Breast cancer remains one of the leading causes of cancer-related mortality worldwide. While HER2-positive breast cancers were once associated with poor outcomes, the introduction of effective HER2-targeted therapies has markedly improved survival. In contrast, triple-negative breast cancer continues to have poorer prognosis and limited treatment options. Single-cell multi-omics data has revolutionized our understanding of breast cancer heterogeneity by revealing the complex cellular ecosystems within tumours. However, analysing these large and high-dimensional multi-omics datasets for biological target discovery has many challenges. We applied quantum algorithms to single-cell patient data that offers a unique advantage of handling complex biological function while preserving critical relationships between genes. We could capture precision oncology and identify targets that may serve as potential therapeutic targets or biomarkers for cell-type specific breast cancer.

The in silico validation of potential drug targets through computational methods provides a rapid and cost-effective screening approach, before proceeding to expensive wet-lab validation studies. For diseases with limited therapeutic options such as breast cancer, such systematic triage is critical to accelerate translational development. We have conducted an in silico, systematic and multistep validation and prioritised the candidate genes based on their biological relevance, novelty, druggability, safety and the competitive research landscape. Our process began by assembling an extensive list of candidate genes that was the output from our quantum led analysis of single-cell patient data. The initial pool included differentially expressed transcription factors, DNA damage response mediators, epigenetic regulators, kinases and tumour microenvironment associated genes. Each gene was evaluated on following criteria:

• Biological relevance – We performed systematic literature mining and extracted each gene’s reported association with breast cancer, particularly with its subtypes. We examined whether the gene is recurrently altered, overexpressed or functionally required in TNBC models or patient samples. Genes with existing evidence for their involvement in breast cancer biology were scored high and prioritised.
• Novelty – Novelty captured whether the target represents an underexplored mechanism within HER2-negative or TNBC. Genes associated with established pathways for breast cancer biology were classified as low novelty, whereas the genes that showed emerging functional evidence for HER2- negative or TNBC biology but remain underexploited pharmacologically wereprioritised as novel opportunities.
• Druggability – We assessed whether the target belongs to a class amenable to small molecules, antibodies or emerging modalities to assess the likelihood that the protein’s structure and function can be modulated. Targets with well- defined catalytic or ligand-binding pockets were rated high. In contrast, transcription factors, although biologically important, were scored lower due to historical challenges in drug development.
• Safety considerations – On-target toxicity risks were considered for each target. Literature based safety filters were applied to identify targets whose inhibition may cause systemic toxicity in patients and were deprioritised.
• Competitive landscape – This category evaluated the maturity and saturation of ongoing drug development efforts. Targets with multiple approved or late-stage clinical agents were deprioritised as compared to early preclinical development agents.

Each gene was scored across these dimensions, and the results were formatted into a structured priority table. This allowed us to triage the list into high, medium and low priority categories for further in vitro validations. Genes with limited breast cancer data, poor tractability or high safety risks were deprioritised, ensuring focus on actionable biology.

Our quantum-enhanced analysis approach has successfully identified several well- established clinical biomarkers for breast cancer such as HER2/ERBB2 and oestrogen receptor (ESR1), providing validation of our methodology. Additionally, we have also successfully identified established immune checkpoint targets, various T-cell activation markers, established chemotherapy resistance markers and DNA repair pathway components including BRCA1 and BRCA2. Identification of these known targets serves as an important benchmark for our quantum algorithm performance and demonstrate that our methodology can recover clinically validated therapeutic targets from single-cell data. This provides confidence that novel targets identified through the same approach might have similar therapeutic relevance and higher success in translational studies.

Beyond validating known biomarkers, quantum algorithms have identified several novel therapeutic targets with potential clinical relevance and patient stratification. Quantum algorithms have revealed novel regulatory interactions and network modules that were not apparent through classical analysis methods. These include previously uncharacterized transcription factor networks, microRNA regulatory circuits, and epigenetic modification pathways that contribute to breast cancer heterogeneity. Our nominated targets for Elixion’s drug discovery pipeline represent cell-type specific therapeutic opportunities for breast cancer that might have been
missed by classical analytical methods.

Elixion’s design phase begins with targets identified through quantum-enhanced single-cell analytics, which provides a level of resolution far beyond what traditional R&D or standard AI-biotechs can achieve. Our cell-type-specific therapeutic targets
have led to a drug design approach that is inherently more precise and differentiated, significantly increasing therapeutic focus and minimising off-target toxicity.

Our in silico drug design framework integrates traditional drug design, quantum-based modelling and artificial intelligence to accelerate the design of given modalities with molecular precision. For small molecules we have employed deep generative models and quantum chemistry–informed simulations to explore chemical space and predict structure–activity relationships. Our hybrid quantum–classical approach allows accurate computation of binding free energies, charge distributions and electronic interactions. This methodology will improve our initial virtual screening, hit-to-lead transition and optimization of pharmacokinetic and physicochemical properties to refine candidate selection for downstream validation.

Computational target identification and drug design, whether classical or quantum- enhanced, requires rigorous experimental validation before clinical translation. The validation approaches used for classically identified targets would also apply to our
nominated candidates. Ultimately, this in silico strategy sets the foundation for a precision medicine pipeline in HER2-negative and TNBC, bridging computational predictions with laboratory and clinical development.

The next critical steps involve integrating these nominated targets into Elixion’s drug discovery pipeline for compound identification and optimization. Following molecule design, the program transitions swiftly to Lead optimisation which runs in parallel with in vitro validation of our target’s biology and mechanism of action, including functional perturbation studies and animal model validation. We are aiming to maximise the precision advantage gained from the initial quantum analysis to ensure that only high-fidelity target candidates proceed to the next drug discovery stage. Our precision medicine pipeline holds significant promise for improving cancer
patient outcomes through more effective target identification and therapeutic development strategies while dramatically reducing overall R&D cost and time-to- validation.