JuloVelo

A local RNA velocity model for cell differentiation inference based on Julia programming language. This project is currently WIP and highly inspired by cellDancer.

Requirement

• Julia >= 1.10

• Python >= 3.7

• scvelo

Optional

• cellDancer == 1.1.7

The minimum example of using JuloVelo

Preprocessing

import scvelo as scv
import scanpy as sc
import numpy as np

# Load anndata from scvelo
adata = scv.datasets.gastrulation_erythroid()

# Preprocessing
scv.pp.filter_and_normalize(adata, min_shared_counts = 20, n_top_genes = 2000)
sc.pp.pca(adata)
sc.pp.neighbors(adata, n_pcs = 50, n_neighbors = 50)
scv.pp.moments(adata, n_pcs = None, n_neighbors = None)

# Clustering
sc.tl.leiden(adata)
sc.pl.umap(adata, color = "leiden")

# Calculate diffusion pseudotime
adata.uns["iroot"] = np.flatnonzero(adata.obs["leiden"] == '6')[0]
adata.uns["root"] = 6 #Base on root leiden cluster
sc.tl.dpt(adata)

# Avoid PooledArray bug in Julia
adata.var["highly_variable_genes"] = ["true" if item == "True" else "false" for item in adata.var["highly_variable_genes"]]

# Write anndata for JuloVelo
adata.write_h5ad("JuloVelo_pre.h5ad")

Training model

using JuloVelo
using Plots

# Load anndata and change iroot to root cluster
adata = read_adata("JuloVelo_pre.h5ad")

# Normalize read count
normalize(adata)

# Filter genes and pre-determine gene kinetics
filter_and_gene_kinetics_predetermination(adata)

# Reshape data for model compatible
reshape_data(adata)

# Reduce sample size
density_sampling(adata)

# Training
Kinetics = train(adata)

# Calculate velocity
velocity_estimation(adata, Kinetics)

# Compute cell velocity on embedding space
compute_cell_velocity(adata)

# Restore pseudotime
estimate_pseudotime(adata)

# Write anndata
write_adata(adata)