QuantumXCT: Learning Interaction-Induced State Transformation in Cell-Cell Communication via Quantum Entanglement and Generative Modeling

Abstract

Inferring cell-cell communication (CCC) from single-cell transcriptomics remains fundamentally limited by reliance on curated ligand-receptor databases, which primarily capture co-expression rather than the system-level effects of signaling on cellular states. Here, we introduce QuantumXCT, a hybrid quantum-classical generative framework that reframes CCC as the problem of learning interaction-induced state transformations between cellular state distributions. By encoding transcriptomic profiles into a high-dimensional Hilbert space, QuantumXCT trains parameterized quantum circuits to learn a unitary transformation that maps a baseline non-interacting cellular state to an interacting state. This approach enables the discovery of communication-driven changes in cellular state distributions without requiring prior biological assumptions. We validate QuantumXCT using both synthetic data with known ground-truth interactions and single-cell RNA-seq data from ovarian cancer-fibroblast co-culture systems. The model accurately recovers complex regulatory dependencies, including feedback structures, and identifies dominant communication hubs such as the PDGFB-PDGFRB-STAT3 axis. Importantly, the learned quantum circuit is interpretable: its entangling topology can be translated into biologically meaningful interaction networks, while post hoc contribution analysis quantifies the relative influence of individual interactions on the observed state transitions. By shifting CCC inference from static interaction lookup to learning data-driven state transformations, QuantumXCT provides a generative framework for modeling intercellular communication. This work establishes a new paradigm for de novo discovery of communication programs in complex biological systems and highlights the potential of quantum machine learning in single-cell biology.

Figures (3)

• Figure 1: A Quantum Generative Framework for Modeling Cell-Cell Communication. Our workflow reframes CCC as a learned transformation between cellular states. We derive two sets of empirical probability distributions from single-cell RNA sequencing: a baseline distribution, $P(X^{\text{Mo}})$, from non-interacting (mono-culture) cells, and a target distribution, $Q(X^{\text{Co}})$, from interacting (co-culture) cells. A parameterized quantum circuit (PQC) is then trained to model this transformation. The circuit is initialized to the global baseline state $\ket{\Psi_{\text{Mo}}}$ (derived from $P(X^{\text{Mo}})$) and applies a parameterized unitary operator, $U(\tau, \theta)$, to produce a final state whose simulated marginal distributions are denoted $P_{\psi'}$. In a hybrid quantum-classical loop, an optimizer iteratively adjusts the circuit's topology ($\tau$) and parameters ($\theta$) to minimize the Kullback-Leibler (KL) divergence, $D_{\text{KL}}(P_{\psi'} \parallel Q_{\text{Co}})$, between the circuit's simulated output and the empirical target. Upon convergence, the optimized unitary, $U(\tau_{\text{opt}}, \theta_{\text{opt}})$, provides a quantitative, data-driven model of the communication dynamics.
• Figure 2: Validation of QuantumXCT on synthetic single-cell data. Our framework successfully learns the transformation from a non-interacting to an interacting cellular state using simulated data with ground-truth rules. (A) Optimization Workflow: Input joint histograms of non-interacting cells define the Initial States ($\ket{\Psi_0}$), while interacting cell histograms define the Final Target States ($\ket{\Psi_f}$). The objective is to identify a unitary transformation, $U(\tau_{\text{opt}}, \theta_{\text{opt}})$, that evolves the initial distribution to best match the target. (B) N-Wise Search Performance: Results for the Iterative Local Search (Algorithm 1, $n=2$). The panel shows the discovered quantum circuit topology, the KL divergence convergence during angle optimization, and the final state occupancy. The optimized QXCT distribution (light purple) shows high fidelity to the "Interacting" ground-truth (dark purple). (C) Multi-Epoch Search Performance: Corresponding results for the Multi-Epoch Search (Algorithm 2). Despite identifying a topologically distinct entangling circuit, the model achieves comparable convergence and occupancy matching. Both heuristics validate the framework's ability to learn the underlying dynamics and identify key regulatory nodes.
• Figure 3: QuantumXCT discovers a core communication hub in an ovarian cancer dataset. Our framework was applied to scRNA-seq data from cancer cells and fibroblasts to model their interaction. (A) N-Wise Algorithm Results: The parsimonious 3-gate quantum circuit (left) discovered by the Iterative Local Search and the corresponding biologically constrained interaction network (right). The network plot visualizes the primary intercellular link (PDGFB-PDGFRB) and the intracellular GRN connections within the identified {PDGFB, PDGFRB, STAT3} hub. (B) Multi-Epoch Algorithm Results: The larger 5-gate circuit discovered by the stochastic search heuristic. Although topologically different, the resulting network converges on the same core hub, highlighting the robustness of the framework. This solution proposes an additional link to TGFBR2, whose low impact is quantified by the contribution analysis in Table \ref{['tab:kl_comp_analysis']}, demonstrating the method's ability to distinguish driver interactions from less significant ones.