Frontiers in Genomics: Shushkov Philip
Computational inference and integration in single-cell transcriptomics

Resumen de la plática
Understanding the transcriptional dynamics at the single-cell level is a central challenge in deciphering developmental processes. Current sequencing technologies allow the measurement of the transcription profiles of hundreds of thousands of individual cells. However, integration of multiple single-cell transcriptomic datasets and inference of developmental progressions of cells is hampered by sampling sparsity, sequencing depth, technological noise, and loss of spatial and temporal information. We employ deep learning models to infer biologically relevant transcriptomic variation and to enable the integration of single-cell transcriptomes derived from different organs and developmental stages. We assemble with these methods a single-cell transcriptomic atlas of the model Arabidopsis that encompasses all vegetative and floral organs across several developmental stages. Furthermore, we use noise-resilient diffusion mapping-based approaches to infer and recapitulate developmental trajectories from large single-cell datasets. Our methods allow us to reconstruct trajectories that occur on substantially different timescales in development and infer waves of gene expression changes along the developmental progressions. We analyze the development of vessel elements during histogenesis and compare it to the development of other vascular cell types, including sieve elements and companion cells, in different plant organs. We recover known regulators of vascular development that follow expected developmental dynamics based on previous studies and identify putative regulatory factors. We identify waves of gene expression changes of transcription factors and their downstream targets, which provide substantial resolution of transcriptional dynamics in vascular cell development. Finally, we demonstrate how to extend our algorithms for single-cell transcriptomic inference to integrate single-cell datasets across different species. The multi-species single-cell atlas allows us to estimate conservation and divergence of transcriptional programs among species at different phylogenetic distances and genomic complexity.

Shushkov Philip. Department of Chemistry. Indiana University, USA

https://www.chem.indiana.edu/faculty/philip-shushkov/

Frontiers in Genomics