Biomarker Discovery with Quantum Neural Networks: A Case-study in CTLA4-Activation Pathways

TL;DR

The paper tackles biomarker discovery as a massive combinatorial problem and proposes a quantum neural network (QNN) framework guided by maximum relevance–minimum redundancy (mRMR) scoring to efficiently identify informative gene sets. An 11-qubit QNN architecture with a four-qubit variant demonstrates how quantum feature maps and a quadratic-programming-based objective can prioritize biomarkers, and the approach is validated on TCGA data across four CTLA4 activation pathways, yielding a top candidate gene set with cross-pathway overlaps. Key contributions include the GSCORE scoring scheme, open-source implementation, convergence analyses, and statistical/clinical significance assessments via cBioPortal and PubMed literature mining. The work highlights the potential practical impact of quantum-assisted biomarker discovery on constrained hardware and as a generalizable framework for immuno-oncology pathway analysis, while outlining avenues for integrating epigenetic markers and extending to other pathways.

Abstract

Biomarker discovery is a challenging task due to the massive search space. Quantum computing and quantum Artificial Intelligence (quantum AI) can be used to address the computational problem of biomarker discovery tasks. We propose a Quantum Neural Networks (QNNs) architecture to discover biomarkers for input activation pathways. The Maximum Relevance, Minimum Redundancy (mRMR) criteria is used to score biomarker candidate sets. Our proposed model is economical since the neural solution can be delivered on constrained hardware. We demonstrate the proof of concept on four activation pathways associated with CTLA4, including (1) CTLA4-activation stand-alone, (2) CTLA4-CD8A-CD8B co-activation, (3) CTLA4-CD2 co-activation, and (4) CTLA4-CD2-CD48-CD53-CD58-CD84 co-activation. The model indicates new biomarkers associated with the mutational activation of CLTA4-associated pathways, including 20 genes: CLIC4, CPE, ETS2, FAM107A, GPR116, HYOU1, LCN2, MACF1, MT1G, NAPA, NDUFS5, PAK1, PFN1, PGAP3, PPM1G, PSMD8, RNF213, SLC25A3, UBA1, and WLS. We open source the implementation at: https://github.com/namnguyen0510/Biomarker-Discovery-with-Quantum-Neural-Networks.

Figures (6)

• Figure 1: The Ansatz Circuit of Our Proposed QNN Models using Four Qubits with Four Neural Blocks. Here, the number of evaluated genes is $2^4 = 16$ genes. We extend the architecture to $11$-qubit ansatz in the numerical result, scoring $2^{11} = 2,048$ genes.
• Figure 1: Experimental Report of Quantum AI-driven Genetic Biomarkers Discovered for CTLA4 Pathway. (Top) GSCORE$^\copyright$ using top-$50\%$ samplers. The neural solutions are well-converged as the score variation is under $200\mu, \mu = 10^{-6}$. (Bottom) Convergence analysis of quantum sampler. The model configuration with a lower score is in darker color (purple), and the model configuration with a higher score is in brighter color (yellow).
• Figure 2: (A)Activation Pathways Analyzed in Our Case-study: We expanded the scope of immunological research by studying the coactivation of CTLA4, CD2, and associated genes, including CD48, CD53, CD58, and CD84. These molecules play significant roles in T-cell immune regulation, and their interactions could potentially lead to the development of more effective therapeutic strategies for various immune-related diseases. (B)Venn Diagram of Discovered-Genetic Biomarker Sets regarding The Quantified Targeted Pathways:MACF1, a protein facilitating actin-microtubule interactions at the cell periphery, is a common genetic biomarker for both CTLA4 and CTLA4-CD8A-CD8B pathways. On the other hand, HSPA1B, a member of the heat shock protein 70 family that stabilizes existing proteins against aggregation, is the common genetic biomarker for CTLA4-CD8A-CD8B and CTLA4-CD2 pathways. These findings highlight the potential of these biomarkers in understanding immune regulation and developing therapeutic strategies, which has not yet been well-studied, discussed in Section \ref{['sec:casestudy_ctla4']}.
• Figure 2: Experimental Report of Quantum AI-driven Genetic Biomarkers Discovered for CTLA4-CD8A-CD8B Pathway. (Top) GSCORE$^\copyright$ using top-$50\%$ samplers. The neural solutions are well-converged as the score variation is under $200\mu, \mu = 10^{-6}$. (Bottom) Convergence analysis of quantum sampler. The model configuration with a lower score is in darker color (purple), and the model configuration with a higher score is in brighter color (yellow).
• Figure 3: Experimental Report of Quantum AI-driven Genetic Biomarkers Discovered for CTLA4-CD2 Pathway. (Top) GSCORE$^\copyright$ using top-$50\%$ samplers. The neural solutions are well-converged as the score variation is under $200\mu, \mu = 10^{-6}$. (Bottom) Convergence analysis of quantum sampler. The model configuration with a lower score is in darker color (purple), and the model configuration with a higher score is in brighter color (yellow).

Theorems & Definitions (1)

• Definition 1