Oral Presentation 44th Lorne Genome Conference 2023

FitnessTracker identifies transcriptional programs driving clonal evolution from barcoded single-cell RNA-sequencing data (#36)

David L Goode 1 2 , Alex Casar Berazaluce 1 3 , Brenda Sanchez Pichardo 4 , Dane Vassiliadis 2 5 , Howard Bondell 3 6
  1. Evolutionary Cancer Genomics Laboratory, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  2. The Sir Peter MacCallum Department of Oncology, University of Melbourne, Melbourne, VIC, Australia
  3. Mathematics and Statistics, UNIVERSITY OF MELBOURNE, Melbourne, Victoria, Australia
  4. Genomic Sciences, National Autonomous University of Mexico, Cuernavaca, Morelos, Mexico
  5. Cancer Biology and Therapeutics Program, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  6. Melbourne Centre for Data Science, University of Melbourne, Melbourne, Victoria, Australia

Clonal evolution is universal in biology. Competition between related but distinct clonal sub-populations of cells shapes properties of biological systems in both natural settings and in human diseases such as cancer. Differences at the transcriptional level can confer survival advantages to clones, enhancing their fitness relative to competitors.

By linking frequency of clonal sub-populations to their transcriptional profiles, the use of barcoded transcripts combined with single-cell RNA-sequencing (barcoded scRNA-seq) is revolutionising our understanding of how changes in gene regulation drive clonal evolution.

We developed FitnessTracker to identify transcriptional programs associated with clonal fitness from barcode scRNA-seq data. FitnessTracker applies a Random Forest machine learning algorithm to the transcriptional profiles of barcoded clonal sub-populations to find genes whose expression can predict clone frequency within and between conditions. Predictive genes are refined into transcriptional programs by clustering to identify sets of genes commonly co-expressed in the same cells. Scoring activity of transcriptional programs across clones determines the contribution of each program to clonal fitness.

Applying FitnessTracker to barcoded scRNA-seq from mouse models of Acute Myeloid Leukemia (AML) successfully uncovered known transcriptional programs controlling differences in proliferation rates between clones. Further, it accurately predicted AML clone size based on transcriptional profiles (R2 >0.8) in both in vitro and in vivo environments.

FitnessTracker identified transcriptional programs that enhanced ability of clones to colonise murine hosts, including a previously reported immune recognition program involving the Slpi gene. Our results reveal how diverse programs involving translation, stemness, mRNA transport and stress response contribute to clonal fitness, disease progression and response to therapy in AML and key genes behind their activity.

By focusing on transcriptional programs, FitnessTracker provides a powerful new way to link gene regulation to clonal fitness from barcoded scRNA-seq data, applicable to any context where clonal evolution occurs including cancer, immunity, and microbiology.