In a paper published in Proceedings of the National Academy of Sciences, researchers from the UChicago Pritzker School of Molecular Engineering and the Chemistry Department have introduced TopicVelo, a new method for utilizing static snapshots from single-cell RNA sequencing (scRNA-seq) data to study cell and gene changes over time. The team took an interdisciplinary approach, combining concepts from classical machine learning, computational biology, and chemistry. By putting together simple, established ideas from unsupervised machine learning and transcriptional modeling, TopicVelo offers a more powerful way to analyze scRNA-seq data.
ScRNA-seq data provides powerful and detailed measurements, but these measurements are static and do not show how cells transition over time. Researchers traditionally use “pseudotime” to infer these dynamic processes, making assumptions based on the similarity of cells’ transcriptional profiles. However, biological processes are more complex, involving various factors that can influence gene expression. RNA velocity approaches look at transcription dynamics within cells, offering insights into the transcription, splicing, and degradation of mRNA. TopicVelo goes beyond deterministic models to embrace a stochastic model that considers the randomness inherent in biological processes.
The team behind TopicVelo recognized that traditional methods assume all cells express the same gene program, overlooking the simultaneous performance of different processes within cells. To address this challenge, they employed probabilistic topic modeling, a machine learning tool used to identify themes from written documents. TopicVelo groups scRNA-seq data based on the processes cells and genes are involved in, inferred from the data itself. By organizing data by topics such as ribosomal synthesis, differentiation, immune response, and cell cycle, TopicVelo provides a new way to understand the dynamics of different processes in cells.
By disentangling processes and organizing them by topic, TopicVelo applies topic weights back onto cells to determine the percentage of each cell’s transcriptional profile involved in specific activities. This approach allows researchers to observe the dynamics of different processes and their significance in various cells, particularly at branch points or when cells face conflicting directions. The results of combining the stochastic model with the topic model are impressive, enabling TopicVelo to reconstruct trajectories that previously required special experimental techniques. These advancements broaden the potential applications of scRNA-seq data analysis.
The researchers’ collaborative effort and interdisciplinary approach highlight the importance of combining expertise from different fields to tackle complex research questions. TopicVelo’s ability to uncover hidden insights from scRNA-seq data demonstrates the power of integrating machine learning techniques with biological data analysis. The team’s emphasis on understanding the intrinsic randomness of biological processes and the need to account for multiple simultaneous processes within cells marks a significant advancement in single-cell transcriptomic analysis. Overall, TopicVelo represents a new frontier in studying dynamic cellular processes and gene expression changes over time.
