
import scanpy as sc

from continuousvi.trajectory import ContinuousTrajectory, ContinuousVI, TrainedContinuousVI


INPUTS = ["./data/trajectory/GSE123335/adata.h5ad", "./data/trajectory/GSE161690/adata.h5ad"]


adata = sc.concat([sc.read_h5ad(f) for f in INPUTS])


adata.obs


import pandas as pd

adata.obs = pd.read_excel("./data/trajectory/cell_info.xlsx")


adata.obs


TRAGET_CELLTYPES = ["NSC", "IPC", "EX"]
adata_sub = adata[adata.obs["annotation"].isin(TRAGET_CELLTYPES)].copy()

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/anndata/_core/aligned_df.py:68: ImplicitModificationWarning: Transforming to str index.
  warnings.warn("Transforming to str index.", ImplicitModificationWarning)


trained = ContinuousTrajectory(adata_sub).train(batch_key="project_id", label_key="annotation", continuous_key=None, n_train=1)

Epoch 91/800:  11%|█▏        | 90/800 [14:46<1:58:18, 10.00s/it, v_num=1, train_loss_step=4.76e+3, train_loss_epoch=4.92e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")

Epoch 92/800:  11%|█▏        | 91/800 [14:56<1:57:48,  9.97s/it, v_num=1, train_loss_step=4.6e+3, train_loss_epoch=4.92e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")

Epoch 93/800:  12%|█▏        | 92/800 [15:06<1:57:36,  9.97s/it, v_num=1, train_loss_step=4.79e+3, train_loss_epoch=4.92e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")

Epoch 94/800:  12%|█▏        | 93/800 [15:16<1:57:53, 10.01s/it, v_num=1, train_loss_step=4.71e+3, train_loss_epoch=4.92e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")

Epoch 95/800:  12%|█▏        | 94/800 [15:26<1:57:57, 10.02s/it, v_num=1, train_loss_step=5.05e+3, train_loss_epoch=4.92e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")


trained.trainedVI.save("./dist/trajectory/models", overwrite=True)

Saving trained model.: 100%|██████████| 1/1 [00:00<00:00,  1.04it/s]

<continuousvi.continuousVI.TrainedContinuousVI at 0x7fc1f438a0e0>


trained = ContinuousTrajectory(adata_sub)


trained.trainedVI = TrainedContinuousVI(adata=adata_sub, batch_key="project_id", label_key="annotation", continuous_key=None, trained_model_path="./dist/trajectory/models")

Loading pre-trained models:   0%|          | 0/1 [00:00<?, ?it/s]

INFO     File dist/trajectory/models/model_0/model.pt already downloaded

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/torch/cuda/__init__.py:734: UserWarning: Can't initialize NVML
  warnings.warn("Can't initialize NVML")
Loading pre-trained models: 100%|██████████| 1/1 [00:00<00:00,  5.50it/s]


trained.trainedVI.calc_embeddings(resolution=0.3)

/mnt/work3/yuyasato/libs/ContinuousVI/src/continuousvi/continuousVI.py:165: FutureWarning: In the future, the default backend for leiden will be igraph instead of leidenalg.

 To achieve the future defaults please pass: flavor="igraph" and n_iterations=2.  directed must also be False to work with igraph's implementation.
  sc.tl.leiden(self.adata, key_added=KEY_CLUSTER, resolution=resolution, directed=False)

<continuousvi.continuousVI.TrainedContinuousVI at 0x724c7d7cf190>


trained.trainedVI.embeddings.umap(color_by=["Sox2", "annotation", "clusters"])

Sampling expression: 100%|██████████| 1/1 [00:24<00:00, 24.55s/it]

<continuousvi.continuousVI.TrainedContinuousVI.Plots at 0x724c7d0f4ca0>


lineages: dict[str, list[int]] = {
    "NSC --> IPC": [1, 3],
    "IPC --> EX": [0, 2],
}
trained.adata = trained.trainedVI.adata.copy()


trained = trained.define_lineages(lineages)


trained = trained.calculate_pseudotime()

Calculating the pseudotime.:   0%|          | 0/2 [00:00<?, ?it/s]

WARNING: Trying to run `tl.dpt` without prior call of `tl.diffmap`. Falling back to `tl.diffmap` with default parameters.

/mnt/work3/yuyasato/libs/ContinuousVI/src/continuousvi/trajectory.py:170: ImplicitModificationWarning: Trying to modify attribute `._uns` of view, initializing view as actual.
  _adata_sub.uns["iroot"] = ContinuousTrajectory._calc_iroot_idx(
Calculating the pseudotime.:  50%|█████     | 1/2 [00:00<00:00,  6.98it/s]/mnt/work3/yuyasato/libs/ContinuousVI/src/continuousvi/trajectory.py:170: ImplicitModificationWarning: Trying to modify attribute `._uns` of view, initializing view as actual.
  _adata_sub.uns["iroot"] = ContinuousTrajectory._calc_iroot_idx(

WARNING: Trying to run `tl.dpt` without prior call of `tl.diffmap`. Falling back to `tl.diffmap` with default parameters.

Calculating the pseudotime.: 100%|██████████| 2/2 [00:00<00:00,  7.43it/s]
Calculating the pseudotime.: 100%|██████████| 2/2 [00:00<00:00,  7.43it/s]


import warnings

warnings.filterwarnings("ignore", message="Can't initialize NVML")

regression_result = trained.regression_lineage(batch_key="project_id", label_key="annotation", n_train=5)

Calculating the regression.:   0%|          | 0/2 [00:00<?, ?it/s]/mnt/work3/yuyasato/libs/ContinuousVI/src/continuousvi/trajectory.py:221: ImplicitModificationWarning: Trying to modify attribute `.obs` of view, initializing view as actual.
  _adata_sub.obs["pseudotime"] = self.pseudotimes[name]["pseudotime"]
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/scvi/data/fields/_dataframe_field.py:187: UserWarning: Category 0 in adata.obs['_scvi_labels'] has fewer than 3 cells. Models may not train properly.
  categorical_mapping = _make_column_categorical(
GPU available: False, used: False
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.

Epoch 130/800:  16%|█▋        | 130/800 [10:46<55:33,  4.98s/it, v_num=1, train_loss_step=4.36e+3, train_loss_epoch=4.85e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/scvi/data/fields/_dataframe_field.py:187: UserWarning: Category 0 in adata.obs['_scvi_labels'] has fewer than 3 cells. Models may not train properly.
  categorical_mapping = _make_column_categorical(
GPU available: False, used: False
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.

Monitored metric elbo_validation did not improve in the last 45 records. Best score: 5128.705. Signaling Trainer to stop.
Epoch 142/800:  18%|█▊        | 142/800 [11:29<53:15,  4.86s/it, v_num=1, train_loss_step=5.16e+3, train_loss_epoch=4.84e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/scvi/data/fields/_dataframe_field.py:187: UserWarning: Category 0 in adata.obs['_scvi_labels'] has fewer than 3 cells. Models may not train properly.
  categorical_mapping = _make_column_categorical(
GPU available: False, used: False
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.

Monitored metric elbo_validation did not improve in the last 45 records. Best score: 5074.546. Signaling Trainer to stop.
Epoch 142/800:  18%|█▊        | 142/800 [11:44<54:23,  4.96s/it, v_num=1, train_loss_step=5.02e+3, train_loss_epoch=4.84e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/scvi/data/fields/_dataframe_field.py:187: UserWarning: Category 0 in adata.obs['_scvi_labels'] has fewer than 3 cells. Models may not train properly.
  categorical_mapping = _make_column_categorical(
GPU available: False, used: False
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.

Monitored metric elbo_validation did not improve in the last 45 records. Best score: 5057.078. Signaling Trainer to stop.
Epoch 124/800:  16%|█▌        | 124/800 [10:12<55:36,  4.94s/it, v_num=1, train_loss_step=4.91e+3, train_loss_epoch=4.86e+3]

/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/scvi/data/fields/_dataframe_field.py:187: UserWarning: Category 0 in adata.obs['_scvi_labels'] has fewer than 3 cells. Models may not train properly.
  categorical_mapping = _make_column_categorical(
GPU available: False, used: False
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.
/home/yuyasato/work3/libs/ContinuousVI/.venv/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=23` in the `DataLoader` to improve performance.

Monitored metric elbo_validation did not improve in the last 45 records. Best score: 5048.759. Signaling Trainer to stop.
Epoch 134/800:  17%|█▋        | 134/800 [10:58<54:35,  4.92s/it, v_num=1, train_loss_step=4.23e+3, train_loss_epoch=4.83e+3]


for name, reg in regression_result.items():
    reg.to_csv(f"./dist/trajectory/regression_{name}.tsv", sep="\t")


# Up
#  Eomes: IPC's marker (https://arc.net/l/quote/uznkomhu)
# Down
#  Nes: Neural stem cell's marker (https://arc.net/l/quote/genbnnhm)
target_lineage = list(trained.defined_lineages.keys())[0]
print(f"@Lineage: {target_lineage}")
regression_result[target_lineage]

@Lineage: NSC --> IPC


# Up:
#  Stmn2: Excitatory marker (https://arc.net/l/quote/vdbdmvqc)
# Down:
#  Neurod1: Progenitor marker (https://arc.net/l/quote/bkxwdxco)

target_lineage = list(trained.defined_lineages.keys())[1]
print(f"@Lineage: {target_lineage}")
regression_result[target_lineage]

@Lineage: IPC --> EX


trained = ContinuousTrajectory(adata_sub)
trained.trainedVI = TrainedContinuousVI(adata=adata_sub, batch_key="project_id", label_key="annotation", continuous_key=None, trained_model_path="./dist/trajectory/models")
trained.trainedVI.calc_embeddings(resolution=0.3)

INFO     File dist/trajectory/models/model_0/model.pt already downloaded

Loading pre-trained models: 100%|██████████| 1/1 [00:00<00:00,  8.56it/s]

<continuousvi.continuousVI.TrainedContinuousVI at 0x724c7a3964a0>


trained.trainedVI.embeddings.umap(color_by=["Nes", "Eomes", "annotation"])

Sampling expression: 100%|██████████| 1/1 [00:27<00:00, 27.68s/it]

<continuousvi.continuousVI.TrainedContinuousVI.Plots at 0x724c7963d600>


trained.trainedVI.embeddings.umap(color_by=["Neurod1", "Stmn2", "annotation"])

Sampling expression: 100%|██████████| 1/1 [00:28<00:00, 28.42s/it]

<continuousvi.continuousVI.TrainedContinuousVI.Plots at 0x724c799895a0>


sc.pl.draw_graph(trained.adata, color=["annotation"], legend_loc="on data")


sc.pl.paga(trained.adata, color=["annotation"])


adata_sub

AnnData object with n_obs × n_vars = 29018 × 19124
    obs: 'index', 'project_id', 'time', 'converted_time', '_scvi_batch', '_scvi_labels', 'leiden', 'annotation'


sc.pp.pca(adata_sub, n_comps=30)
sc.pp.neighbors(adata_sub)
sc.tl.umap(adata_sub)


sc.pl.umap(adata_sub, color=["project_id", "annotation"])

	project_id	time
p0-WT2_AAAACATGAACC	GSE123335	P0
p0-WT2_AAAAGACACCAC	GSE123335	P0
p0-WT2_AAAAGGCTCGTG	GSE123335	P0
p0-WT2_AAAATACGCAAG	GSE123335	P0
p0-WT2_AAACACCGGATA	GSE123335	P0
...	...	...
E18_2_CGCGTTAGAGGC	GSE161690	E18
E18_2_ACCCTGATCTGG	GSE161690	E18
E18_2_TCGCCTGTGGAC	GSE161690	E18
E18_2_GAGCTTCTCCGC	GSE161690	E18
E18_2_ATCGTGGCGTTC	GSE161690	E18

	index	project_id	time	converted_time	_scvi_batch	_scvi_labels	leiden	annotation
0	p0-WT2_AAAACATGAACC	GSE123335	P0	19	0	0	15	OPC
1	p0-WT2_AAAAGACACCAC	GSE123335	P0	19	0	0	11	Other
2	p0-WT2_AAAAGGCTCGTG	GSE123335	P0	19	0	0	16	EC
3	p0-WT2_AAAATACGCAAG	GSE123335	P0	19	0	0	4	IN
4	p0-WT2_AAACACCGGATA	GSE123335	P0	19	0	0	4	IN
...	...	...	...	...	...	...	...	...
29013	E18_2_CGCGTTAGAGGC	GSE161690	E18	18	1	0	2	Other
29014	E18_2_ACCCTGATCTGG	GSE161690	E18	18	1	0	9	oRG
29015	E18_2_TCGCCTGTGGAC	GSE161690	E18	18	1	0	2	Other
29016	E18_2_GAGCTTCTCCGC	GSE161690	E18	18	1	0	7	Other
29017	E18_2_ATCGTGGCGTTC	GSE161690	E18	18	1	0	2	Other

	gene	slope	intercept	r2
9774	Igfbpl1	3.919986	0.042765	0.576759
5396	Eomes	3.663179	-0.265159	0.691716
16120	Sox4	2.803581	0.561999	0.494562
12101	Neurod6	2.788903	-0.131581	0.585619
6286	Gadd45g	2.788765	0.007986	0.438869
...	...	...	...	...
15330	Sfrp1	-1.881385	1.432907	0.180729
1714	Aldoc	-1.922470	1.132011	0.226049
17751	Tuba1b	-1.939113	1.377589	0.182468
13951	Ptn	-2.352473	1.612374	0.231150
5601	Fabp7	-3.261634	2.394049	0.279647

	gene	slope	intercept	r2
11194	Meg3	4.257950	0.155017	0.575140
11032	Mapt	4.246047	0.186789	0.460254
16434	Stmn2	3.787384	0.823437	0.267230
9872	Ina	3.707868	0.180680	0.400880
17758	Tubb2a	3.465068	0.190151	0.423433
...	...	...	...	...
9774	Igfbpl1	-1.278309	1.897858	0.074115
12098	Neurod1	-1.284228	0.803582	0.109375
14709	Robo2	-1.461415	0.719719	0.256030
14612	Rnd2	-1.474085	0.735223	0.264990
12379	Nrp1	-1.560335	0.818908	0.158502

Additional data¶

Raw data (2 data)¶