单细胞样本对齐： 模型训练

在本小节，我们就要开始对齐单细胞了

1. 数据准备

1.1 导入包

import anndata
import itertools
import networkx as nx
import numpy as np
import pandas as pd
import scanpy as sc
import scglue
from matplotlib import rcParams

1.2 导入数据

rna = anndata.read_h5ad("rna_preprocessed.h5ad")
atac = anndata.read_h5ad("atac_preprocessed.h5ad")
graph = nx.read_graphml("prior.graphml.gz")

2 模型训练

2.1 设置模型参数

scglue.plot.set_publication_params()
rcParams["figure.figsize"] = (4, 4)

scglue.models.configure_dataset(
    rna, "NB", use_highly_variable=True,
    use_layer="raw", use_rep="X_pca"
)

scglue.models.configure_dataset(
    atac, "NB", use_highly_variable=True,
    use_rep="X_lsi"
)

graph = graph.subgraph(itertools.chain(
    rna.var.query("highly_variable").index,
    atac.var.query("highly_variable").index
))

2.2 模型训练

glue = scglue.models.SCGLUEModel(
    {"rna": rna, "atac": atac}, sorted(graph.nodes),
    random_seed=0
)

glue.compile()

glue.fit(
    {"rna": rna, "atac": atac},
    graph, edge_weight="weight", edge_sign="sign",
    directory="glue"
)

[INFO] SCGLUEModel: Setting `graph_batch_size` = 28027
[INFO] SCGLUEModel: Setting `align_burnin` = 10
[INFO] SCGLUEModel: Setting `max_epochs` = 38
[INFO] SCGLUEModel: Setting `patience` = 5
[INFO] SCGLUEModel: Setting `reduce_lr_patience` = 3
[INFO] SCGLUETrainer: Using training directory: "glue"
[INFO] SCGLUETrainer: [Epoch 10] train={'g_nll': 0.383, 'g_kl': 0.004, 'g_elbo': 0.387, 'x_rna_nll': 0.246, 'x_rna_kl': 0.009, 'x_rna_elbo': 0.254, 'x_atac_nll': 0.029, 'x_atac_kl': 0.0, 'x_atac_elbo': 0.03, 'dsc_loss': 0.684, 'gen_loss': 0.286}, val={'g_nll': 0.425, 'g_kl': 0.004, 'g_elbo': 0.429, 'x_rna_nll': 0.246, 'x_rna_kl': 0.009, 'x_rna_elbo': 0.254, 'x_atac_nll': 0.029, 'x_atac_kl': 0.0, 'x_atac_elbo': 0.029, 'dsc_loss': 0.692, 'gen_loss': 0.287}, 24.2s elapsed
[INFO] SCGLUETrainer: [Epoch 20] train={'g_nll': 0.338, 'g_kl': 0.004, 'g_elbo': 0.342, 'x_rna_nll': 0.244, 'x_rna_kl': 0.009, 'x_rna_elbo': 0.253, 'x_atac_nll': 0.029, 'x_atac_kl': 0.0, 'x_atac_elbo': 0.029, 'dsc_loss': 0.688, 'gen_loss': 0.282}, val={'g_nll': 0.442, 'g_kl': 0.004, 'g_elbo': 0.447, 'x_rna_nll': 0.245, 'x_rna_kl': 0.008, 'x_rna_elbo': 0.253, 'x_atac_nll': 0.029, 'x_atac_kl': 0.0, 'x_atac_elbo': 0.029, 'dsc_loss': 0.691, 'gen_loss': 0.286}, 24.3s elapsed
Epoch    21: reducing learning rate of group 0 to 2.0000e-04.
Epoch    21: reducing learning rate of group 0 to 2.0000e-04.
2021-10-22 16:24:42,283 ignite.handlers.early_stopping.EarlyStopping INFO: EarlyStopping: Stop training
[INFO] EarlyStopping: Retoring checkpoint "22"...

2.3 保存模型

glue.save("glue/final.dill")