Research Day: Recent developments in statistical modelling of RNA velocity

RNA velocity provides an approach for inferring cellular state transitions from single-cell RNA sequencing (scRNA-seq) data, in particular from the number of molecules of spliced and unspliced mRNA. The unspliced mRNA content is a leading indicator of spliced mRNA, meaning that it is a predictor of the spliced mRNA content in the cell's near future. This causal relationship can be usefully exploited to identify directions of differentiation pathways. Mathematically, it is modelled by first order differential equations for spliced and unspliced mRNA concentration, and the derivative of the spliced mRNA concentration is the RNA velocity. Many related methods have been developed, with the two most popular implementations were released in 2017–2018: velocyto [1], which introduced the method for scRNA-seq data, and scVelo [2], which extended it to fit a more sophisticated dynamical model.

However, the reliability and relevance of such velocity estimates has been called into question [3-4], resulting in a flurry of recent improvements to address the challenges [5-9]. With many methods developed in parallel, it remains unclear how they compare and whether any of them address all the issues identified by critics.

The main focus of the discussion will be on the (lack of) reliability of RNA velocity models, their underlying assumptions, how to use the information derived from them, how to estimate the uncertainty of the output and how to improve these methods.

Schedule

10:00-10:45 Linus Schumacher  (Centre for Regenerative Medicine)  "RNA velocity: Introduction and recent developments"

10:45-11:15 coffee break

11:15-12:00 Ava Khamseh (School of Informatics and MRC Human Genetics Unit) "Stator: From higher-order interactions to cell types, subtypes and states."

12:00-12:45 Augustinas Sukys (Ramon Grima Group, School of Biology) "Transcriptome-wide single cell analysis of the variation of gene expression with cell-cycle phase"

12:45-1:30pm - lunch

1:30-2:15pm Natalia Bochkina (School of Mathematics) "Bayesian modelling of RNA velocity: identifiability and model comparison"

2:15-3pm - Brainstorming and discussion.

References 

1. RNA velocity of single cells
2. Generalizing RNA velocity to transient cell states through dynamical modeling |Nature. 
3. RNA velocity unraveled
4. Pumping the brakes on RNA velocity – understanding and interpreting RNA velocity estimates
5. UniTVelo: temporally unified RNA velocity reinforces single-cell trajectory inference | Nature Communications
6. Pyro-Velocity: Probabilistic RNA Velocity inference from single-cell data
7. Bayesian Inference of RNA Velocity from Multi-Lineage Single-Cell Data
8. Inferring single-cell transcriptomic dynamics with structured latent gene expression dynamics
9. A relay velocity model infers cell-dependent RNA velocity