During the reporting period, the project has focused on establishing the experimental and computational foundations required to study tumor heterogeneity and develop predictive models of therapeutic response at subclone and single-cell resolution, in line with the objectives described in the Description of Action.
The progress achieved during the reporting period directly supports the project objectives by establishing the experimental and computational foundations required to (i) identify drugs with low cross-resistance to consolidation therapy, (ii) develop predictive models of drug response at single-cell resolution, and (iii) enable subsequent functional validation of candidate therapies and rare-cell resistance mechanisms.
High-throughput drug perturbation experiments
We established high-throughput drug screening pipelines and performed screening across ten neuroblastoma organoid models and seven neuroblastoma cell lines. RNA sequencing of treated organoids is currently ongoing to link phenotypic drug responses to transcriptional programs. These experiments generate the first perturbation datasets that will support downstream single-cell and predictive modelling analyses.
Quantifying and benchmarking intratumor heterogeneity
Early experiments highlighted the importance of robust quantification of tumor heterogeneity, motivating the development and benchmarking of improved metrics. We generated benchmarking datasets and initiated systematic comparisons of multiple heterogeneity measurements, including approaches rarely used in single-cell analyses, establishing a more rigorous framework for defining and quantifying intratumor heterogeneity. In parallel, we performed a systematic comparison between cisplatin and ENU as strategies to experimentally induce heterogeneity, establishing an experimental framework to generate controlled diversity in tumor models.
Functional lineage tracing and single-cell workflows
We implemented expressed cell barcode (ECB) lineage tracing in neuroblastoma models and performed single-cell RNA sequencing on barcoded populations to assess barcode capture rates and establish workflows for lineage-resolved single-cell analysis. These efforts lay the groundwork for linking lineage history, heterogeneity, and therapy response.
Development of CRISPR and single-cell perturbation methodologies
We initiated the development of CRISPR screening methodologies enabling dual (and higher-order) perturbations compatible with single-cell sequencing readouts. Although not part of the original project plan, this direction was pursued to support the development of single-cell perturbation prediction models in a more controlled genetic setting compared to drug response. The rapidly evolving field of single-cell perturbation modelling highlights CRISPR-based perturbations as a tractable intermediate step toward predicting complex drug responses. This strategic extension strengthens the project’s core objective of predicting therapy responses in heterogeneous tumor populations.
Computational infrastructure for perturbation prediction
We established a computational pipeline to benchmark commonly used deep-learning models for single-cell perturbation prediction. This effort included downloading, curating, and integrating approximately 150 million publicly available single-cell profiles. To support model development and benchmarking, we developed a synthetic data generator tailored for single-cell perturbation experiments.
Publications in preparation
These activities have resulted in several manuscripts in preparation, including studies on high-throughput drug screening, quantitative frameworks for intratumor heterogeneity, and synthetic data generation for single-cell perturbation prediction.
