scvelo

Name: scvelo
Author: LeonChaoX

698stars61forks

How to install scvelo?

leonchaox/qinyan-academic-skills/scvelo

$npx -y skills add LeonChaoX/qinyan-academic-skills --skill scvelo

Installs into the current project.

Summary

RNA velocity analysis with scVelo. Estimate cell state transitions from unspliced/spliced mRNA dynamics, infer trajectory directions, compute latent time, and identify driver genes in single-cell RNA-seq data. Complements Scanpy/scVI-tools for trajectory inference.

Files · 1

View on GitHub

SKILL.md

1	# scVelo — RNA Velocity Analysis
2
3	## Overview
4
5	scVelo is the leading Python package for RNA velocity analysis in single-cell RNA-seq data. It infers cell state transitions by modeling the kinetics of mRNA splicing — using the ratio of unspliced (pre-mRNA) to spliced (mature mRNA) abundances to determine whether a gene is being upregulated or downregulated in each cell. This allows reconstruction of developmental trajectories and identification of cell fate decisions without requiring time-course data.
6
7	Installation: `pip install scvelo`
8
9	Key resources:
10	- Documentation: https://scvelo.readthedocs.io/
11	- GitHub: https://github.com/theislab/scvelo
12	- Paper: Bergen et al. (2020) Nature Biotechnology. PMID: 32747759
13
14	## When to Use This Skill
15
16	Use scVelo when:
17
18	- Trajectory inference from snapshot data: Determine which direction cells are differentiating
19	- Cell fate prediction: Identify progenitor cells and their downstream fates
20	- Driver gene identification: Find genes whose dynamics best explain observed trajectories
21	- Developmental biology: Model hematopoiesis, neurogenesis, epithelial-to-mesenchymal transitions
22	- Latent time estimation: Order cells along a pseudotime derived from splicing dynamics
23	- Complement to Scanpy: Add directional information to UMAP embeddings
24
25	## Prerequisites
26
27	scVelo requires count matrices for both unspliced and spliced RNA. These are generated by:
28	1. STARsolo or kallisto\|bustools with `lamanno` mode
29	2. velocyto CLI: `velocyto run10x` / `velocyto run`
30	3. alevin-fry / simpleaf with spliced/unspliced output
31
32	Data is stored in an `AnnData` object with `layers["spliced"]` and `layers["unspliced"]`.
33
34	## Standard RNA Velocity Workflow
35
36	### 1. Setup and Data Loading
37
38	```python
39	import scvelo as scv
40	import scanpy as sc
41	import numpy as np
42	import matplotlib.pyplot as plt
43
44	# Configure settings
45	scv.settings.verbosity = 3 # Show computation steps
46	scv.settings.presenter_view = True
47	scv.settings.set_figure_params('scvelo')
48
49	# Load data (AnnData with spliced/unspliced layers)
50	# Option A: Load from loom (velocyto output)
51	adata = scv.read("cellranger_output.loom", cache=True)
52
53	# Option B: Merge velocyto loom with Scanpy-processed AnnData
54	adata_processed = sc.read_h5ad("processed.h5ad") # Has UMAP, clusters
55	adata_velocity = scv.read("velocyto.loom")
56	adata = scv.utils.merge(adata_processed, adata_velocity)
57
58	# Verify layers
59	print(adata)
60	# obs × var: N × G
61	# layers: 'spliced', 'unspliced' (required)
62	# obsm['X_umap'] (required for visualization)
63	```
64
65	### 2. Preprocessing
66
67	```python
68	# Filter and normalize (follows Scanpy conventions)
69	scv.pp.filter_and_normalize(
70	adata,
71	min_shared_counts=20, # Minimum counts in spliced+unspliced
72	n_top_genes=2000 # Top highly variable genes
73	)
74
75	# Compute first and second order moments (means and variances)
76	# knn_connectivities must be computed first
77	sc.pp.neighbors(adata, n_neighbors=30, n_pcs=30)
78	scv.pp.moments(
79	adata,
80	n_pcs=30,
81	n_neighbors=30
82	)
83	```
84
85	### 3. Velocity Estimation — Stochastic Model
86
87	The stochastic model is fast and suitable for exploratory analysis:
88
89	```python
90	# Stochastic velocity (faster, less accurate)
91	scv.tl.velocity(adata, mode='stochastic')
92	scv.tl.velocity_graph(adata)
93
94	# Visualize
95	scv.pl.velocity_embedding_stream(
96	adata,
97	basis='umap',
98	color='leiden',
99	title="RNA Velocity (Stochastic)"
100	)
101	```
102
103	### 4. Velocity Estimation — Dynamical Model (Recommended)
104
105	The dynamical model fits the full splicing kinetics and is more accurate:
106
107	```python
108	# Recover dynamics (computationally intensive; ~10-30 min for 10K cells)
109	scv.tl.recover_dynamics(adata, n_jobs=4)
110
111	# Compute velocity from dynamical model
112	scv.tl.velocity(adata, mode='dynamical')
113	scv.tl.velocity_graph(adata)
114	```
115
116	### 5. Latent Time
117
118	The dynamical model enables computation of a shared latent time (pseudotime):
119
120	```python
121	# Compute latent time
122	scv.tl.latent_time(adata)
123
124	# Visualize latent time on UMAP
125	scv.pl.scatter(
126	adata,
127	color='latent_time',
128	color_map='gnuplot',
129	size=80,
130	title='Latent time'
131	)
132
133	# Identify top genes ordered by latent time
134	top_genes = adata.var['fit_likelihood'].sort_values(ascending=False).index[:300]
135	scv.pl.heatmap(
136	adata,
137	var_names=top_genes,
138	sortby='latent_time',
139	col_color='leiden',
140	n_convolve=100
141	)
142	```
143
144	### 6. Driver Gene Analysis
145
146	```python
147	# Identify genes with highest velocity fit
148	scv.tl.rank_velocity_genes(adata, groupby='leiden', min_corr=0.3)
149	df = scv.DataFrame(adata.uns['rank_velocity_genes']['names'])
150	print(df.head(10))
151
152	# Speed and coherence
153	scv.tl.velocity_confidence(ad

People who liked this also viewed

6 picks

Type

scvelo

698stars61forks

View on GitHub

How to install scvelo?

leonchaox/qinyan-academic-skills/scvelo

$npx -y skills add LeonChaoX/qinyan-academic-skills --skill scvelo

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# scVelo — RNA Velocity Analysis
2
3	## Overview
4
5	scVelo is the leading Python package for RNA velocity analysis in single-cell RNA-seq data. It infers cell state transitions by modeling the kinetics of mRNA splicing — using the ratio of unspliced (pre-mRNA) to spliced (mature mRNA) abundances to determine whether a gene is being upregulated or downregulated in each cell. This allows reconstruction of developmental trajectories and identification of cell fate decisions without requiring time-course data.
6
7	Installation: `pip install scvelo`
8
9	Key resources:
10	- Documentation: https://scvelo.readthedocs.io/
11	- GitHub: https://github.com/theislab/scvelo
12	- Paper: Bergen et al. (2020) Nature Biotechnology. PMID: 32747759
13
14	## When to Use This Skill
15
16	Use scVelo when:
17
18	- Trajectory inference from snapshot data: Determine which direction cells are differentiating
19	- Cell fate prediction: Identify progenitor cells and their downstream fates
20	- Driver gene identification: Find genes whose dynamics best explain observed trajectories
21	- Developmental biology: Model hematopoiesis, neurogenesis, epithelial-to-mesenchymal transitions
22	- Latent time estimation: Order cells along a pseudotime derived from splicing dynamics
23	- Complement to Scanpy: Add directional information to UMAP embeddings
24
25	## Prerequisites
26
27	scVelo requires count matrices for both unspliced and spliced RNA. These are generated by:
28	1. STARsolo or kallisto\|bustools with `lamanno` mode
29	2. velocyto CLI: `velocyto run10x` / `velocyto run`
30	3. alevin-fry / simpleaf with spliced/unspliced output
31
32	Data is stored in an `AnnData` object with `layers["spliced"]` and `layers["unspliced"]`.
33
34	## Standard RNA Velocity Workflow
35
36	### 1. Setup and Data Loading
37
38	```python
39	import scvelo as scv
40	import scanpy as sc
41	import numpy as np
42	import matplotlib.pyplot as plt
43
44	# Configure settings
45	scv.settings.verbosity = 3 # Show computation steps
46	scv.settings.presenter_view = True
47	scv.settings.set_figure_params('scvelo')
48
49	# Load data (AnnData with spliced/unspliced layers)
50	# Option A: Load from loom (velocyto output)
51	adata = scv.read("cellranger_output.loom", cache=True)
52
53	# Option B: Merge velocyto loom with Scanpy-processed AnnData
54	adata_processed = sc.read_h5ad("processed.h5ad") # Has UMAP, clusters
55	adata_velocity = scv.read("velocyto.loom")
56	adata = scv.utils.merge(adata_processed, adata_velocity)
57
58	# Verify layers
59	print(adata)
60	# obs × var: N × G
61	# layers: 'spliced', 'unspliced' (required)
62	# obsm['X_umap'] (required for visualization)
63	```
64
65	### 2. Preprocessing
66
67	```python
68	# Filter and normalize (follows Scanpy conventions)
69	scv.pp.filter_and_normalize(
70	adata,
71	min_shared_counts=20, # Minimum counts in spliced+unspliced
72	n_top_genes=2000 # Top highly variable genes
73	)
74
75	# Compute first and second order moments (means and variances)
76	# knn_connectivities must be computed first
77	sc.pp.neighbors(adata, n_neighbors=30, n_pcs=30)
78	scv.pp.moments(
79	adata,
80	n_pcs=30,
81	n_neighbors=30
82	)
83	```
84
85	### 3. Velocity Estimation — Stochastic Model
86
87	The stochastic model is fast and suitable for exploratory analysis:
88
89	```python
90	# Stochastic velocity (faster, less accurate)
91	scv.tl.velocity(adata, mode='stochastic')
92	scv.tl.velocity_graph(adata)
93
94	# Visualize
95	scv.pl.velocity_embedding_stream(
96	adata,
97	basis='umap',
98	color='leiden',
99	title="RNA Velocity (Stochastic)"
100	)
101	```
102
103	### 4. Velocity Estimation — Dynamical Model (Recommended)
104
105	The dynamical model fits the full splicing kinetics and is more accurate:
106
107	```python
108	# Recover dynamics (computationally intensive; ~10-30 min for 10K cells)
109	scv.tl.recover_dynamics(adata, n_jobs=4)
110
111	# Compute velocity from dynamical model
112	scv.tl.velocity(adata, mode='dynamical')
113	scv.tl.velocity_graph(adata)
114	```
115
116	### 5. Latent Time
117
118	The dynamical model enables computation of a shared latent time (pseudotime):
119
120	```python
121	# Compute latent time
122	scv.tl.latent_time(adata)
123
124	# Visualize latent time on UMAP
125	scv.pl.scatter(
126	adata,
127	color='latent_time',
128	color_map='gnuplot',
129	size=80,
130	title='Latent time'
131	)
132
133	# Identify top genes ordered by latent time
134	top_genes = adata.var['fit_likelihood'].sort_values(ascending=False).index[:300]
135	scv.pl.heatmap(
136	adata,
137	var_names=top_genes,
138	sortby='latent_time',
139	col_color='leiden',
140	n_convolve=100
141	)
142	```
143
144	### 6. Driver Gene Analysis
145
146	```python
147	# Identify genes with highest velocity fit
148	scv.tl.rank_velocity_genes(adata, groupby='leiden', min_corr=0.3)
149	df = scv.DataFrame(adata.uns['rank_velocity_genes']['names'])
150	print(df.head(10))
151
152	# Speed and coherence
153	scv.tl.velocity_confidence(ad

People who liked this also viewed

6 picks

Type