• Smita Krishnaswamy, Yale University, United States

• Laura Richards, University of Toronto, Canada
• Owen Whitley, University of Toronto, Canada
• Florence Cavalli, The Hospital for Sick Children, Canada
• Fiona Coutinho, The Hospital for Sick Children, Canada
• Michelle Kushida, The Hospital for Sick Children, Canada
• Nataliia Svergun, The Hospital for Sick Children, Canada
• Nagmeh Rastegar, The Hospital for Sick Children, Canada
• H. Artee Luchman, University of Calgary, Canada
• Samuel Weiss, University of Calgary, Canada
• Peter Dirks, The Hospital for Sick Children, Canada
• Gary Bader, University of Toronto, Canada
• Trevor Pugh, University Health Network, Princess Margaret Cancer Centre, Canada

Glioblastomas (GBM) harbour diverse populations of cells, including a rare subpopulation of glioblastoma stem cells (GSCs) that drive tumourigenesis. To characterize functional diversity within the tumour-initiating fraction of GBM, we performed single-cell RNA-sequencing on >69,000 GSCs cultured from the tumours of 26 patients. We identified a high degree of transcriptional heterogeneity within GSCs, with 2-6 transcriptional clonotypes per sample. Inference of copy number variation (CNVs) from scRNA-seq data revealed subclonal somatic CNVs drive a portion of transcriptional diversity within GSCs, but do not fully explain inter-sample heterogeneity. Instead, we found GSCs mapped along a transcriptional gradient spanning two cellular states reminiscent of normal neural developmental and inflammatory wound response processes. Further single-cell RNA-sequencing of patient tumours found the majority of cancer cells organize along an astrocyte maturation gradient orthogonal to the GSC gradient yet retained expression of founder transcriptional programs found in GSCs. Our results support a model whereby GBM heterogeneity can be explained by a fundamental GSC-based gradient between neural tissue regeneration and wound response transcriptional programs. This new paradigm has implications for understanding GBM origins with clinical implications for designing targeted therapies against the heterogeneous tumour-initiating fraction of GBM.

• Claudia Kleinman, McGill University, Canada

• Benjamin J. Raphael, Princeton University, United States

• Xiangqi Bai, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, China
• Lin Wan, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, China
• Li C. Xia, Department of Medicine, Stanford University School of Medicine, United States

Latest high-throughput single-cell RNA-sequencing (scRNA-seq) and DNA-sequencing (scDNA-seq) technologies enabled cell-resolved investigation of pathological tissue clones. However, it is still technically challenging to simultaneously measure the genome and transcriptome content of a single cell. In this work, we developed CCNMF – a new computational tool utilizing the Coupled-Clone Non-negative Matrix Factorization technique to jointly infer clonal structures in single-cell genomics and transcriptomics data. We benchmarked CCNMF using both simulated and real cell mixture derived datasets and fully demonstrated its robustness and accuracy. We also applied CCNMF to the paired scRNA and scDNA data from a triple-negative breast cancer xenograft, resolved its underlying clonal structures, and identified differential genes between cell clusters. In summary, CCNMF presents a joint and coherent approach to resolve the clonal genome and transcriptome structures, which will facilitate a better understanding of the cellular and tissue changes associated with disease development.

• Yingxin Lin, The University of Sydney, Australia
• Jean Yang, The University of Sydney, Australia
• Ellis Patrick, The University of Sydney, Australia

Complex cancer progression, such as tumour growth and invasion, is known to be impacted by the immune system and its spatial interaction with tumours. Two recent studies of breast cancer using multiplexed ion beam imaging by time-of-flight (MIBI-TOF) and imaging mass cytometry (IMC) have revealed that single-cell heterogeneity within the spatial tumour-microenvironment context is closely associated with patient subtypes (Keren et al. 2018, Jackson et al. 2020). However, a complete understanding of the ability of different features generated from these state-of-the-art technologies, such as cell type composition and tumour spatial patterns, to predict patient clinical outcomes is still limited.

To this end, we will introduce a procedure to identify informative spatial and non-spatial features and develop a corresponding model to predict patient survival outcomes. We investigate integration methods such as optimal transport to reconstruct the spatial pattern of additional antibodies not measured in the imagining data using information from CyTOF. This allows us to expand the feature space by characterising the spatial pattern of existing and imputed protein expression, individual cell type, and cell type interaction using spatial statistics. Finally, we generalize the predictive features from different experiments and evaluate them via patient survival prediction performance.

• Sohrab Shah, Memorial Sloan Kettering Cancer Center, United States

• Lingfei Wang, Broad Institute of MIT and Harvard, United States
• Jacques Deguine, Broad Institute of MIT and Harvard, United States
• Ramnik Xavier, Broad Institute of MIT and Harvard, United States

Single-cell RNA sequencing (ScRNA-seq) may provide unprecedented technical and statistical power to study gene expression and regulation within and across cell-types. However, due to its sparsity and technical variations, developing a superior single-cell computational method for differential expression (DE) and co-expression remains challenging. Here we present Normalisr, a parameter-free normalization-association two-step inferential framework for scRNA-seq that solves case-control DE, co-expression, and pooled CRISPRi scRNA-seq screen under one umbrella of linear association testing. Normalisr addresses those challenges with posterior mRNA abundances, nonlinear cellular summary covariates, and mean and variance normalization. All these enable linear association testing to achieve optimal sensitivity, specificity, and speed in all above scenarios. Normalisr recovers high-quality transcriptome-wide co-expression networks from conventional scRNA-seq of T cells in human melanoma and robust gene regulations from pooled CRISPRi scRNA-seq screens. Normalisr provides a unified framework for optimal, scalable hypothesis testings in scRNA-seq.