课前准备-单细胞新版velocity（cellrank 2）

作者，Evil Genius

参考文章CellRank 2: unified fate mapping in multiview single-cell data(nature methods)，2024年6月。

官网示例在cellrank documentation

分析框架

CellRank 2为使用马尔可夫链研究单细胞命运决策提供了统一的框架

• 自动确定初始和最终状态，计算命运概率，绘制轨迹特异性基因表达趋势图表，并识别谱系相关基因
• 采用概率系统描述，其中每个细胞构成马尔可夫链中的一个状态，边缘表示细胞-细胞转移概率
• CellRank 2提供了一组基于基因表达、RNA速率、伪时间、发育潜力、实验时间点和代谢标记数据的转换概率的不同kernels 。
• 引入了一种random walk-based的可视化方案

CellRank’s Key Applications

• Estimate differentiation direction based on a varied number of biological priors, including pseudotime, developmental potential, RNA velocity, experimental time points, and more.
• Compute initial, terminal and intermediate macrostates [="https://links.jianshu.com/go?to=https%3A%2F%2Fcellrank.readthedocs.io%2Fen%2Flatest%2Freferences.html%23id21">Reuter et al., 2019, ="https://links.jianshu.com/go?to=https%3A%2F%2Fcellrank.readthedocs.io%2Fen%2Flatest%2Freferences.html%23id4">Reuter et al., 2022].
• Infer fate probabilities and driver genes.
• Visualize and cluster gene expression trends

基于velocyto

• use scvelo to compute RNA velocity
• set up CellRank’s VelocityKernel and compute a transition matrix based on RNA velocity.
• combine the VelocityKernel with the ConnectivityKernel to emphasize gene expression similarity.
• visualize the transition matrix in a low-dimensional embedding.

#######示例

Preprocess the data

Combine RNA velocity with expression similarity in high dimensions

• Set up the VelocityKernel

默认情况下，使用确定性模式来计算transiton matrix。如果要传播速率向量中的不确定性，查看随机模式和蒙特卡罗模式。随机模式使用KNN图估计速率向量上的分布，并使用分析近似将该分布传播到过渡矩阵中。

Combine with gene expression similarity

RNA velocity can be a very noise quantity; to make our computations more robust, we combine the VelocityKernel with the similarity-based ConnectivityKernel

可视化

基于Diffusion pseudotime (DPT)

Check RNA velocity on this data

Use pseudotime to recover directed differentiation

Choosing the right pseudotime

Compute a transition matrix

生活很好，有你更好