Statistical Bioinformatics Seminar

Title: Multi-state evolutionary model quantifies tumour cellular plasticity 

Abstract: Cell-state transition dynamics are important in many diseases in cancer,
transitions among distinct cell states can affect treatment effectiveness and
metastasis.  Cancer evolutionary studies based on phylogenetics often assume weak or no
selection, especially in the recent past, to estimate effective population sizes and
evolutionary forces across a relatively long time span.  However, this assumption is
violated during tumour growth when cells rapidly proliferate and outcompete one another
under strong selection pressures.  

We adopt a stochastic modelling approach where cells divide according to a birth-death
branching process and couple their fates with a Markov model for phenotypic
transitions.  Both genotype and phenotype are simultaneously inherited by the next
generation.  Phylogenies are constructed for a subsampled population of cells and
coupled cell state information at a single time-point is used to estimate transition
rates.  We show that it is possible to determine phenotypic transition dynamics for
specific population trajectories by ‘fine-graining’ node depth levels.  This feature
is crucial for addressing the proliferation of tumours which are mixtures of clones with
different selection advantages.  

We pair our analysis with cell state annotations derived specifically for single-cell
RNA sequencing (scRNAseq) data to define evolutionary relatedness between cell
phenotypes.  We then apply our computational framework to published metastatic
pancreatic cancer phylogenies reconstructed using CRISPR-based lineage tracing in mice,
where scRNAseq information is available for each leaf.  By comparing across metastatic
locations in a single mouse, we are able to reveal changes in selection pressures and
cell phenotypic transition rates during metastatic progression.  Inferred cell state
transitions are supported by inferred RNA velocities.  

About the speaker: Dr Gladys Poon completed both her undergraduate (Physics) and
postgraduate (PhD in Oncology) studies at the University of Cambridge.  During her time
in the Blundell lab, Gladys used population genetics to quantify the expected mutation
burden and unknown drivers in human cancers, particularly in acute myeloid leukemia.
She also combined newly gained experimental expertise with her quantitative background
to interrogate clonal evolutionary patterns in leukemic bone marrow, revealing levels of
positive selection during the human lifespan.  Her research bridged between mathematics,
cancer genomics and computational modelling.  

Gladys is now with the SMA lab at HKU – working on cellular plasticity in
hepatocellular carcinoma.  This lab uses lineage tracing experiments in mice and
modelling to understand the role of cellular plasticity during tumor evolution.