Colorectal cancer (CRC) cell atlases have refined descriptive maps of tumour ecosystems, yet cross-sample integration often obscures disease-relevant patient-specific variation and remains largely correlative, limiting insight into the mechanisms and state transitions that drive progression and treatment response.
Here, we develop a continual learning framework to construct a comparative single-cell CRC atlas spanning over 300 patients and 1.5 million cells, preserving inter-patient variation while aligning healthy and malignant contexts. We resolve distinct non-canonical malignant cell states, including an endoderm-like state enriched in microsatellite-stable, KRAS-mutant CRC with features of oncofetal plasticity. Cell states are recapitulated in patient-derived organoids, establishing a tractable model of reprogramming.
By linking the observational atlas to a large-scale perturbation atlas using relative representations, we map perturbations that drive cells toward defined phenotypic extremes. We connect cell states to therapeutic responses, showing that MAPK inhibition induces a shift away from a proliferative phenotype and converges towards a plastic, endoderm-like state. Together, this framework moves beyond static atlases to enable mechanistic modeling of cell-state regulation and causal inference toward cell-state–directed therapies.
Read our preprint on bioRxiv.
Roadmap article on cancer drug-tolerant persister cells, covering their key biological features such as phenotypic plasticity and heterogeneity, adaptation to treatment, and potential to evade immune surveillance. We also cover potential therapeutic opportunities and key technologies to profile cancer drug-tolerant persister cells.
Read our Roadmap article on Nature Reviews Cancer.
Precision medicine is empowered by identifying robust drug response biomarkers. However, experimental noise in drug high-throughput screens can lead to biased estimates of biomarker-drug response associations.
Here, we introduce a novel, flexible dose-response fitting algorithm based on Gaussian processes (GPs) to produce confidence intervals for drug response. In addition to this new dose-response metric, we provide a Bayesian framework which incorporates the uncertainty estimates for the identification of biomarkers. This suggests the importance of quantifying uncertainty for in vitro drug experiments and advances the development of precision medicine by improving reproducibility of biomarker association studies.
Read our manuscript on eLife.
GP drug response fitting code repositoryComparisons of drugs have predominantly focused on identifying how individuals respond similarly, and there is a lack of data-driven insight into subpopulations of individuals with differential drug response.
We applied a novel machine learning algorithm, SEgmentation And Biomarker Enrichment of Differential treatment response (SEABED), to compare drug response between cancer drug pairs. This data-driven approach paves a new way for patient stratification to identify novel cancer subpopulations, their genetic biomarkers, and effective drug combinations.
Read our manuscript on npj Systems Biology & Applications.
SEABED code repositoryReview article on common applied AI and machine learning approaches used in translational medicine, particularly in the areas of drug discovery, imaging, and genomic medicine. We also discuss limitations of these technologies in our review article as they do not come without their limitations and shortcomings.
Read our review on EBioMedicine.