Gene co-expression graphs are a rich source of information, revealing critical insights into cellular functions, states, and activities. Yet, extracting meaningful signals from these graphs presents a formidable challenge. This complexity arises due to the presence of multiple, overlapping sources of information and the inherent noise, which is particularly pronounced in data derived from single-cell sequencing. In this talk, we explore our latest endeavors and progress in unraveling these intricate signals in gene co-expression graphs. In this talk, we focus on the identification and interpretation of various topological structures within these graphs. We demonstrate how these structures can help in pinpointing feature genes and analyzing cellular pathway activities. We also develop novel Graph Neural Network (GNN) encoders, and new testing frameworks to explore cell heterogeneities and functions. The practical implications of these advancements are profound. By applying these tools, we have successfully identified novel markers of cellular senescence and distinct populations of senescent cells.

Zoom Link