Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Master's Award: Canada Graduate Scholarships

University of British Columbia

$17,500.00

Interpretable representations for gene expression states from scRNA data via transformers by injecting prior knowledge

FY 2023-24

Summary
Single cell RNA sequencing (scRNA-seq) measures the expression patterns of human genes on a cell-by-cell basis, which communicates the function and intent of individual cells from within a possibly heterogeneous population. Observing the space of possible expression patterns will reveal prevalent distributions, otherwise known as cell types, and finer features such as expression responses from environmental stimuli or perturbations and cell states. With the breakthrough machine learning approach of transformers which consists of attention mechanisms that direct the rest of the model to focus on sections of the data that are most relevant to the task at hand and their highly parallelizable architecture, transformer models like TOSICA and scGPT have been successfully applied to infer the cell type landscape from scRNA sequencing. I intend to build upon these approaches to improve their interpretability and generalizability by injecting prior knowledge through the use of pre-trained and guided representation spaces (embeddings) and higher order structures in the data such as metabolic/signalling pathways and protein-protein interaction networks. This way, prior knowledge is brought to bear to improve downstream analysis and surfacing the computation to interpretable embeddings instead of hiding it behind matrix operations. These embeddings will have the potential to act as a data-driven distillation of the existing body of science in the single-cell sequencing domain, facilitating research in the wider field of medicine, health, and biology.
Program
Master's Award: Canada Graduate Scholarships
Location
British Columbia, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database

As published

Id
505908
Name
Cai Runtong
Region
British Columbia
Country
Canada
Orgtype
Institution Paid; Research Institution
Prcname2
Special Cases - Awards Programs
Projecttitle
Interpretable representations for gene expression states from scRNA data via transformers by injecting prior knowledge
Programtype2
Studentship Programs
Orgnameinp2
University of British Columbia
Cihramount2
$17,500
Cihrequipment2
$0
Programname2
Master's Award: Canada Graduate Scholarships
Keyworddelim
Cell Type; Machine Learning; Scrna; Self-Supervised
Abstract
Single cell RNA sequencing (scRNA-seq) measures the expression patterns of human genes on a cell-by-cell basis, which communicates the function and intent of individual cells from within a possibly heterogeneous population. Observing the space of possible expression patterns will reveal prevalent distributions, otherwise known as cell types, and finer features such as expression responses from environmental stimuli or perturbations and cell states. With the breakthrough machine learning approach of transformers which consists of attention mechanisms that direct the rest of the model to focus on sections of the data that are most relevant to the task at hand and their highly parallelizable architecture, transformer models like TOSICA and scGPT have been successfully applied to infer the cell type landscape from scRNA sequencing. I intend to build upon these approaches to improve their interpretability and generalizability by injecting prior knowledge through the use of pre-trained and guided representation spaces (embeddings) and higher order structures in the data such as metabolic/signalling pathways and protein-protein interaction networks. This way, prior knowledge is brought to bear to improve downstream analysis and surfacing the computation to interpretable embeddings instead of hiding it behind matrix operations. These embeddings will have the potential to act as a data-driven distillation of the existing body of science in the single-cell sequencing domain, facilitating research in the wider field of medicine, health, and biology.
Competitiondate
202312
Approvedterm2
1 yr 0 mth
Pinamesdelim
Cai, Runtong
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $17,500~ $0~