that has been considered difficult to target due to its structural complexity. We have demonstrated how it can be strengthened. Scientists have long been interested in the target of KRA, as it is a protein implicated in a variety of cancers, including lung, colorectal, and pancreatic cancers. This makes it an important target for therapeutic development.
However, the structure’s complexity and role in cell signaling have historically made it difficult to effectively inhibit it.
In this study, the researchers synthesized 15 candidate molecules, two of which showed promise as KRAS inhibitors in preliminary tests, paving the way for further development.
Drug discovery is a time-consuming and expensive process, often requiring billions of dollars of investment over 10 years or more. Generative models, which use machine learning to predict new molecular structures with desired properties, are emerging as a way to navigate the vast chemical space of potential drug candidates. The integration of quantum computing into this process provides new approaches to address the limitations of classical methods, especially when designing molecules for challenging targets like KRAS.
“This work demonstrates the potential of quantum computing to generate experimentally validated hits that compare to classical models,” the team wrote.
According to researchers, quantum computing leverages principles such as superposition and entanglement to explore high-dimensional data spaces. The team’s hybrid model outperformed purely classical methods of generating viable drug candidates. This suggests that there may be at least a synergistic benefit.
They wrote: “Purely classical algorithms have made significant advances in drug discovery, but hybrid classical square approaches are becoming more powerful than quantum circuit-born machines (QCBMs), which exploit quantum effects such as superpositions and entanglement. The introduction of QCBMS demonstrates this progress, providing generative models that can outperform classical models in certain aspects. QCBMS is a quantum generative model that leverages quantum circuits to learn complex probability distributions, allowing you to generate new samples similar to your training day.”
methodology
Researchers used a three-step process to design and test KRAS inhibitors.
The team first compiled a dataset of 1.1 million molecules starting with 650 known KRAS inhibitors from the literature. They expanded this set by screening 100 million compounds from commercial libraries and using an algorithm to create analogs of known inhibitors. Molecules were filtered for their synthetic potential and drug-like properties.
Second, the team focused on molecular design. Hybrid models combine quantum circuit-based generative models with classical machine learning networks. 16 Kut’s quantum processor generates a “prior distribution” of molecules that refines the classical network into a viable candidate. Reward functions aligned with prioritization of KRAS binding characteristics guided an iterative training process.
The final step was experimental validation. The researchers synthesized the molecule in 15 minutes from the computational output and tested it using laboratory techniques. Two compounds, ISM061-018-2 and ISM061-022, showed strong binding to KRAS and inhibited its activity in cell-based assays without significant toxicity.
promising results
The molecule ISM061-018-2 showed broad activity in several KRAS variants, including the G12D variant commonly associated with cancer. It demonstrated high binding affinity (1.4 μM) and selective inhibition of KRAS interactions in cell-based assays. Also note: ISM061-022 displayed a clear mode of action showing increased selectivity for certain KRAS variants such as G12R and Q61H, but the effect was less pronounced for the G12D variant .
Both molecules avoided off-target effects in preliminary tests, strengthening their potential as starting points for further development.
initial
The team stops short of saying the study proves “quantum advantage.” This achieves results that cannot be achieved with classical methods. The model’s success relies on a hybrid approach, suggesting that quantum computing alone is not yet sufficient for drug discovery tasks. Current reliance on small-scale quantum processors limits the scope of molecular exploration.
Compounds were also designed in this study. This is an early stage candidate. The team wrote that work remains to optimize the efficacy and safety profile before clinical trials begin. Experimental methods such as protein crystallography are required to confirm the exact binding mechanism.
Future work: Qubit Boost, Transformer-based algorithm
The researchers plan to enhance the model by increasing the number of qubits in the quantum component and exploring transformer-based generation algorithms to improve the diversity and quality of the molecules. Scaling these approaches has the potential to further reduce the time required for drug discovery, potentially compressing years of work into months.
By showing how quantum computing can complement classical methods, this work lays the foundation for more efficient drug discovery workflows. Although this technology is still maturing, its ability to process complex datasets and generate innovative solutions points to a transformative future for pharmaceutical research.
For a deeper technical investigation, please see this paper.
The team brought together a huge number of scientists from all over the world for their research. University of Toronto researchers include Mohammad Ghazi Vakili, Aran Aspul Guzik, Jamie Snyder, Anna Ryakishva, Ardalan Hosseini Mansob, Zhong Yao, Lela Bitar, Daniel Tahoulas, Igor Staglayar, Dora・Includes Cherina. From St. Jude Children’s Research Hospital, Christoph Golgla is listed. Harvard University is represented by Christoph Gorgula, Krishna M. Padmanabha Das, and Huell Cox III. Stanford University includes Akshatkumar Nigam. Insilico Medicine AI Limited researchers are Dmitry Bezrukov, Alex Aliper, Eugene Radchenko, Xiao Ding, Jinxin Liu, Fanye Meng, Feng Ren, and Alex Zhavoronkov. Zapata AI is represented by Daniel Varoli, while Insilico Medicine Canada, Inc. includes Daniil Polykovsky. From Dana-Farber Cancer Institute, Krishna M. Padmanabha Das is listed. The University of Zagreb includes Lela Bitar, while the University of Split is represented by Dora Celina and Igor Stagljar. Aqi, Inc. includes Yudong Cao, and the Mediterranean Institute of Science (underground) is represented by Igor Stagljar.
