Hajhashemi, S. | bioRxiv Science

bioRxiv2026

A Measurable Systemic Immune Phenotype Links Circulating Myeloid Dysfunction to Tumour Spatial Organization in Gastroesophageal Adenocarcinoma

Background: Both static and longitudinal measures of circulating neutrophil-to-lymphocyte ratio (cNLR) are prognostic markers across solid malignancies, including gastroesophageal adenocarcinoma (GEA). Despite this consistency, cNLR remains poorly integrated into clinical decision-making and is generally regarded as a nonspecific inflammatory or disease-burden signal rather than a defined biological state. Analysis of a large, prospectively assembled GEA cohort showed that cNLR was only weakly associated with tumour burden or pathological response, yet both baseline and postoperative cNLR independently predicted outcome. Methods: These observations prompted a retrospective biological investigation of what host biology cNLR captures. Using available clinically linked biospecimens, we performed multiscale analyses across complementary patient subsets encompassing peripheral-blood immune function and soluble signalling, tumour single-cell transcriptional programmes, and spatial myeloid organization within tissue. Results: High baseline cNLR was associated with reduced circulating cytotoxic and immune-trafficking mediators, elevated angiogenesis-associated factors, neutrophil-mediated suppression of lymphocyte proliferation, impaired tumour-cell killing, and enhanced spontaneous neutrophil extracellular trap formation. Single-cell profiling identified coordinated transcriptional differences across tumour compartments, including innate, cytokine and myeloid programmes. Spatial profiling showed that myeloid abundance and organization varied across anatomical compartments; greater tumour-periphery PDL1 positive myeloid clustering was associated with inferior recurrence-free survival. Baseline cNLR showed little relationship with static myeloid abundance, whereas pretreatment cNLR dynamics were associated with altered cross-compartment spatial relationships between CD8 positive cells and myeloid cells. Notably, higher early postoperative cNLR was associated with greater PDL1 positive myeloid clustering at the tumour periphery, linking the spatial architecture of the resected tumour with a systemic myeloid phenotype persisting after surgery. Conclusions: The data identify a prognostic, multilevel myeloid-associated host phenotype in GEA that is incompletely explained by tumour burden. Integrating systemic immune function, tumour single cell transcriptomics and spatial myeloid organization demonstrates that these biological domains provide distinct, yet convergent views of the host immune phenotype captured by cNLR. These findings establish a biological framework for the clinical utility of cNLR and highlight myeloid-associated immune processes as promising targets for therapeutic intervention.

cancer biology↗

bioRxiv2026

DiffDose: Differentiable Programming for Personalized Dose-Regimen Optimal Control

Dose-regimen design requires choosing how much drug to give, when to give it, and how treatment should vary across patients. Mechanistic pharmacokinetic-pharmacodynamic (PK/PD) and quantitative systems pharmacology (QSP) models can predict treatment responses, but optimizing dosing inputs depends on model-specific sensitivity derivations or derivative-free search. Here, we introduce DiffDose, a differentiable programming framework for mechanistic open-loop dose-regimen optimization that uses automatic differentiation (AD) to handle clinically interpretable dose amounts and administration times as differentiable controls. We evaluate our method in three settings: fixed-schedule dose-amplitude optimization in OptiDose PK/PD benchmarks; dose-timing optimization in a chemotherapy-induced neutropenia model with state-dependent delay; and individualized mosunetuzumab dosing in a QSP virtual population. Across these examples, AD produced gradients consistent with references, reduced model-specific derivative work, and shortened benchmark time to solution. DiffDose thereby turns mechanistic PK/PD and QSP models from tools that evaluate prespecified regimens into gradient-based engines for individualized regimen design.

systems biology↗

bioRxiv2026

A latent space thermodynamic model of cell differentiation

Inferring the governing dynamics of differentiation that capture cell state evolution remains a central challenge in single-cell biology. We present Latent Space Dynamics (LSD), a thermodynamics-inspired framework that models cell differentiation as evolution on a learned Waddington landscape in latent space. LSD jointly infers a low-dimensional cell state, a differentiable potential function governing developmental flow, and a local entropy term that quantifies cellular plasticity. Using a neural ordinary differential equation, LSD reconstructs continuous differentiation trajectories from time-ordered single-cell data. Across diverse developmental systems, LSD accurately recovers lineage hierarchies, predicts fate commitment for unseen cell types, and outperforms existing trajectory inference approaches in directional accuracy. Moreover, in silico gene perturbations reveal how individual regulators reshape the landscape, and entropy provides a quantitative measure of plasticity in development and cancer.

bioinformatics↗