Vasanthakumari, P. | bioRxiv Science

bioRxiv2026

sctrial: Participant-Level Differential Analysis for Longitudinal Single-Cell Experiments

Longitudinal single-cell RNA sequencing studies in clinical trials and translational cohorts offer a powerful view of treatment response, disease progression, and cellular dynamics, but their hierarchical structure poses a major inferential challenge: thousands of cells are measured within the same participants across repeated time points, whereas the participant, not the cell, is the true unit of biological replication. Conventional cell-level workflows can therefore yield inflated significance and misleading confidence in reported associations. Here, we present sctrial, an open source analytical framework for repeated-measures single-cell studies that uses design-specific participant-level estimands, including difference-in-differences for two-group longitudinal comparisons, and small-cluster-aware uncertainty quantification. In simulation benchmarks using a hierarchical gamma-Poisson generative model, sctrial maintained well-calibrated error rates in mixed-signal gene panels where established multi-subject methods showed inflated false positive rates among unaffected genes. We applied sctrial to five independent datasets spanning melanoma immunotherapy, COVID-19 severity, BNT162b2 vaccination, AML chemotherapy, and CAR-T therapy. Across these studies, sctrial identified immune programs whose direction and magnitude differed across therapeutic and disease contexts, while benchmarking analyses showed that many associations highlighted by conventional cell-level workflows were attenuated or no longer supported when inference was performed at the participant level. These analyses illustrate how participant-aware inference can reduce pseudoreplication-driven signal inflation and provide a more rigorous basis for interpreting longitudinal single-cell data. sctrial enables reproducible participant-level analysis of longitudinal single-cell experiments and facilitates more reliable biological interpretation in translational and clinical studies. The software is implemented in Python and compatible the AnnData/scverse ecosystem.

bioinformatics↗

bioRxiv2024

Data Imbalance in Drug Response Prediction - Multi-Objective Optimization Approach in Deep Learning Setting

Drug response prediction (DRP) methods tackle the complex task of associating the effectiveness of small molecules with the specific genetic makeup of the patient. Anti-cancer DRP is a particularly challenging task requiring costly experiments as underlying pathogenic mechanisms are broad and associated with multiple genomic pathways. The scientific community has exerted significant efforts to generate public drug screening datasets, giving a path to various machine learning (ML) models that attempt to reason over complex data space of small compounds and biological characteristics of tumors. However, the data depth is still lacking compared to computer vision or natural language processing domains, limiting current learning capabilities. To combat this issue and increase the generalizability of the DRP models, we are exploring strategies that explicitly address the imbalance in the DRP datasets. We reframe the problem as a multi-objective optimization across multiple drugs to maximize deep learning model performance. We implement this approach by constructing Multi-Objective Optimization Regularized by Loss Entropy (MOORLE) loss function and plugging it into a Deep Learning model. We demonstrate the utility of proposed drug discovery methods and make suggestions for further potential application of the work to promote equitable outcomes in the healthcare field. Availabilityhttps://github.com/AlexandrNP/MOORLE Contactonarykov@anl.gov

cancer biology↗