194_单细胞谱系追踪¶

# CRISPR条码数据分析（使用Cassiopeia）
# pip install cassiopeia-lineage
import cassiopeia as cas
import pandas as pd
import numpy as np

# 读取条码数据
# 格式：每行一个细胞，每列一个条码位点，值为等位基因状态
# 0=未编辑, 1,2,3...=不同编辑结果, -1=dropout
barcode_matrix = pd.read_csv("barcode_allele_table.csv", index_col=0)

# 创建CassiopeiaTree对象
tree = cas.data.CassiopeiaTree(
    character_matrix=barcode_matrix,
    cell_meta=cell_metadata  # 可选的细胞注释
)

# 使用Cassiopeia-Greedy重建谱系树
greedy_solver = cas.solver.VanillaGreedySolver()
greedy_solver.solve(tree)

# 或使用更精确的ILP求解器
# ilp_solver = cas.solver.ILPSolver()
# ilp_solver.solve(tree)

# 或使用最大似然方法
# ml_solver = cas.solver.MaximumLikelihoodSolver()
# ml_solver.solve(tree)

# 可视化谱系树
cas.pl.plot_matplotlib(tree, 
                        meta_data=["cell_type"],
                        clade_colors={"cell_type": color_map})

# 基于体细胞SNV的谱系追踪
import numpy as np
from scipy.cluster.hierarchy import linkage, dendrogram
from scipy.spatial.distance import pdist

# 突变矩阵：行=细胞，列=突变位点，值=0/1
mutation_matrix = np.array([
    [1, 1, 0, 0, 0, 1],  # Cell1
    [1, 1, 0, 0, 1, 0],  # Cell2
    [1, 1, 1, 0, 0, 0],  # Cell3
    [0, 0, 0, 1, 0, 0],  # Cell4
    [0, 0, 0, 1, 1, 0],  # Cell5
])
cell_names = [f"Cell{i}" for i in range(1, 6)]

# 基于Hamming距离的层次聚类
dist = pdist(mutation_matrix, metric='hamming')
Z = linkage(dist, method='average')

import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(8, 5))
dendrogram(Z, labels=cell_names, ax=ax)
ax.set_title("Cell Lineage Tree (Somatic Mutations)")
ax.set_ylabel("Distance")
plt.tight_layout()
plt.savefig("lineage_tree.png", dpi=300)

# 使用mgatk + Signac进行线粒体谱系分析
library(Signac)
library(Seurat)

# 加载mgatk结果
mgatk <- ReadMGATK("mgatk_output/final/")
seurat_obj[["mito"]] <- CreateAssayObject(counts = mgatk$counts)

# 识别信息性变异
variants <- IdentifyVariants(seurat_obj, assay = "mito", refallele = mgatk$refallele)
informative <- subset(variants, n_cells_conf_detected >= 5 & strand_correlation >= 0.65)

# 构建谱系树
library(ape)
vaf_matrix <- GetAssayData(seurat_obj, assay = "mito")[informative$variant, ]
dist_mat <- dist(t(as.matrix(vaf_matrix)))
tree <- nj(dist_mat)

# 将谱系与细胞类型关联
plot(tree, type = "fan", show.tip.label = FALSE)
tiplabels(pch = 19, col = cell_type_colors[seurat_obj$cell_type], cex = 0.5)
legend("topleft", legend = names(cell_type_colors), col = cell_type_colors, pch = 19)

# 使用LineageOT进行谱系感知的最优传输
# pip install lineageot
import lineageot as lot

# 构建时间点之间的细胞对应关系
# 输入：多个时间点的scRNA-seq数据 + 谱系树
coupling = lot.fit_lineage_coupling(
    adata_early,      # 早期时间点
    adata_late,       # 晚期时间点
    lineage_tree,     # 谱系树
    alpha=0.5         # 谱系信息权重
)

# 使用CoSpar进行克隆动态推断
# pip install cospar
import cospar as cs

# 准备数据
adata_orig = cs.pp.initialize_adata_object(adata)
cs.tl.infer_Tmap(adata_orig, smooth_array=[15, 10, 5],
                  method='OT', OT_epsilon=0.02)
cs.pl.fate_map(adata_orig, selected_fates=['CellType_A', 'CellType_B'])

# Cassiopeia谱系重建
import cassiopeia as cas
tree = cas.data.CassiopeiaTree(character_matrix=barcodes)
cas.solver.VanillaGreedySolver().solve(tree)

# R中构建谱系树
library(ape)
tree <- nj(dist(t(mutation_matrix)))
plot(tree, type = "fan")