Ctrl+K

Note

This page was generated from combosciplex_Rdkit_embeddings.ipynb. Interactive online version: Colab badge.

Predicting combinatorial drug perturbations using RDKit embeddings for drugs#

While CPA can predict unseen combinations using its default setting, it can not predict completely unseen perturbations. We have extended CPA to work with external perturbation embeddings (chemCPA. Ultimately, these embeddings should capture the similarities and differences between perturbations. They can be obtained from any resources (e.g., generative models encoding drugs or Chemo-physical properties of molecules). Here, we showcase how to leverage an external embedding from RDKit, which returns a vector representing the Chemo-physical features of molecules. This will allow us to predict perturbation completely absent in the training data, conditional on similar molecules in training data; otherwise, predictions will not be reliable. Here, we train CPA on the Combo Sci-Plex data as we did here using RDKit embeddings extracted with RDKit python package. The model’s performance was finally evaluated on the same held-out drugs (OOD drugs) as the previous tutorial. The rationale is the same if you would like to include other perturbation embedding as shown in the chemCPA paper.

[1]:

import sys
#if branch is stable, will install via pypi, else will install from source
branch = "latest"
IN_COLAB = "google.colab" in sys.modules

if IN_COLAB and branch == "stable":
    !pip install cpa-tools
    !pip install scanpy
elif IN_COLAB and branch != "stable":
    !pip install --quiet --upgrade jsonschema
    !pip install git+https://github.com/theislab/cpa
    !pip install scanpy

[2]:

from sklearn.metrics import r2_score
import numpy as np

import os
# os.chdir('/home/mohsen/projects/cpa/') # For local run
# os.environ['CUDA_VISIBLE_DEVICES'] = '0' # For dedicated GPU usage

Import CPA package

[3]:

import cpa
import scanpy as sc

Global seed set to 0

[4]:

sc.settings.set_figure_params(dpi=100)

[5]:

data_path = '/home/mohsen/projects/cpa/datasets/combo_sciplex_prep_hvg_filtered.h5ad'

Data Loading#

Combo sciplex dataset can be loaded using scanpy’s read function. The dataset is already preprocessed and stored in the data folder. The following columns in adata.obs are required for training the model:

  • condition_ID: Single/Combinatorial CHEMBL IDs of the drugs applied to the cells

  • log_dose: Log-transformed dosages of the drugs applied to the cells

  • cell_type: Cell types of the cells (All A549 in this case)

  • smiles_rkdit: Canonical SMILES representation of the drugs separated by “..” for combinatorial perturbations

  • cov_drug_dose: Covariates+Drug+Dosage information of the drugs applied to the cells (used for accessing top DEGs from adata.uns[‘rank_genes_groups_cov’])

  • split_1ct_MEC: Train/valid/ood split information of the cells

[6]:

try:
    adata = sc.read(data_path)
except:
    import gdown
    gdown.download('https://drive.google.com/uc?export=download&id=1RRV0_qYKGTvD3oCklKfoZQFYqKJy4l6t')
    data_path = 'combo_sciplex_prep_hvg_filtered.h5ad'
    adata = sc.read(data_path)

adata

[6]:

AnnData object with n_obs × n_vars = 63378 × 5000
    obs: 'ncounts', 'well', 'plate', 'cell_line', 'replicate', 'time', 'dose_value', 'pathway1', 'pathway2', 'perturbation', 'target', 'pathway', 'dose_unit', 'celltype', 'disease', 'tissue_type', 'organism', 'perturbation_type', 'ngenes', 'percent_mito', 'percent_ribo', 'nperts', 'chembl-ID', 'condition', 'condition_ID', 'control', 'cell_type', 'smiles_rdkit', 'source', 'sample', 'Size_Factor', 'n.umi', 'RT_well', 'Drug1', 'Drug2', 'Well', 'n_genes', 'n_genes_by_counts', 'total_counts', 'total_counts_mt', 'pct_counts_mt', 'leiden', 'split', 'condition_old', 'pert_type', 'batch', 'split_1ct_MEC', 'split_2ct_MEC', 'split_3ct_MEC', 'batch_cov', 'batch_cov_cond', 'log_dose', 'cov_drug_dose'
    var: 'ensembl_id-0', 'ncounts-0', 'ncells-0', 'symbol-0', 'symbol-1', 'id-1', 'n_cells-1', 'mt-1', 'n_cells_by_counts-1', 'mean_counts-1', 'pct_dropout_by_counts-1', 'total_counts-1', 'highly_variable-1', 'means-1', 'dispersions-1', 'dispersions_norm-1', 'highly_variable', 'means', 'dispersions', 'dispersions_norm', 'highly_variable_nbatches', 'highly_variable_intersection'
    uns: 'cell_type_colors', 'hvg', 'neighbors', 'pathway1_colors', 'pca', 'rank_genes_groups_cov', 'rank_genes_groups_cov_detailed', 'source_colors', 'umap'
    obsm: 'X_pca', 'X_umap'
    varm: 'PCs'
    layers: 'counts'
    obsp: 'connectivities', 'distances'

Data preparation#

IMPORTANT: Currenlty because of the standartized evaluation procedure, we need to provide adata.obs[‘control’] (0 if not control, 1 for cells to use as control). And we also need to provide de_genes in .uns[‘rank_genes_groups’].

[7]:

adata.obs['split_1ct_MEC'].value_counts()

[7]:

train    49683
ood       8209
valid     5486
Name: split_1ct_MEC, dtype: int64

In the following, we will extract OOD drugs and visualize cells perturbed by them.

[10]:

ood_conds = list(adata[adata.obs['split_1ct_MEC'] == 'ood'].obs['condition_ID'].value_counts().index)
ood_conds

[10]:

['CHEMBL1213492+CHEMBL491473',
 'CHEMBL483254+CHEMBL4297436',
 'CHEMBL356066+CHEMBL402548',
 'CHEMBL483254+CHEMBL383824',
 'CHEMBL4297436+CHEMBL383824']

[11]:

adata.obs['condition_split'] = adata.obs['condition_ID'].apply(lambda x: x if x in ood_conds else 'other')

[12]:

sc.settings.verbosity = 3

[13]:

sc.pp.neighbors(adata)
sc.tl.umap(adata)

computing neighbors
    using 'X_pca' with n_pcs = 50
    finished: added to `.uns['neighbors']`
    `.obsp['distances']`, distances for each pair of neighbors
    `.obsp['connectivities']`, weighted adjacency matrix (0:00:28)
computing UMAP
    finished: added
    'X_umap', UMAP coordinates (adata.obsm) (0:00:30)

[15]:

sc.settings.set_figure_params(dpi=100)

[24]:

sc.pl.umap(adata,
           color='condition_split',
           groups=ood_conds,
           palette=sc.pl.palettes.godsnot_102,
           na_in_legend=False,
           na_color='grey',
           frameon=False)
../_images/tutorials_combosciplex_Rdkit_embeddings_20_0.png 
[25]:

adata.X = adata.layers['counts'].copy()

Data setup#

Data setup is the first step required for training CPA. Just like scvi-tools models, you can call cpa.CPA.setup_anndata to preprocess your data. This function will accept the following arguments:

  • adata: AnnData object containing the data to be preprocessed

  • perturbation_key: The key in adata.obs that contains the perturbation information

  • control_group: The name of the control group in perturbation_key

  • batch_key: The key in adata.obs that contains the batch information

  • dosage_key: The key in adata.obs that contains the dosage information

  • categorical_covariate_keys: A list of keys in adata.obs that contain categorical covariates

  • is_count_data: Whether the adata.X is count data or not

  • deg_uns_key: The key in adata.uns that contains the differential expression results

  • deg_uns_cat_key: The key in adata.obs that contains the category information of each cell which can be used as to access differential expression results in adata.uns[deg_uns_key]. For example, if deg_uns_key is rank_genes_groups_cov and deg_uns_cat_key is cov_cond, then adata.uns[deg_uns_key][cov_cond] will contain the differential expression results for each category in cov_cond.

  • max_comb_len: The maximum number of perturbations that are applied to each cell. For example, if max_comb_len is 2, then the model will be trained to predict the effect of single perturbations and the effect of double perturbations.

[9]:

cpa.CPA.setup_anndata(adata,
                      perturbation_key='condition_ID',
                      dosage_key='log_dose',
                      control_group='CHEMBL504',
                      batch_key=None,
                      smiles_key='smiles_rdkit',
                      is_count_data=True,
                      categorical_covariate_keys=['cell_type'],
                      deg_uns_key='rank_genes_groups_cov',
                      deg_uns_cat_key='cov_drug_dose',
                      max_comb_len=2,
                     )

100%|████████████████████████████████████████████████████████████████████████████████████████| 63378/63378 [00:01<00:00, 51900.44it/s]
100%|███████████████████████████████████████████████████████████████████████████████████████| 63378/63378 [00:00<00:00, 848088.64it/s]
100%|███████████████████████████████████████████████████████████████████████████████████████████████| 32/32 [00:00<00:00, 1246.29it/s]
No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)

INFO     Generating sequential column names
INFO     Generating sequential column names
INFO     Generating sequential column names
INFO     Generating sequential column names

Training CPA#

You can specify all the parameters for the model in a dictionary of parameters. If they are not specified, default values will be selected.

  • ae_hparams are technical parameters of the architecture of the autoencoder.

    • n_latent: number of latent dimensions for the autoencoder

    • recon_loss: the type of reconstruction loss function to use

    • doser_type: the type of doser to use

    • n_hidden_encoder: number of hidden neurons in each hidden layer of the encoder

    • n_layers_encoder: number of hidden layers in the encoder

    • n_hidden_decoder: number of hidden neurons in each hidden layer of the decoder

    • n_layers_decoder: number of hidden layers in the decoder

    • use_batch_norm_encoder: if True, batch normalization will be used in the encoder

    • use_layer_norm_encoder: if True, layer normalization will be used in the encoder

    • use_batch_norm_decoder: if True, batch normalization will be used in the decoder

    • use_layer_norm_decoder: if True, layer normalization will be used in the decoder

    • dropout_rate_encoder: dropout rate used in the encoder

    • dropout_rate_decoder: dropout rate used in the decoder

    • variational: if True, variational autoencoder will be employed as the main perturbation response predictor

    • seed: number for setting the seed for generating random numbers.

  • trainer_params are training parameters of CPA.

    • n_epochs_adv_warmup: number of epochs for adversarial warmup

    • n_epochs_kl_warmup: number of epochs for KL divergence warmup

    • n_epochs_pretrain_ae: number of epochs to pre-train the autoencoder

    • adv_steps: number of steps used to train adversarial classifiers after a single step of training the autoencoder

    • mixup_alpha: mixup interpolation coefficient

    • n_epochs_mixup_warmup: number of epochs for mixup warmup

    • lr: learning rate of the trainer

    • wd: weight decay of the trainer

    • doser_lr: learning rate of doser parameters

    • doser_wd: weight decay of doser parameters

    • adv_lr: learning rate of adversarial classifiers

    • adv_wd: weight decay rate of adversarial classifiers

    • pen_adv: penalty for adversarial classifiers

    • reg_adv: regularization for adversarial classifiers

    • n_layers_adv: number of hidden layers in adversarial classifiers

    • n_hidden_adv: number of hidden neurons in each hidden layer of adversarial classifiers

    • use_batch_norm_adv: if True, batch normalization will be used in the adversarial classifiers

    • use_layer_norm_adv: if True, layer normalization will be used in the adversarial classifiers

    • dropout_rate_adv: dropout rate used in the adversarial classifiers

    • step_size_lr: learning rate step size

    • do_clip_grad: if True, gradient clipping will be used

    • adv_loss: the type of loss function to use for adversarial training

    • gradient_clip_value: value to clip gradients to, if do_clip_grad is True

[10]:

ae_hparams = {'n_latent': 64,
 'recon_loss': 'nb',
 'doser_type': 'linear',
 'n_hidden_encoder': 256,
 'n_layers_encoder': 3,
 'n_hidden_decoder': 512,
 'n_layers_decoder': 2,
 'use_batch_norm_encoder': True,
 'use_layer_norm_encoder': False,
 'use_batch_norm_decoder': True,
 'use_layer_norm_decoder': False,
 'dropout_rate_encoder': 0.25,
 'dropout_rate_decoder': 0.25,
 'variational': False,
 'seed': 6478}

trainer_params = {'n_epochs_kl_warmup': None,
 'n_epochs_pretrain_ae': 50,
 'n_epochs_adv_warmup': 100,
 'n_epochs_mixup_warmup': 10,
 'mixup_alpha': 0.1,
 'adv_steps': None,
 'n_hidden_adv': 128,
 'n_layers_adv': 3,
 'use_batch_norm_adv': False,
 'use_layer_norm_adv': False,
 'dropout_rate_adv': 0.2,
 'reg_adv': 10.0,
 'pen_adv': 0.1,
 'lr': 0.0003,
 'wd': 4e-07,
 'adv_lr': 0.0003,
 'adv_wd': 4e-07,
 'adv_loss': 'cce',
 'doser_lr': 0.0003,
 'doser_wd': 4e-07,
 'do_clip_grad': False,
 'gradient_clip_value': 1.0,
 'step_size_lr': 10}

Model instantiation#

NOTE: Run the following 3 cells if you haven’t already trained CPA from scratch.

Here, we create a CPA model using cpa.CPA given all hyper-parameters.

[11]:

adata.obs['split_1ct_MEC'].value_counts()

[11]:

train    49683
ood       8209
valid     5486
Name: split_1ct_MEC, dtype: int64

[12]:

model = cpa.CPA(adata=adata,
                split_key='split_1ct_MEC',
                train_split='train',
                valid_split='valid',
                test_split='ood',
                use_rdkit_embeddings=True,
                **ae_hparams,
               )

Global seed set to 6478

(18, 2048)

Training CPA#

After creating a CPA object, we train the model with the following arguments: * max_epochs: Maximum number of epochs to train the models. * use_gpu: If True, will use the available GPU to train the model. * batch_size: Number of samples to use in each mini-batches. * early_stopping_patience: Number of epochs with no improvement in early stopping callback. * check_val_every_n_epoch: Interval of checking validation losses. * save_path: Path to save the model after the training has finished.

[13]:

model.train(max_epochs=2000,
            use_gpu=True,
            batch_size=512,
            plan_kwargs=trainer_params,
            early_stopping_patience=10,
            check_val_every_n_epoch=5,
            save_path='/home/mohsen/projects/cpa/lightning_logs/combo_rdkit/',
           )

100%|█████████████████████████████████████████████████████████████████████████████████████████████████| 32/32 [00:00<00:00, 54.40it/s]
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs
You are using a CUDA device ('NVIDIA RTX A6000') that has Tensor Cores. To properly utilize them, you should set `torch.set_float32_matmul_precision('medium' | 'high')` which will trade-off precision for performance. For more details, read https://pytorch.org/docs/stable/generated/torch.set_float32_matmul_precision.html#torch.set_float32_matmul_precision
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]

Epoch 5/2000:   0%|         | 4/2000 [01:10<11:14:26, 20.27s/it, v_num=1, recon=1.33e+3, r2_mean=0.904, adv_loss=2.25, acc_pert=0.171]


Epoch 00004: cpa_metric reached. Module best state updated.

Epoch 10/2000:   0%| | 9/2000 [03:10<11:32:22, 20.87s/it, v_num=1, recon=1.32e+3, r2_mean=0.912, adv_loss=2.12, acc_pert=0.192, val_re


Epoch 00009: cpa_metric reached. Module best state updated.


disnt_basal = 0.10935998205906987
disnt_after = 0.1947580928900672
val_r2_mean = 0.9087304059503858
val_r2_var = 0.5239273883678295
Epoch 15/2000:   1%| | 14/2000 [03:49<5:23:20,  9.77s/it, v_num=1, recon=1.31e+3, r2_mean=0.912, adv_loss=2.08, acc_pert=0.217, val_re


Epoch 00014: cpa_metric reached. Module best state updated.

Epoch 20/2000:   1%| | 19/2000 [04:29<4:31:19,  8.22s/it, v_num=1, recon=1.3e+3, r2_mean=0.915, adv_loss=2.07, acc_pert=0.228, val_rec


Epoch 00019: cpa_metric reached. Module best state updated.


disnt_basal = 0.09836006051939437
disnt_after = 0.20331588620822566
val_r2_mean = 0.9018172625740771
val_r2_var = 0.5401924108013962
Epoch 25/2000:   1%| | 24/2000 [05:10<4:23:27,  8.00s/it, v_num=1, recon=1.29e+3, r2_mean=0.915, adv_loss=2.1, acc_pert=0.222, val_rec


Epoch 00024: cpa_metric reached. Module best state updated.

Epoch 30/2000:   1%| | 29/2000 [05:47<4:02:34,  7.38s/it, v_num=1, recon=1.29e+3, r2_mean=0.916, adv_loss=2.14, acc_pert=0.218, val_re


Epoch 00029: cpa_metric reached. Module best state updated.


disnt_basal = 0.08712921352109765
disnt_after = 0.2052801964435556
val_r2_mean = 0.9123962264671068
val_r2_var = 0.5435857327297481
Epoch 35/2000:   2%| | 34/2000 [06:27<4:16:16,  7.82s/it, v_num=1, recon=1.29e+3, r2_mean=0.915, adv_loss=2.19, acc_pert=0.215, val_re


Epoch 00034: cpa_metric reached. Module best state updated.

Epoch 40/2000:   2%| | 39/2000 [07:05<4:03:35,  7.45s/it, v_num=1, recon=1.28e+3, r2_mean=0.917, adv_loss=2.22, acc_pert=0.212, val_re
disnt_basal = 0.07843184176780146
disnt_after = 0.20770758128973216
val_r2_mean = 0.9067388287297002
val_r2_var = 0.5569063208319924
Epoch 45/2000:   2%| | 44/2000 [07:46<4:21:04,  8.01s/it, v_num=1, recon=1.28e+3, r2_mean=0.916, adv_loss=2.26, acc_pert=0.205, val_re


Epoch 00044: cpa_metric reached. Module best state updated.

Epoch 50/2000:   2%| | 49/2000 [08:25<4:04:41,  7.53s/it, v_num=1, recon=1.28e+3, r2_mean=0.917, adv_loss=2.31, acc_pert=0.198, val_re


Epoch 00049: cpa_metric reached. Module best state updated.


disnt_basal = 0.07375605171504182
disnt_after = 0.22323710480713033
val_r2_mean = 0.9130540171054878
val_r2_var = 0.5530647197957793
Epoch 55/2000:   3%| | 54/2000 [09:05<4:19:32,  8.00s/it, v_num=1, recon=1.27e+3, r2_mean=0.918, adv_loss=2.47, acc_pert=0.181, val_re


Epoch 00054: cpa_metric reached. Module best state updated.

Epoch 60/2000:   3%| | 59/2000 [09:50<4:33:56,  8.47s/it, v_num=1, recon=1.27e+3, r2_mean=0.919, adv_loss=2.67, acc_pert=0.152, val_re


Epoch 00059: cpa_metric reached. Module best state updated.


disnt_basal = 0.06024990624800466
disnt_after = 0.21984197099233702
val_r2_mean = 0.9087773384871306
val_r2_var = 0.5638583727958628
Epoch 65/2000:   3%| | 64/2000 [10:34<4:42:46,  8.76s/it, v_num=1, recon=1.27e+3, r2_mean=0.918, adv_loss=2.84, acc_pert=0.125, val_re


Epoch 00064: cpa_metric reached. Module best state updated.

Epoch 70/2000:   3%| | 69/2000 [11:20<4:46:58,  8.92s/it, v_num=1, recon=1.27e+3, r2_mean=0.919, adv_loss=2.94, acc_pert=0.112, val_re


Epoch 00069: cpa_metric reached. Module best state updated.


disnt_basal = 0.05042097894349415
disnt_after = 0.2285000618741189
val_r2_mean = 0.904481914509025
val_r2_var = 0.5684133487117009
Epoch 75/2000:   4%| | 74/2000 [12:07<4:53:52,  9.16s/it, v_num=1, recon=1.27e+3, r2_mean=0.918, adv_loss=3.01, acc_pert=0.101, val_re


Epoch 00074: cpa_metric reached. Module best state updated.

Epoch 80/2000:   4%| | 79/2000 [12:53<4:47:11,  8.97s/it, v_num=1, recon=1.26e+3, r2_mean=0.919, adv_loss=3.06, acc_pert=0.0929, val_r


Epoch 00079: cpa_metric reached. Module best state updated.


disnt_basal = 0.04626068302570279
disnt_after = 0.2414682926452566
val_r2_mean = 0.90780227653908
val_r2_var = 0.563313795460595
Epoch 85/2000:   4%| | 84/2000 [13:38<4:38:39,  8.73s/it, v_num=1, recon=1.26e+3, r2_mean=0.919, adv_loss=3.08, acc_pert=0.0874, val_r


Epoch 00084: cpa_metric reached. Module best state updated.

Epoch 90/2000:   4%| | 89/2000 [14:23<4:46:46,  9.00s/it, v_num=1, recon=1.26e+3, r2_mean=0.918, adv_loss=3.11, acc_pert=0.0841, val_r


Epoch 00089: cpa_metric reached. Module best state updated.


disnt_basal = 0.04528630029442043
disnt_after = 0.25225609797660486
val_r2_mean = 0.9069772950727931
val_r2_var = 0.5681825789939675
Epoch 95/2000:   5%| | 94/2000 [15:10<4:56:06,  9.32s/it, v_num=1, recon=1.26e+3, r2_mean=0.919, adv_loss=3.13, acc_pert=0.0783, val_r


Epoch 00094: cpa_metric reached. Module best state updated.

Epoch 100/2000:   5%| | 99/2000 [15:54<4:40:07,  8.84s/it, v_num=1, recon=1.26e+3, r2_mean=0.919, adv_loss=3.15, acc_pert=0.0771, val_


Epoch 00099: cpa_metric reached. Module best state updated.


disnt_basal = 0.04494676974137365
disnt_after = 0.26197003263361773
val_r2_mean = 0.9158698047451701
val_r2_var = 0.5516319541818766
Epoch 105/2000:   5%| | 104/2000 [16:40<4:44:37,  9.01s/it, v_num=1, recon=1.26e+3, r2_mean=0.92, adv_loss=3.16, acc_pert=0.0744, val_


Epoch 00104: cpa_metric reached. Module best state updated.

Epoch 110/2000:   5%| | 109/2000 [17:30<5:01:15,  9.56s/it, v_num=1, recon=1.26e+3, r2_mean=0.919, adv_loss=3.17, acc_pert=0.0696, val
disnt_basal = 0.04339071490366559
disnt_after = 0.2604380221300531
val_r2_mean = 0.9125259323954983
val_r2_var = 0.5690988555099025
Epoch 115/2000:   6%| | 114/2000 [18:15<4:47:45,  9.15s/it, v_num=1, recon=1.26e+3, r2_mean=0.918, adv_loss=3.18, acc_pert=0.0698, val


Epoch 00114: cpa_metric reached. Module best state updated.

Epoch 120/2000:   6%| | 119/2000 [19:04<4:58:09,  9.51s/it, v_num=1, recon=1.26e+3, r2_mean=0.921, adv_loss=3.19, acc_pert=0.0673, val


Epoch 00119: cpa_metric reached. Module best state updated.


disnt_basal = 0.04304725258172981
disnt_after = 0.27402960880763183
val_r2_mean = 0.9129360739630883
val_r2_var = 0.5630708128113538
Epoch 125/2000:   6%| | 124/2000 [19:50<4:46:14,  9.16s/it, v_num=1, recon=1.26e+3, r2_mean=0.921, adv_loss=3.19, acc_pert=0.0665, val


Epoch 00124: cpa_metric reached. Module best state updated.

Epoch 130/2000:   6%| | 129/2000 [20:35<4:37:01,  8.88s/it, v_num=1, recon=1.26e+3, r2_mean=0.919, adv_loss=3.2, acc_pert=0.0651, val_


Epoch 00129: cpa_metric reached. Module best state updated.


disnt_basal = 0.04174844759476427
disnt_after = 0.28627420062442704
val_r2_mean = 0.9105156136281564
val_r2_var = 0.5637243307800806
Epoch 140/2000:   7%| | 139/2000 [22:02<4:23:46,  8.50s/it, v_num=1, recon=1.26e+3, r2_mean=0.92, adv_loss=3.21, acc_pert=0.0621, val_
disnt_basal = 0.04217145517764959
disnt_after = 0.2805999259742059
val_r2_mean = 0.9146810872386198
val_r2_var = 0.5651895538323655
Epoch 145/2000:   7%| | 144/2000 [22:43<4:14:15,  8.22s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.22, acc_pert=0.061, val_r


Epoch 00144: cpa_metric reached. Module best state updated.

Epoch 150/2000:   7%| | 149/2000 [23:28<4:29:36,  8.74s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.22, acc_pert=0.0607, val


Epoch 00149: cpa_metric reached. Module best state updated.


disnt_basal = 0.04115490579605907
disnt_after = 0.30270976116181003
val_r2_mean = 0.9097078623193684
val_r2_var = 0.5575797549803249
Epoch 155/2000:   8%| | 154/2000 [24:10<4:13:55,  8.25s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.22, acc_pert=0.0594, val_


Epoch 00154: cpa_metric reached. Module best state updated.

Epoch 160/2000:   8%| | 159/2000 [24:55<4:27:57,  8.73s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0601, val


Epoch 00159: cpa_metric reached. Module best state updated.


disnt_basal = 0.041702075277698675
disnt_after = 0.30185033605274075
val_r2_mean = 0.9131247589885184
val_r2_var = 0.5600758447791591
Epoch 165/2000:   8%| | 164/2000 [25:36<4:05:18,  8.02s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0592, val


Epoch 00164: cpa_metric reached. Module best state updated.

Epoch 170/2000:   8%| | 169/2000 [26:18<4:14:26,  8.34s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0582, val


Epoch 00169: cpa_metric reached. Module best state updated.


disnt_basal = 0.04054660772157918
disnt_after = 0.30944831027394487
val_r2_mean = 0.9047880594176476
val_r2_var = 0.5663980542609989
Epoch 175/2000:   9%| | 174/2000 [27:41<9:27:12, 18.64s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0568, val_


Epoch 00174: cpa_metric reached. Module best state updated.

Epoch 180/2000:   9%| | 179/2000 [29:28<9:23:10, 18.56s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0579, val_


Epoch 00179: cpa_metric reached. Module best state updated.


disnt_basal = 0.04031620204251352
disnt_after = 0.31916067204719945
val_r2_mean = 0.9093311804312246
val_r2_var = 0.5613407139424924
Epoch 185/2000:   9%| | 184/2000 [30:16<5:36:17, 11.11s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0575, val


Epoch 00184: cpa_metric reached. Module best state updated.

Epoch 190/2000:   9%| | 189/2000 [31:04<4:53:26,  9.72s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.23, acc_pert=0.0579, val


Epoch 00189: cpa_metric reached. Module best state updated.


disnt_basal = 0.04147435332487317
disnt_after = 0.32289914177116436
val_r2_mean = 0.9104435285333833
val_r2_var = 0.5612664056145383
Epoch 195/2000:  10%| | 194/2000 [31:51<4:46:24,  9.52s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0573, val_


Epoch 00194: cpa_metric reached. Module best state updated.

Epoch 200/2000:  10%| | 199/2000 [32:38<4:39:35,  9.31s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.057, val_r


Epoch 00199: cpa_metric reached. Module best state updated.


disnt_basal = 0.04115023126718784
disnt_after = 0.33321823779642357
val_r2_mean = 0.908866039731286
val_r2_var = 0.558030357959202
Epoch 205/2000:  10%| | 204/2000 [33:25<4:38:22,  9.30s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0573, val


Epoch 00204: cpa_metric reached. Module best state updated.

Epoch 210/2000:  10%| | 209/2000 [34:54<9:26:02, 18.96s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.23, acc_pert=0.0567, val
disnt_basal = 0.04024653138473878
disnt_after = 0.33018913026085883
val_r2_mean = 0.9091093530157198
val_r2_var = 0.5620534755164124
Epoch 215/2000:  11%| | 214/2000 [36:55<9:29:23, 19.13s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0559, val


Epoch 00214: cpa_metric reached. Module best state updated.

Epoch 220/2000:  11%| | 219/2000 [37:42<5:18:25, 10.73s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0555, val
disnt_basal = 0.04064928305878278
disnt_after = 0.332652966633433
val_r2_mean = 0.9123359880463443
val_r2_var = 0.5672176745202807
Epoch 225/2000:  11%| | 224/2000 [38:33<5:00:35, 10.16s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0568, val_


Epoch 00224: cpa_metric reached. Module best state updated.

Epoch 230/2000:  11%| | 229/2000 [39:24<5:03:03, 10.27s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0566, val_
disnt_basal = 0.04036115876415716
disnt_after = 0.3378564856058243
val_r2_mean = 0.910251923683115
val_r2_var = 0.566620121098528
Epoch 240/2000:  12%| | 239/2000 [41:00<4:30:41,  9.22s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0564, val_
disnt_basal = 0.04045372147394693
disnt_after = 0.34173993593466334
val_r2_mean = 0.9095593357728387
val_r2_var = 0.5628890019875986
Epoch 245/2000:  12%| | 244/2000 [41:47<4:32:58,  9.33s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0553, val_


Epoch 00244: cpa_metric reached. Module best state updated.

Epoch 250/2000:  12%| | 249/2000 [42:31<4:18:21,  8.85s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.23, acc_pert=0.0561, val


Epoch 00249: cpa_metric reached. Module best state updated.


disnt_basal = 0.039513698714692805
disnt_after = 0.34576276899510144
val_r2_mean = 0.9027658825370198
val_r2_var = 0.5758106174693766
Epoch 255/2000:  13%|▏| 254/2000 [43:18<4:24:13,  9.08s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.058, val_


Epoch 00254: cpa_metric reached. Module best state updated.

Epoch 260/2000:  13%|▏| 259/2000 [44:03<4:15:25,  8.80s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0575, val
disnt_basal = 0.03994983206016651
disnt_after = 0.3439185323833249
val_r2_mean = 0.910643367052881
val_r2_var = 0.5654164647012447
Epoch 265/2000:  13%|▏| 264/2000 [44:46<4:03:03,  8.40s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0558, val_


Epoch 00264: cpa_metric reached. Module best state updated.

Epoch 270/2000:  13%|▏| 269/2000 [45:31<4:12:00,  8.74s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0566, val
disnt_basal = 0.040166761968135005
disnt_after = 0.3504748097196458
val_r2_mean = 0.9097813715035666
val_r2_var = 0.5645362028369197
Epoch 275/2000:  14%|▏| 274/2000 [46:17<4:21:53,  9.10s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.057, val_


Epoch 00274: cpa_metric reached. Module best state updated.

Epoch 280/2000:  14%|▏| 279/2000 [47:00<4:07:23,  8.63s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0553, val


Epoch 00279: cpa_metric reached. Module best state updated.


disnt_basal = 0.040456163244333036
disnt_after = 0.3572522499600077
val_r2_mean = 0.910852282336264
val_r2_var = 0.5594399948714157
Epoch 285/2000:  14%|▏| 284/2000 [47:48<4:30:46,  9.47s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0576, val


Epoch 00284: cpa_metric reached. Module best state updated.

Epoch 290/2000:  14%|▏| 289/2000 [48:35<4:22:33,  9.21s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0559, val
disnt_basal = 0.03926832712372904
disnt_after = 0.3497102540936567
val_r2_mean = 0.9121499475003894
val_r2_var = 0.570735256479244
Epoch 295/2000:  15%|▏| 294/2000 [49:22<4:21:00,  9.18s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0551, val


Epoch 00294: cpa_metric reached. Module best state updated.

Epoch 300/2000:  15%|▏| 299/2000 [50:09<4:30:07,  9.53s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0568, val_


Epoch 00299: cpa_metric reached. Module best state updated.


disnt_basal = 0.03920168682453528
disnt_after = 0.36158794127635674
val_r2_mean = 0.9137285146818851
val_r2_var = 0.5629803686600774
Epoch 310/2000:  15%|▏| 309/2000 [52:19<5:36:45, 11.95s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0554, val


Epoch 00309: cpa_metric reached. Module best state updated.


disnt_basal = 0.03981924943625341
disnt_after = 0.3652186213880025
val_r2_mean = 0.9131556992177609
val_r2_var = 0.5631247309321908
Epoch 320/2000:  16%|▏| 319/2000 [53:49<4:05:29,  8.76s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0562, val


Epoch 00319: cpa_metric reached. Module best state updated.


disnt_basal = 0.04053643754989322
disnt_after = 0.37015536414977995
val_r2_mean = 0.9128284351994292
val_r2_var = 0.5662032074799843
Epoch 330/2000:  16%|▏| 329/2000 [55:20<4:13:39,  9.11s/it, v_num=1, recon=1.25e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0556, val_
disnt_basal = 0.040080655093989045
disnt_after = 0.3678346731429833
val_r2_mean = 0.9095952502806178
val_r2_var = 0.5599004443647083
Epoch 340/2000:  17%|▏| 339/2000 [56:49<4:05:14,  8.86s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.23, acc_pert=0.0569, val
disnt_basal = 0.03988857249217172
disnt_after = 0.3704604220867891
val_r2_mean = 0.9106467877975618
val_r2_var = 0.5663637701509777
Epoch 345/2000:  17%|▏| 344/2000 [57:48<5:50:59, 12.72s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0557, val_


Epoch 00344: cpa_metric reached. Module best state updated.

Epoch 350/2000:  17%|▏| 349/2000 [58:36<4:47:45, 10.46s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0567, val
disnt_basal = 0.03922064593854288
disnt_after = 0.3690430275185857
val_r2_mean = 0.9116486545765038
val_r2_var = 0.5657429996162955
Epoch 360/2000:  18%|▏| 359/2000 [1:00:18<5:28:15, 12.00s/it, v_num=1, recon=1.25e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0553, v
disnt_basal = 0.0403999673155449
disnt_after = 0.3734554326820138
val_r2_mean = 0.913087622886555
val_r2_var = 0.5652885898596509
Epoch 365/2000:  18%|▏| 364/2000 [1:01:30<6:21:43, 14.00s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0558, v


Epoch 00364: cpa_metric reached. Module best state updated.

Epoch 370/2000:  18%|▏| 369/2000 [1:02:43<5:58:14, 13.18s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0576, v
disnt_basal = 0.040913813357305866
disnt_after = 0.37245901582466023
val_r2_mean = 0.9121225030735285
val_r2_var = 0.5635640514820112
Epoch 380/2000:  19%|▏| 379/2000 [1:07:00<12:46:45, 28.38s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0564,


Epoch 00379: cpa_metric reached. Module best state updated.


disnt_basal = 0.039450868211319774
disnt_after = 0.3810250290713748
val_r2_mean = 0.9111390364691865
val_r2_var = 0.567552351791048
Epoch 390/2000:  19%|▏| 389/2000 [1:09:44<7:36:34, 17.00s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0565, v
disnt_basal = 0.039263429869721284
disnt_after = 0.37796660162953394
val_r2_mean = 0.9129030499100778
val_r2_var = 0.5679200446627295
Epoch 400/2000:  20%|▏| 399/2000 [1:11:17<4:32:52, 10.23s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.056, va
disnt_basal = 0.039448362508479856
disnt_after = 0.38306228513894836
val_r2_mean = 0.9112247363084092
val_r2_var = 0.5607114321856386
Epoch 410/2000:  20%|▏| 409/2000 [1:13:29<5:08:16, 11.63s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0567, v


Epoch 00409: cpa_metric reached. Module best state updated.


disnt_basal = 0.040036109266303084
disnt_after = 0.3852261108922756
val_r2_mean = 0.9092650594133318
val_r2_var = 0.5641676907988912
Epoch 415/2000:  21%|▏| 414/2000 [1:14:38<6:46:17, 15.37s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0557, v


Epoch 00414: cpa_metric reached. Module best state updated.

Epoch 420/2000:  21%|▏| 419/2000 [1:15:31<5:35:45, 12.74s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0564, v


Epoch 00419: cpa_metric reached. Module best state updated.


disnt_basal = 0.03985408537328855
disnt_after = 0.38631324868336364
val_r2_mean = 0.9112534330348777
val_r2_var = 0.5670535080762021
Epoch 430/2000:  21%|▏| 429/2000 [1:17:30<6:08:04, 14.06s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0551, va
disnt_basal = 0.04051310007875685
disnt_after = 0.38639743931358783
val_r2_mean = 0.9099488471092197
val_r2_var = 0.5686952479760654
Epoch 440/2000:  22%|▏| 439/2000 [1:20:45<9:44:56, 22.48s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.23, acc_pert=0.0559, va
disnt_basal = 0.03956727544223899
disnt_after = 0.3850694862411169
val_r2_mean = 0.9107194233823706
val_r2_var = 0.5699542847145286
Epoch 445/2000:  22%|▏| 444/2000 [1:22:10<7:22:45, 17.07s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0554, v


Epoch 00444: cpa_metric reached. Module best state updated.

Epoch 450/2000:  22%|▏| 449/2000 [1:22:46<3:45:54,  8.74s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.056, va
disnt_basal = 0.04032557922243797
disnt_after = 0.3854489771448042
val_r2_mean = 0.9121489962343414
val_r2_var = 0.5680684710191155
Epoch 455/2000:  23%|▏| 454/2000 [1:23:21<3:04:32,  7.16s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0554, v


Epoch 00454: cpa_metric reached. Module best state updated.

Epoch 460/2000:  23%|▏| 459/2000 [1:23:58<3:08:30,  7.34s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0574, v


Epoch 00459: cpa_metric reached. Module best state updated.


disnt_basal = 0.04082144897444196
disnt_after = 0.393595290106351
val_r2_mean = 0.9105518125123045
val_r2_var = 0.5637211177485558
Epoch 470/2000:  23%|▏| 469/2000 [1:25:07<2:53:05,  6.78s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0554, v
disnt_basal = 0.04000654212230136
disnt_after = 0.39207859489413543
val_r2_mean = 0.9072011171366632
val_r2_var = 0.5625318035934911
Epoch 480/2000:  24%|▏| 479/2000 [1:26:16<2:55:48,  6.93s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0563, v
disnt_basal = 0.03941728266625964
disnt_after = 0.39080270029069025
val_r2_mean = 0.9118343844558254
val_r2_var = 0.563226621158998
Epoch 485/2000:  24%|▏| 484/2000 [1:26:53<3:00:19,  7.14s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0563, v


Epoch 00484: cpa_metric reached. Module best state updated.

Epoch 490/2000:  24%|▏| 489/2000 [1:27:29<2:59:15,  7.12s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0541, v


Epoch 00489: cpa_metric reached. Module best state updated.


disnt_basal = 0.03966032767287265
disnt_after = 0.3974854925612059
val_r2_mean = 0.9125428593118584
val_r2_var = 0.5673860899527065
Epoch 500/2000:  25%|▏| 499/2000 [1:28:41<2:55:01,  7.00s/it, v_num=1, recon=1.25e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0564, v


Epoch 00499: cpa_metric reached. Module best state updated.


disnt_basal = 0.04011059166043884
disnt_after = 0.40303497237209807
val_r2_mean = 0.9112178507986787
val_r2_var = 0.5679613636028455
Epoch 510/2000:  25%|▎| 509/2000 [1:29:50<2:52:34,  6.94s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0564, v
disnt_basal = 0.03966800904716618
disnt_after = 0.40238836002755796
val_r2_mean = 0.9136719005276458
val_r2_var = 0.5598738438352592
Epoch 520/2000:  26%|▎| 519/2000 [1:30:59<2:46:12,  6.73s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0539, va
disnt_basal = 0.03960290847682412
disnt_after = 0.4008921205035158
val_r2_mean = 0.9125659736719999
val_r2_var = 0.5624201327863366
Epoch 530/2000:  26%|▎| 529/2000 [1:32:09<2:52:51,  7.05s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0564, v
disnt_basal = 0.040439245104308234
disnt_after = 0.4026161051268289
val_r2_mean = 0.9085394224333843
val_r2_var = 0.5630331177904148
Epoch 540/2000:  27%|▎| 539/2000 [1:33:17<2:39:52,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0537, v
disnt_basal = 0.03923726365872967
disnt_after = 0.40108718106710894
val_r2_mean = 0.9120809209065807
val_r2_var = 0.5652563539819686
Epoch 545/2000:  27%|▎| 544/2000 [1:33:52<2:45:16,  6.81s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0558, va


Epoch 00544: cpa_metric reached. Module best state updated.

Epoch 550/2000:  27%|▎| 549/2000 [1:34:26<2:43:00,  6.74s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0546, v
disnt_basal = 0.03915781291126416
disnt_after = 0.4030755817464731
val_r2_mean = 0.9123560565489309
val_r2_var = 0.5630666194138704
Epoch 560/2000:  28%|▎| 559/2000 [1:35:34<2:42:03,  6.75s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0566, v


Epoch 00559: cpa_metric reached. Module best state updated.


disnt_basal = 0.04046507063700886
disnt_after = 0.40739666400036006
val_r2_mean = 0.9103421920076363
val_r2_var = 0.5684774481888972
Epoch 570/2000:  28%|▎| 569/2000 [1:36:50<3:10:06,  7.97s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0541, v
disnt_basal = 0.03983058922661337
disnt_after = 0.4067729732602119
val_r2_mean = 0.9094641451482418
val_r2_var = 0.5637151167448923
Epoch 575/2000:  29%|▎| 574/2000 [1:37:25<2:50:54,  7.19s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0566, v


Epoch 00574: cpa_metric reached. Module best state updated.

Epoch 580/2000:  29%|▎| 579/2000 [1:38:01<2:49:19,  7.15s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0546, v
disnt_basal = 0.03989654234166392
disnt_after = 0.409478339726352
val_r2_mean = 0.9119101698551115
val_r2_var = 0.5607152105581882
Epoch 590/2000:  29%|▎| 589/2000 [1:39:12<2:39:40,  6.79s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0563, v


Epoch 00589: cpa_metric reached. Module best state updated.


disnt_basal = 0.04008174666683633
disnt_after = 0.4091380709161543
val_r2_mean = 0.9118934548663771
val_r2_var = 0.5687675744997531
Epoch 600/2000:  30%|▎| 599/2000 [1:40:21<2:36:38,  6.71s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0559, v
disnt_basal = 0.03945936974349657
disnt_after = 0.40948105689516134
val_r2_mean = 0.9115343384871178
val_r2_var = 0.5633621569033022
Epoch 605/2000:  30%|▎| 604/2000 [1:40:56<2:42:27,  6.98s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0559, v


Epoch 00604: cpa_metric reached. Module best state updated.

Epoch 610/2000:  30%|▎| 609/2000 [1:41:31<2:41:46,  6.98s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0553, v
disnt_basal = 0.03983297856496686
disnt_after = 0.40913880713434386
val_r2_mean = 0.9121930996576946
val_r2_var = 0.560360529607394
Epoch 620/2000:  31%|▎| 619/2000 [1:42:38<2:30:22,  6.53s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0557, v
disnt_basal = 0.03920774379521638
disnt_after = 0.40717118796240553
val_r2_mean = 0.9099390392351633
val_r2_var = 0.5730916909898571
Epoch 630/2000:  31%|▎| 629/2000 [1:43:46<2:31:26,  6.63s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0547, v
disnt_basal = 0.03930856699282243
disnt_after = 0.4135294977667994
val_r2_mean = 0.9113208601370405
val_r2_var = 0.5621969009489323
Epoch 635/2000:  32%|▎| 634/2000 [1:44:23<2:43:30,  7.18s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0556, v


Epoch 00634: cpa_metric reached. Module best state updated.

Epoch 640/2000:  32%|▎| 639/2000 [1:44:58<2:41:00,  7.10s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0567, v
disnt_basal = 0.03998190355022367
disnt_after = 0.41648636298660446
val_r2_mean = 0.9098926597974117
val_r2_var = 0.5641110952855762
Epoch 650/2000:  32%|▎| 649/2000 [1:46:10<2:39:46,  7.10s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0539, v
disnt_basal = 0.04062512074898763
disnt_after = 0.4182319889384008
val_r2_mean = 0.9078854786025153
val_r2_var = 0.5657896132581564
Epoch 660/2000:  33%|▎| 659/2000 [1:47:18<2:28:44,  6.65s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0565, v
disnt_basal = 0.03920011381094231
disnt_after = 0.4145832285443369
val_r2_mean = 0.9110531895248978
val_r2_var = 0.5683941782925667
Epoch 670/2000:  33%|▎| 669/2000 [1:48:30<2:38:16,  7.13s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0577, v
disnt_basal = 0.03973629514952688
disnt_after = 0.4161916465414118
val_r2_mean = 0.911522991729505
val_r2_var = 0.5664939310414221
Epoch 675/2000:  34%|▎| 674/2000 [1:49:05<2:36:10,  7.07s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0559, v


Epoch 00674: cpa_metric reached. Module best state updated.

Epoch 680/2000:  34%|▎| 679/2000 [1:49:43<2:37:32,  7.16s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0559, v
disnt_basal = 0.040222764175929605
disnt_after = 0.41784593320564567
val_r2_mean = 0.9137453535189132
val_r2_var = 0.560379992631148
Epoch 685/2000:  34%|▎| 684/2000 [1:50:17<2:30:54,  6.88s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0557, v


Epoch 00684: cpa_metric reached. Module best state updated.

Epoch 690/2000:  34%|▎| 689/2000 [1:50:52<2:27:09,  6.73s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0547, v
disnt_basal = 0.040333695251299255
disnt_after = 0.41985034210547184
val_r2_mean = 0.9100548554751205
val_r2_var = 0.568388536120906
Epoch 700/2000:  35%|▎| 699/2000 [1:52:05<2:36:43,  7.23s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0562, v


Epoch 00699: cpa_metric reached. Module best state updated.


disnt_basal = 0.041027925356964846
disnt_after = 0.4233228058604713
val_r2_mean = 0.9117571914236154
val_r2_var = 0.5636767169040461
Epoch 710/2000:  35%|▎| 709/2000 [1:53:13<2:22:35,  6.63s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0567, v


Epoch 00709: cpa_metric reached. Module best state updated.


disnt_basal = 0.039409497507492015
disnt_after = 0.42247398161870414
val_r2_mean = 0.9113324079850706
val_r2_var = 0.5669508355635183
Epoch 720/2000:  36%|▎| 719/2000 [1:54:20<2:20:36,  6.59s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0548, v
disnt_basal = 0.040729672080764695
disnt_after = 0.42138687909010686
val_r2_mean = 0.9110367306153783
val_r2_var = 0.5653694127545212
Epoch 730/2000:  36%|▎| 729/2000 [1:55:33<2:30:27,  7.10s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0552, v
disnt_basal = 0.03994646459073874
disnt_after = 0.4238356298508862
val_r2_mean = 0.9092787140949006
val_r2_var = 0.5650023759816232
Epoch 740/2000:  37%|▎| 739/2000 [1:56:41<2:18:24,  6.59s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0584, v


Epoch 00739: cpa_metric reached. Module best state updated.


disnt_basal = 0.039365748605678166
disnt_after = 0.42551301453613705
val_r2_mean = 0.9107407674644933
val_r2_var = 0.5681525763839183
Epoch 745/2000:  37%|▎| 744/2000 [1:57:15<2:19:34,  6.67s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0559, v


Epoch 00744: cpa_metric reached. Module best state updated.

Epoch 750/2000:  37%|▎| 749/2000 [1:57:49<2:18:34,  6.65s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0559, v
disnt_basal = 0.040018295381203425
disnt_after = 0.42385808265663216
val_r2_mean = 0.9114000237349308
val_r2_var = 0.571694547881181
Epoch 760/2000:  38%|▍| 759/2000 [1:58:58<2:16:28,  6.60s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0572, v
disnt_basal = 0.039818688661333784
disnt_after = 0.42718406511559215
val_r2_mean = 0.9102089328797981
val_r2_var = 0.5672914313547538
Epoch 770/2000:  38%|▍| 769/2000 [2:00:05<2:15:50,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0563, v
disnt_basal = 0.039839481851915076
disnt_after = 0.42604594068150026
val_r2_mean = 0.9101505229368755
val_r2_var = 0.5634078730637777
Epoch 780/2000:  39%|▍| 779/2000 [2:01:12<2:13:42,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.055, va
disnt_basal = 0.03937555023414729
disnt_after = 0.4275122479827335
val_r2_mean = 0.9102302637164441
val_r2_var = 0.5639477269818085
Epoch 790/2000:  39%|▍| 789/2000 [2:02:19<2:11:06,  6.50s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0552, v
disnt_basal = 0.03967748970781883
disnt_after = 0.4260601845986839
val_r2_mean = 0.912395669755711
val_r2_var = 0.565398547183785
Epoch 795/2000:  40%|▍| 794/2000 [2:02:52<2:10:25,  6.49s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0547, v


Epoch 00794: cpa_metric reached. Module best state updated.

Epoch 800/2000:  40%|▍| 799/2000 [2:03:25<2:10:45,  6.53s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0548, va
disnt_basal = 0.040129055712868296
disnt_after = 0.4287790500286537
val_r2_mean = 0.9085923342993766
val_r2_var = 0.5592593986177284
Epoch 810/2000:  40%|▍| 809/2000 [2:04:32<2:08:53,  6.49s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0576, v
disnt_basal = 0.03978977292012623
disnt_after = 0.4283306146154132
val_r2_mean = 0.9117273846860686
val_r2_var = 0.5662687559320468
Epoch 820/2000:  41%|▍| 819/2000 [2:05:41<2:16:21,  6.93s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0575, v


Epoch 00819: cpa_metric reached. Module best state updated.


disnt_basal = 0.040101563567173065
disnt_after = 0.43167082605914064
val_r2_mean = 0.9089021478036439
val_r2_var = 0.5689769230707727
Epoch 825/2000:  41%|▍| 824/2000 [2:06:14<2:10:35,  6.66s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0562, v


Epoch 00824: cpa_metric reached. Module best state updated.

Epoch 830/2000:  41%|▍| 829/2000 [2:06:49<2:11:04,  6.72s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0555, v
disnt_basal = 0.040385451888514716
disnt_after = 0.4318560382270798
val_r2_mean = 0.9108197165258004
val_r2_var = 0.5655405997025846
Epoch 840/2000:  42%|▍| 839/2000 [2:07:57<2:14:07,  6.93s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0564, v
disnt_basal = 0.03936340357283188
disnt_after = 0.4319931075824341
val_r2_mean = 0.9129863130524501
val_r2_var = 0.5640985833273994
Epoch 850/2000:  42%|▍| 849/2000 [2:09:04<2:07:27,  6.64s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0561, va
disnt_basal = 0.0397372158875321
disnt_after = 0.43239291608041097
val_r2_mean = 0.910749152653948
val_r2_var = 0.5605080904382648
Epoch 860/2000:  43%|▍| 859/2000 [2:10:13<2:06:01,  6.63s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0556, v


Epoch 00859: cpa_metric reached. Module best state updated.


disnt_basal = 0.039972875586427564
disnt_after = 0.43313296679628627
val_r2_mean = 0.9127384392099348
val_r2_var = 0.5674906118148907
Epoch 870/2000:  43%|▍| 869/2000 [2:11:23<2:15:48,  7.20s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0554, v
disnt_basal = 0.040441499509153815
disnt_after = 0.4330431447578728
val_r2_mean = 0.9108001782436562
val_r2_var = 0.5647250526280515
Epoch 880/2000:  44%|▍| 879/2000 [2:12:36<2:12:14,  7.08s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0554, v


Epoch 00879: cpa_metric reached. Module best state updated.


disnt_basal = 0.039467169581103086
disnt_after = 0.4326138905831209
val_r2_mean = 0.9132694804708567
val_r2_var = 0.5685608671971845
Epoch 890/2000:  44%|▍| 889/2000 [2:13:45<2:04:42,  6.74s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0569, v
disnt_basal = 0.03968554007372383
disnt_after = 0.4341542950645927
val_r2_mean = 0.9110849236799813
val_r2_var = 0.5667977005945712
Epoch 895/2000:  45%|▍| 894/2000 [2:14:18<2:02:46,  6.66s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0555, v


Epoch 00894: cpa_metric reached. Module best state updated.

Epoch 900/2000:  45%|▍| 899/2000 [2:14:54<2:04:36,  6.79s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0551, v
disnt_basal = 0.04002801836739362
disnt_after = 0.43453322998773847
val_r2_mean = 0.9109650386704339
val_r2_var = 0.5653133494685395
Epoch 910/2000:  45%|▍| 909/2000 [2:16:01<2:00:13,  6.61s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0557, v
disnt_basal = 0.04010239246493758
disnt_after = 0.4334131410466604
val_r2_mean = 0.9077517069148695
val_r2_var = 0.5662006065500305
Epoch 920/2000:  46%|▍| 919/2000 [2:17:11<2:02:32,  6.80s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.057, va


Epoch 00919: cpa_metric reached. Module best state updated.


disnt_basal = 0.03991615041537686
disnt_after = 0.43582988000614975
val_r2_mean = 0.9111762901749274
val_r2_var = 0.56778900651418
Epoch 930/2000:  46%|▍| 929/2000 [2:18:20<2:00:24,  6.75s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0567, va
disnt_basal = 0.03973688255140672
disnt_after = 0.4371336071421345
val_r2_mean = 0.9102193476375106
val_r2_var = 0.5642672155039881
Epoch 935/2000:  47%|▍| 934/2000 [2:18:53<1:56:25,  6.55s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0559, v


Epoch 00934: cpa_metric reached. Module best state updated.

Epoch 940/2000:  47%|▍| 939/2000 [2:19:27<1:57:00,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0546, v
disnt_basal = 0.03983495373206482
disnt_after = 0.43510781907019336
val_r2_mean = 0.9115133506280404
val_r2_var = 0.5662647381374731
Epoch 950/2000:  47%|▍| 949/2000 [2:20:35<1:56:28,  6.65s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0546, v
disnt_basal = 0.04041140069112158
disnt_after = 0.4391370876532196
val_r2_mean = 0.9110394890862282
val_r2_var = 0.5700022297274786
Epoch 960/2000:  48%|▍| 959/2000 [2:21:46<2:03:29,  7.12s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0556, v
disnt_basal = 0.04023616731677624
disnt_after = 0.4362614262738232
val_r2_mean = 0.9117268436284177
val_r2_var = 0.5671996998064445
Epoch 970/2000:  48%|▍| 969/2000 [2:22:56<1:57:34,  6.84s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0579, v
disnt_basal = 0.03993376411890781
disnt_after = 0.4411488578450985
val_r2_mean = 0.9109238099786352
val_r2_var = 0.561231227095218
Epoch 980/2000:  49%|▍| 979/2000 [2:24:08<1:55:48,  6.81s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0555, v


Epoch 00979: cpa_metric reached. Module best state updated.


disnt_basal = 0.039397143297156614
disnt_after = 0.4410704913546864
val_r2_mean = 0.9108223037625865
val_r2_var = 0.5698059643373932
Epoch 990/2000:  49%|▍| 989/2000 [2:25:16<1:52:38,  6.69s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0552, v


Epoch 00989: cpa_metric reached. Module best state updated.


disnt_basal = 0.04059361643608334
disnt_after = 0.44338921658590724
val_r2_mean = 0.9126005628602364
val_r2_var = 0.5661275423088706
Epoch 1000/2000:  50%|▍| 999/2000 [2:26:23<1:51:41,  6.69s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0558,
disnt_basal = 0.03945138453362174
disnt_after = 0.44181219223775725
val_r2_mean = 0.9112518897361628
val_r2_var = 0.5649537538438533
Epoch 1005/2000:  50%|▌| 1004/2000 [2:26:57<1:49:22,  6.59s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0561,


Epoch 01004: cpa_metric reached. Module best state updated.

Epoch 1010/2000:  50%|▌| 1009/2000 [2:27:30<1:49:23,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0571,
disnt_basal = 0.039715777382755106
disnt_after = 0.4400858996343968
val_r2_mean = 0.9115443992293643
val_r2_var = 0.5662362932757496
Epoch 1020/2000:  51%|▌| 1019/2000 [2:28:39<1:50:39,  6.77s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0579,
disnt_basal = 0.0398945606745382
disnt_after = 0.44173132868575654
val_r2_mean = 0.9097537799716396
val_r2_var = 0.5687487466568095
Epoch 1030/2000:  51%|▌| 1029/2000 [2:29:50<1:52:00,  6.92s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0563,
disnt_basal = 0.03945085294649067
disnt_after = 0.43940896767287696
val_r2_mean = 0.9134739537030359
val_r2_var = 0.5657271012312636
Epoch 1040/2000:  52%|▌| 1039/2000 [2:31:01<1:51:59,  6.99s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0549,
disnt_basal = 0.03963863469464625
disnt_after = 0.4398847180837058
val_r2_mean = 0.9120007895861412
val_r2_var = 0.5633916844823946
Epoch 1050/2000:  52%|▌| 1049/2000 [2:32:08<1:43:55,  6.56s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0553,
disnt_basal = 0.0407381431245832
disnt_after = 0.44254126870720895
val_r2_mean = 0.9128285265438175
val_r2_var = 0.5664128749150126
Epoch 1055/2000:  53%|▌| 1054/2000 [2:32:42<1:45:35,  6.70s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0553,


Epoch 01054: cpa_metric reached. Module best state updated.

Epoch 1060/2000:  53%|▌| 1059/2000 [2:33:15<1:43:00,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.058,
disnt_basal = 0.03945341352487521
disnt_after = 0.44027887205230304
val_r2_mean = 0.9098600775304465
val_r2_var = 0.5712726742330223
Epoch 1070/2000:  53%|▌| 1069/2000 [2:34:24<1:44:50,  6.76s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0556,
disnt_basal = 0.040885178748950785
disnt_after = 0.44270544655399013
val_r2_mean = 0.9128227749657551
val_r2_var = 0.5616051295791009
Epoch 1080/2000:  54%|▌| 1079/2000 [2:35:35<1:44:19,  6.80s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0559,


Epoch 01079: cpa_metric reached. Module best state updated.


disnt_basal = 0.03872682935783932
disnt_after = 0.4446248129765948
val_r2_mean = 0.9127585293869377
val_r2_var = 0.5652296820072212
Epoch 1090/2000:  54%|▌| 1089/2000 [2:36:41<1:39:41,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0546,
disnt_basal = 0.03932429227718138
disnt_after = 0.44341833735649044
val_r2_mean = 0.9109610174641464
val_r2_var = 0.5671557458964261
Epoch 1100/2000:  55%|▌| 1099/2000 [2:37:48<1:39:58,  6.66s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.056,
disnt_basal = 0.039865390514568375
disnt_after = 0.44632313461844875
val_r2_mean = 0.9096050481202226
val_r2_var = 0.5655384408905851
Epoch 1110/2000:  55%|▌| 1109/2000 [2:38:57<1:43:22,  6.96s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0554,
disnt_basal = 0.0394382041607409
disnt_after = 0.4445881810517868
val_r2_mean = 0.9111838328717934
val_r2_var = 0.562625375660983
Epoch 1115/2000:  56%|▌| 1114/2000 [2:39:31<1:39:33,  6.74s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0547,


Epoch 01114: cpa_metric reached. Module best state updated.

Epoch 1120/2000:  56%|▌| 1119/2000 [2:40:05<1:38:21,  6.70s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0559,
disnt_basal = 0.039695891628164545
disnt_after = 0.443741103232636
val_r2_mean = 0.9104009192399304
val_r2_var = 0.5646421090520992
Epoch 1130/2000:  56%|▌| 1129/2000 [2:41:12<1:34:32,  6.51s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0566,
disnt_basal = 0.03964405818036127
disnt_after = 0.44602787205486644
val_r2_mean = 0.9114504568905943
val_r2_var = 0.5667104877606787
Epoch 1140/2000:  57%|▌| 1139/2000 [2:42:20<1:38:27,  6.86s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0543,
disnt_basal = 0.040035254314409845
disnt_after = 0.4462739480869741
val_r2_mean = 0.9095784275218695
val_r2_var = 0.5703863491514316
Epoch 1150/2000:  57%|▌| 1149/2000 [2:43:28<1:33:12,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0564,
disnt_basal = 0.039782358132807775
disnt_after = 0.44617357055534707
val_r2_mean = 0.9109337042879175
val_r2_var = 0.5653985417651809
Epoch 1160/2000:  58%|▌| 1159/2000 [2:44:36<1:33:51,  6.70s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0552,
disnt_basal = 0.040289368921927825
disnt_after = 0.4460170764507765
val_r2_mean = 0.9099192791916303
val_r2_var = 0.5603986946420638
Epoch 1165/2000:  58%|▌| 1164/2000 [2:45:10<1:33:15,  6.69s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0555,


Epoch 01164: cpa_metric reached. Module best state updated.

Epoch 1170/2000:  58%|▌| 1169/2000 [2:45:44<1:30:54,  6.56s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0551,
disnt_basal = 0.039659923088613484
disnt_after = 0.4456627729711075
val_r2_mean = 0.9105888085333185
val_r2_var = 0.5659774931191595
Epoch 1180/2000:  59%|▌| 1179/2000 [2:46:52<1:31:12,  6.67s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0541,
disnt_basal = 0.03882958976970779
disnt_after = 0.44761502621687227
val_r2_mean = 0.9108926708048041
val_r2_var = 0.5652395065384681
Epoch 1190/2000:  59%|▌| 1189/2000 [2:47:59<1:29:17,  6.61s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0549,
disnt_basal = 0.040539341476762865
disnt_after = 0.44775312996430067
val_r2_mean = 0.9115842305851303
val_r2_var = 0.5679958175729822
Epoch 1195/2000:  60%|▌| 1194/2000 [2:48:33<1:28:11,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0573,


Epoch 01194: cpa_metric reached. Module best state updated.

Epoch 1200/2000:  60%|▌| 1199/2000 [2:49:07<1:28:23,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0548,
disnt_basal = 0.04010349733203377
disnt_after = 0.44889055816284446
val_r2_mean = 0.910777935476014
val_r2_var = 0.5625951033248644
Epoch 1210/2000:  60%|▌| 1209/2000 [2:50:15<1:30:40,  6.88s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0532,


Epoch 01209: cpa_metric reached. Module best state updated.


disnt_basal = 0.0400787023582781
disnt_after = 0.45049128117328335
val_r2_mean = 0.9107607974347843
val_r2_var = 0.5625823045820499
Epoch 1220/2000:  61%|▌| 1219/2000 [2:51:23<1:25:48,  6.59s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0573,


Epoch 01219: cpa_metric reached. Module best state updated.


disnt_basal = 0.04042053347151885
disnt_after = 0.45016382608402916
val_r2_mean = 0.9105550098900842
val_r2_var = 0.5696414289651094
Epoch 1230/2000:  61%|▌| 1229/2000 [2:52:31<1:23:08,  6.47s/it, v_num=1, recon=1.24e+3, r2_mean=0.92, adv_loss=3.24, acc_pert=0.0577,
disnt_basal = 0.03935432695492893
disnt_after = 0.448162785216132
val_r2_mean = 0.9124219893205044
val_r2_var = 0.5676547563838638
Epoch 1240/2000:  62%|▌| 1239/2000 [2:53:38<1:23:16,  6.57s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0558,
disnt_basal = 0.03910866825741823
disnt_after = 0.449844174840016
val_r2_mean = 0.910750227343755
val_r2_var = 0.5659096210893959
Epoch 1250/2000:  62%|▌| 1249/2000 [2:54:45<1:22:49,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0547,


Epoch 01249: cpa_metric reached. Module best state updated.


disnt_basal = 0.04005837644727899
disnt_after = 0.45128589792278295
val_r2_mean = 0.9116241014766373
val_r2_var = 0.5718753947954788
Epoch 1260/2000:  63%|▋| 1259/2000 [2:55:52<1:20:39,  6.53s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0558,
disnt_basal = 0.041013348748899586
disnt_after = 0.45203367148527496
val_r2_mean = 0.9116415225125695
val_r2_var = 0.568181695360126
Epoch 1270/2000:  63%|▋| 1269/2000 [2:56:58<1:19:26,  6.52s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0544,
disnt_basal = 0.040079499920979235
disnt_after = 0.451969804098721
val_r2_mean = 0.9107683969266486
val_r2_var = 0.5626417771734372
Epoch 1280/2000:  64%|▋| 1279/2000 [2:58:04<1:18:27,  6.53s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0559,
disnt_basal = 0.03968806239642345
disnt_after = 0.4518606234904323
val_r2_mean = 0.9122602389717744
val_r2_var = 0.5667908167598223
Epoch 1285/2000:  64%|▋| 1284/2000 [2:58:39<1:18:56,  6.61s/it, v_num=1, recon=1.24e+3, r2_mean=0.921, adv_loss=3.24, acc_pert=0.0539,


Epoch 01284: cpa_metric reached. Module best state updated.

Epoch 1290/2000:  64%|▋| 1289/2000 [2:59:13<1:19:05,  6.67s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0552,


Epoch 01289: cpa_metric reached. Module best state updated.


disnt_basal = 0.04108327848704389
disnt_after = 0.45481212339777266
val_r2_mean = 0.9124768253312584
val_r2_var = 0.5663197441086871
Epoch 1300/2000:  65%|▋| 1299/2000 [3:00:20<1:17:43,  6.65s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0565,
disnt_basal = 0.03987365171991119
disnt_after = 0.45233114784100753
val_r2_mean = 0.9104704680266202
val_r2_var = 0.5659425653993883
Epoch 1305/2000:  65%|▋| 1304/2000 [3:00:54<1:16:45,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0549,


Epoch 01304: cpa_metric reached. Module best state updated.

Epoch 1310/2000:  65%|▋| 1309/2000 [3:01:28<1:17:36,  6.74s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0547,
disnt_basal = 0.03956814775582223
disnt_after = 0.4534955289747813
val_r2_mean = 0.9090922693612197
val_r2_var = 0.568383286296318
Epoch 1320/2000:  66%|▋| 1319/2000 [3:02:35<1:14:37,  6.58s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0555,
disnt_basal = 0.03931314447108622
disnt_after = 0.4513691495645237
val_r2_mean = 0.9123028946243953
val_r2_var = 0.5600161604608349
Epoch 1330/2000:  66%|▋| 1329/2000 [3:03:45<1:14:28,  6.66s/it, v_num=1, recon=1.24e+3, r2_mean=0.922, adv_loss=3.24, acc_pert=0.0557,
disnt_basal = 0.04001099383091684
disnt_after = 0.4539275528216418
val_r2_mean = 0.9094210716610405
val_r2_var = 0.5686852831631799
Epoch 1340/2000:  67%|▋| 1339/2000 [3:04:52<1:13:15,  6.65s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0537,
disnt_basal = 0.03894327219927795
disnt_after = 0.45272663449528605
val_r2_mean = 0.9094103430256698
val_r2_var = 0.5657511393630545
Epoch 1350/2000:  67%|▋| 1349/2000 [3:05:59<1:11:52,  6.62s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0571,
disnt_basal = 0.040455072842637586
disnt_after = 0.4540434061979121
val_r2_mean = 0.9114404830467019
val_r2_var = 0.5665313258315577
Epoch 1355/2000:  68%|▋| 1355/2000 [3:06:41<1:28:52,  8.27s/it, v_num=1, recon=1.24e+3, r2_mean=0.923, adv_loss=3.24, acc_pert=0.0564,

[14]:

cpa.pl.plot_history(model)
../_images/tutorials_combosciplex_Rdkit_embeddings_34_0.png

If you already trained CPA, you can restore model weights by running the following cell:

[15]:

# model = cpa.CPA.load(dir_path='/home/mohsen/projects/cpa/lightning_logs/combo_rdkit/',
#                      adata=adata, use_gpu=True)

Latent space UMAP visualization#

Here, we visualize the latent representations of all cells. We computed basal and final latent representations with model.get_latent_representation function. The function will return a python dictionary as output with the following keys:

  • latent_basal: Basal latent representation of the cells (before perturbation) as anndata object

  • latent_after: Final latent representation of the cells (after perturbation) as anndata object

  • latent_corrected: Batch-corrected (if batch_key was provided when calling setup_anndata) latent representation of the cells (after perturbation) as anndata object

[16]:

latent_outputs = model.get_latent_representation(adata, batch_size=1024)

100%|█████████████████████████████████████████████████████████████████████████████████████████████████| 62/62 [00:02<00:00, 23.61it/s]

[17]:

sc.settings.verbosity = 3

[18]:

latent_basal_adata = latent_outputs['latent_basal']
latent_adata = latent_outputs['latent_after']

[19]:

sc.pp.neighbors(latent_basal_adata)
sc.tl.umap(latent_basal_adata)

computing neighbors
WARNING: You’re trying to run this on 64 dimensions of `.X`, if you really want this, set `use_rep='X'`.
         Falling back to preprocessing with `sc.pp.pca` and default params.
computing PCA
    with n_comps=50
    finished (0:00:00)
    finished: added to `.uns['neighbors']`
    `.obsp['distances']`, distances for each pair of neighbors
    `.obsp['connectivities']`, weighted adjacency matrix (0:03:03)
computing UMAP
    finished: added
    'X_umap', UMAP coordinates (adata.obsm) (0:00:31)

[20]:

latent_basal_adata

[20]:

AnnData object with n_obs × n_vars = 63378 × 64
    obs: 'ncounts', 'well', 'plate', 'cell_line', 'replicate', 'time', 'dose_value', 'pathway1', 'pathway2', 'perturbation', 'target', 'pathway', 'dose_unit', 'celltype', 'disease', 'tissue_type', 'organism', 'perturbation_type', 'ngenes', 'percent_mito', 'percent_ribo', 'nperts', 'chembl-ID', 'condition', 'condition_ID', 'control', 'cell_type', 'smiles_rdkit', 'source', 'sample', 'Size_Factor', 'n.umi', 'RT_well', 'Drug1', 'Drug2', 'Well', 'n_genes', 'n_genes_by_counts', 'total_counts', 'total_counts_mt', 'pct_counts_mt', 'leiden', 'split', 'condition_old', 'pert_type', 'batch', 'split_1ct_MEC', 'split_2ct_MEC', 'split_3ct_MEC', 'batch_cov', 'batch_cov_cond', 'log_dose', 'cov_drug_dose', 'CPA_cat', 'CPA_CHEMBL504', '_scvi_condition_ID', '_scvi_cell_type', '_scvi_CPA_cat'
    uns: 'neighbors', 'umap'
    obsm: 'X_pca', 'X_umap'
    obsp: 'distances', 'connectivities'

Basal Latent#

The basal representation should be free of the variation(s) of the `’condition_ID’ as observed below

[21]:

sc.pl.umap(latent_basal_adata, color=['condition_ID'], frameon=False, wspace=0.2)
../_images/tutorials_combosciplex_Rdkit_embeddings_46_0.png 
[22]:

sc.pp.neighbors(latent_adata)
sc.tl.umap(latent_adata)

computing neighbors
WARNING: You’re trying to run this on 64 dimensions of `.X`, if you really want this, set `use_rep='X'`.
         Falling back to preprocessing with `sc.pp.pca` and default params.
computing PCA
    with n_comps=50
    finished (0:00:00)
    finished: added to `.uns['neighbors']`
    `.obsp['distances']`, distances for each pair of neighbors
    `.obsp['connectivities']`, weighted adjacency matrix (0:00:08)
computing UMAP
    finished: added
    'X_umap', UMAP coordinates (adata.obsm) (0:00:34)

Final Latent#

Here, you can visualize that when the drug embedding is added to the basal representation, the cells treated with different drugs will be separated from each other.

[23]:

sc.pl.umap(latent_adata, color=['condition_ID'], frameon=False, wspace=0.2)
../_images/tutorials_combosciplex_Rdkit_embeddings_50_0.png

Evaluation#

Next, we will evaluate the model’s prediction performance on the whole dataset, including OOD (test) cells. The model will report metrics on how well we have captured the variation in top n differentially expressed genes when compared to control cells (DMSO, CHEMBL 504) for each condition. The metrics calculate the mean accuracy (r2_mean_deg), the variance (r2_var_deg) and similar metrics (r2_mean_lfc_deg and log fold change)to measure the log fold change of the predicted cells vs control((LFC(control, ground truth) ~ LFC(control, predicted cells)). The R2 is the sklearn.metrics.r2_score from sklearn.

[ ]:

model.predict(adata, batch_size=1024)

[25]:

import numpy as np
import pandas as pd
from sklearn.metrics import r2_score
from collections import defaultdict
from tqdm import tqdm

n_top_degs = [10, 20, 50, None] # None means all genes

results = defaultdict(list)
ctrl_adata = adata[adata.obs['condition_ID'] == 'CHEMBL504'].copy()
for cat in tqdm(adata.obs['cov_drug_dose'].unique()):
    if 'CHEMBL504' not in cat:
        cat_adata = adata[adata.obs['cov_drug_dose'] == cat].copy()

        deg_cat = f'{cat}'
        deg_list = adata.uns['rank_genes_groups_cov'][deg_cat]

        x_true = cat_adata.layers['counts'].toarray()
        x_pred = cat_adata.obsm['CPA_pred']
        x_ctrl = ctrl_adata.layers['counts'].toarray()

        x_true = np.log1p(x_true)
        x_pred = np.log1p(x_pred)
        x_ctrl = np.log1p(x_ctrl)

        for n_top_deg in n_top_degs:
            if n_top_deg is not None:
                degs = np.where(np.isin(adata.var_names, deg_list[:n_top_deg]))[0]
            else:
                degs = np.arange(adata.n_vars)
                n_top_deg = 'all'

            x_true_deg = x_true[:, degs]
            x_pred_deg = x_pred[:, degs]
            x_ctrl_deg = x_ctrl[:, degs]

            r2_mean_deg = r2_score(x_true_deg.mean(0), x_pred_deg.mean(0))
            r2_var_deg = r2_score(x_true_deg.var(0), x_pred_deg.var(0))

            r2_mean_lfc_deg = r2_score(x_true_deg.mean(0) - x_ctrl_deg.mean(0), x_pred_deg.mean(0) - x_ctrl_deg.mean(0))
            r2_var_lfc_deg = r2_score(x_true_deg.var(0) - x_ctrl_deg.var(0), x_pred_deg.var(0) - x_ctrl_deg.var(0))

            cov, cond, dose = cat.split('_')

            results['cell_type'].append(cov)
            results['condition'].append(cond)
            results['dose'].append(dose)
            results['n_top_deg'].append(n_top_deg)
            results['r2_mean_deg'].append(r2_mean_deg)
            results['r2_var_deg'].append(r2_var_deg)
            results['r2_mean_lfc_deg'].append(r2_mean_lfc_deg)
            results['r2_var_lfc_deg'].append(r2_var_lfc_deg)

df = pd.DataFrame(results)

100%|█████████████████████████████████████████████████████████████████████████████████████████████████| 32/32 [00:15<00:00,  2.12it/s]

n_top_deg shows how many DEGs genes were used to calculate the metric.

[26]:

df[df['n_top_deg'] == 20]

[26]:
cell_type condition dose n_top_deg r2_mean_deg r2_var_deg r2_mean_lfc_deg r2_var_lfc_deg
1 A549 CHEMBL483254 3.0 20 0.943252 0.222424 0.985652 0.758756
5 A549 CHEMBL491473+CHEMBL2170177 3.0+3.0 20 0.953437 -0.723978 0.345032 -19.256924
9 A549 CHEMBL1213492+CHEMBL257991 3.0+3.0 20 0.815992 -2.668683 0.308896 -6.126473
13 A549 CHEMBL483254+46245047 3.0+3.0 20 0.948573 0.245916 0.985587 0.741571
17 A549 CHEMBL483254+CHEMBL2170177 3.0+3.0 20 0.944003 0.019561 0.982641 0.619355
21 A549 CHEMBL356066 3.0 20 0.933226 -0.177728 0.977254 0.643084
25 A549 CHEMBL356066+CHEMBL2170177 3.0+3.0 20 0.943791 0.077367 0.982595 0.729290
29 A549 CHEMBL483254+CHEMBL1200485 3.0+3.0 20 0.951563 0.166435 0.985437 0.678770
33 A549 CHEMBL1213492+CHEMBL491473 3.0+3.0 20 0.607602 -1.442659 0.467476 -3.483045
37 A549 CHEMBL1213492 3.0 20 0.845446 -3.124852 0.555419 -1.616240
41 A549 CHEMBL356066+CHEMBL402548 3.0+3.0 20 0.748116 -0.468033 0.923252 0.609934
45 A549 CHEMBL483254+CHEMBL1421 3.0+3.0 20 0.934773 -0.018929 0.982338 0.645029
49 A549 CHEMBL1213492+CHEMBL109480 3.0+3.0 20 0.706345 -1.185113 0.873071 -1.703947
53 A549 CHEMBL1213492+CHEMBL460499 3.0+3.0 20 0.863053 -3.360452 0.682881 -4.661797
57 A549 CHEMBL483254+CHEMBL4297436 3.0+3.0 20 0.942252 0.051656 0.983361 0.665030
61 A549 CHEMBL483254+CHEMBL257991 3.0+3.0 20 0.940087 0.079230 0.983258 0.683227
65 A549 CHEMBL483254+CHEMBL601719 3.0+3.0 20 0.943840 -0.008598 0.981825 0.608500
69 A549 CHEMBL383824+CHEMBL2354444 3.0+3.0 20 0.936724 -1.103101 0.957802 0.116940
73 A549 CHEMBL1213492+CHEMBL4297436 3.0+3.0 20 0.792289 -1.255896 0.387688 -1.011227
77 A549 CHEMBL1213492+CHEMBL1200485 3.0+3.0 20 0.931320 -0.043086 0.966058 0.680356
81 A549 CHEMBL483254+CHEMBL383824 3.0+3.0 20 0.496878 -1.205903 0.835792 0.192209
85 A549 CHEMBL383824 3.0 20 0.896525 -2.160942 0.858206 -1.138908
89 A549 CHEMBL483254+CHEMBL116438 3.0+3.0 20 0.950202 0.287852 0.986677 0.770033
93 A549 CHEMBL1213492+CHEMBL1421 3.0+3.0 20 0.842568 -3.738260 0.780868 -3.050616
97 A549 CHEMBL4297436+CHEMBL383824 3.0+3.0 20 0.888726 -1.714776 0.927265 -0.432541
101 A549 CHEMBL356066+CHEMBL1421 3.0+3.0 20 0.901944 0.031103 0.975665 0.713307
105 A549 CHEMBL4297436 3.0 20 0.728050 -3.757220 -1.039214 -15.514841
109 A549 CHEMBL1213492+CHEMBL116438 3.0+3.0 20 0.884812 -3.038191 0.601723 -5.172740
113 A549 CHEMBL1213492+CHEMBL601719 3.0+3.0 20 0.810477 -0.981200 0.772785 -3.749685
117 A549 46245047+CHEMBL491473 3.0+3.0 20 0.875358 -2.926035 -1.173367 -27.171973
121 A549 CHEMBL1421 3.0 20 0.936153 -0.578474 0.856687 -1.833810

We can further visualize these per condition

[27]:

for cat in adata.obs["cov_drug_dose"].unique():
    if "CHEMBL504" not in cat:
        cat_adata = adata[adata.obs["cov_drug_dose"] == cat].copy()

        cat_adata.X = np.log1p(cat_adata.layers["counts"].A)
        cat_adata.obsm["CPA_pred"] = np.log1p(cat_adata.obsm["CPA_pred"])

        deg_list = adata.uns["rank_genes_groups_cov"][f'{cat}'][:20]

        print(cat, f"{cat_adata.shape}")
        cpa.pl.mean_plot(
            cat_adata,
            pred_obsm_key="CPA_pred",
            path_to_save=None,
            deg_list=deg_list,
            # gene_list=deg_list[:5],
            show=True,
            verbose=True,
        )

A549_CHEMBL483254_3.0 (1578, 5000)
Top 20 DEGs var:  0.9432570119362333
All genes var:  0.2383808579912715
../_images/tutorials_combosciplex_Rdkit_embeddings_58_1.png 
A549_CHEMBL491473+CHEMBL2170177_3.0+3.0 (2161, 5000)
Top 20 DEGs var:  0.9534399036528307
All genes var:  0.32357108908961274
../_images/tutorials_combosciplex_Rdkit_embeddings_58_3.png 
A549_CHEMBL1213492+CHEMBL257991_3.0+3.0 (2260, 5000)
Top 20 DEGs var:  0.8160014574452075
All genes var:  0.27984262715152175
../_images/tutorials_combosciplex_Rdkit_embeddings_58_5.png 
A549_CHEMBL483254+46245047_3.0+3.0 (1889, 5000)
Top 20 DEGs var:  0.9485761406990977
All genes var:  0.3656485031352644
../_images/tutorials_combosciplex_Rdkit_embeddings_58_7.png 
A549_CHEMBL483254+CHEMBL2170177_3.0+3.0 (1814, 5000)
Top 20 DEGs var:  0.9440058158529224
All genes var:  0.28778099233742227
../_images/tutorials_combosciplex_Rdkit_embeddings_58_9.png 
A549_CHEMBL356066_3.0 (1869, 5000)
Top 20 DEGs var:  0.9332284720946527
All genes var:  0.20723961915360356
../_images/tutorials_combosciplex_Rdkit_embeddings_58_11.png 
A549_CHEMBL356066+CHEMBL2170177_3.0+3.0 (3298, 5000)
Top 20 DEGs var:  0.9437920785334201
All genes var:  0.22203679591558112
../_images/tutorials_combosciplex_Rdkit_embeddings_58_13.png 
A549_CHEMBL483254+CHEMBL1200485_3.0+3.0 (2013, 5000)
Top 20 DEGs var:  0.9515651464946858
All genes var:  0.33489523506046215
../_images/tutorials_combosciplex_Rdkit_embeddings_58_15.png 
A549_CHEMBL1213492+CHEMBL491473_3.0+3.0 (2783, 5000)
Top 20 DEGs var:  0.6076002851215299
All genes var:  0.27443908963125596
../_images/tutorials_combosciplex_Rdkit_embeddings_58_17.png 
A549_CHEMBL1213492_3.0 (1682, 5000)
Top 20 DEGs var:  0.8454551331901641
All genes var:  0.24572231027227376
../_images/tutorials_combosciplex_Rdkit_embeddings_58_19.png 
A549_CHEMBL356066+CHEMBL402548_3.0+3.0 (1939, 5000)
Top 20 DEGs var:  0.7481165747181706
All genes var:  0.12389914380865896
../_images/tutorials_combosciplex_Rdkit_embeddings_58_21.png 
A549_CHEMBL483254+CHEMBL1421_3.0+3.0 (1955, 5000)
Top 20 DEGs var:  0.934776396223971
All genes var:  0.2366556337809702
../_images/tutorials_combosciplex_Rdkit_embeddings_58_23.png 
A549_CHEMBL1213492+CHEMBL109480_3.0+3.0 (1310, 5000)
Top 20 DEGs var:  0.706362704506477
All genes var:  0.3362485928159149
../_images/tutorials_combosciplex_Rdkit_embeddings_58_25.png 
A549_CHEMBL1213492+CHEMBL460499_3.0+3.0 (2692, 5000)
Top 20 DEGs var:  0.863054790510513
All genes var:  0.23900141664137675
../_images/tutorials_combosciplex_Rdkit_embeddings_58_27.png 
A549_CHEMBL483254+CHEMBL4297436_3.0+3.0 (1971, 5000)
Top 20 DEGs var:  0.9422541889114032
All genes var:  0.22341950424153467
../_images/tutorials_combosciplex_Rdkit_embeddings_58_29.png 
A549_CHEMBL483254+CHEMBL257991_3.0+3.0 (1826, 5000)
Top 20 DEGs var:  0.9400900790227729
All genes var:  0.24910902525384204
../_images/tutorials_combosciplex_Rdkit_embeddings_58_31.png 
A549_CHEMBL483254+CHEMBL601719_3.0+3.0 (1641, 5000)
Top 20 DEGs var:  0.9438439375563483
All genes var:  0.2952641755499209
../_images/tutorials_combosciplex_Rdkit_embeddings_58_33.png 
A549_CHEMBL383824+CHEMBL2354444_3.0+3.0 (476, 5000)
Top 20 DEGs var:  0.936723645221692
All genes var:  0.3553705779281776
../_images/tutorials_combosciplex_Rdkit_embeddings_58_35.png 
A549_CHEMBL1213492+CHEMBL4297436_3.0+3.0 (2353, 5000)
Top 20 DEGs var:  0.792297524159662
All genes var:  0.3107678007497654
../_images/tutorials_combosciplex_Rdkit_embeddings_58_37.png 
A549_CHEMBL1213492+CHEMBL1200485_3.0+3.0 (2734, 5000)
Top 20 DEGs var:  0.9313229171757604
All genes var:  0.17166578684608935
../_images/tutorials_combosciplex_Rdkit_embeddings_58_39.png 
A549_CHEMBL483254+CHEMBL383824_3.0+3.0 (996, 5000)
Top 20 DEGs var:  0.4968727606901995
All genes var:  0.1787646931836686
../_images/tutorials_combosciplex_Rdkit_embeddings_58_41.png 
A549_CHEMBL383824_3.0 (758, 5000)
Top 20 DEGs var:  0.8965285137419556
All genes var:  0.3831937702713043
../_images/tutorials_combosciplex_Rdkit_embeddings_58_43.png 
A549_CHEMBL483254+CHEMBL116438_3.0+3.0 (2244, 5000)
Top 20 DEGs var:  0.9502036607016588
All genes var:  0.26971633067733725
../_images/tutorials_combosciplex_Rdkit_embeddings_58_45.png 
A549_CHEMBL1213492+CHEMBL1421_3.0+3.0 (2421, 5000)
Top 20 DEGs var:  0.8425749877700159
All genes var:  0.2750751945668547
../_images/tutorials_combosciplex_Rdkit_embeddings_58_47.png 
A549_CHEMBL4297436+CHEMBL383824_3.0+3.0 (520, 5000)
Top 20 DEGs var:  0.8887265210694321
All genes var:  0.21741516085853863
../_images/tutorials_combosciplex_Rdkit_embeddings_58_49.png 
A549_CHEMBL356066+CHEMBL1421_3.0+3.0 (1231, 5000)
Top 20 DEGs var:  0.9019485977562606
All genes var:  0.19927103640546096
../_images/tutorials_combosciplex_Rdkit_embeddings_58_51.png 
A549_CHEMBL4297436_3.0 (2756, 5000)
Top 20 DEGs var:  0.7280733638487724
All genes var:  0.28689414242694167
../_images/tutorials_combosciplex_Rdkit_embeddings_58_53.png 
A549_CHEMBL1213492+CHEMBL116438_3.0+3.0 (2736, 5000)
Top 20 DEGs var:  0.8848140790799369
All genes var:  0.29790094480564633
../_images/tutorials_combosciplex_Rdkit_embeddings_58_55.png 
A549_CHEMBL1213492+CHEMBL601719_3.0+3.0 (2662, 5000)
Top 20 DEGs var:  0.8104838141770024
All genes var:  0.26790973907234417
../_images/tutorials_combosciplex_Rdkit_embeddings_58_57.png 
A549_46245047+CHEMBL491473_3.0+3.0 (3016, 5000)
Top 20 DEGs var:  0.8753626170879585
All genes var:  0.29425056709554254
../_images/tutorials_combosciplex_Rdkit_embeddings_58_59.png 
A549_CHEMBL1421_3.0 (2343, 5000)
Top 20 DEGs var:  0.9361548093504224
All genes var:  0.19595293836598193
../_images/tutorials_combosciplex_Rdkit_embeddings_58_61.png

Visualizing similarity between drug embeddings#

CPA learns an embedding for each covariate, and those can visualised to compare the similarity between perturbation (i.e. which perturbation have similar gene expression responses)

[28]:

cpa_api = cpa.ComPertAPI(adata, model,
                         de_genes_uns_key='rank_genes_groups_cov',
                         pert_category_key='cov_drug_dose',
                         control_group='CHEMBL504',
                         )

[29]:

cpa_plots = cpa.pl.CompertVisuals(cpa_api, fileprefix=None)

[30]:

cpa_api.num_measured_points['train']

[30]:

{'A549_46245047+CHEMBL491473_3.0+3.0': 2723,
 'A549_CHEMBL1213492+CHEMBL109480_3.0+3.0': 1175,
 'A549_CHEMBL1213492+CHEMBL116438_3.0+3.0': 2447,
 'A549_CHEMBL1213492+CHEMBL1200485_3.0+3.0': 2479,
 'A549_CHEMBL1213492+CHEMBL1421_3.0+3.0': 2158,
 'A549_CHEMBL1213492+CHEMBL257991_3.0+3.0': 2037,
 'A549_CHEMBL1213492+CHEMBL4297436_3.0+3.0': 2127,
 'A549_CHEMBL1213492+CHEMBL460499_3.0+3.0': 2425,
 'A549_CHEMBL1213492+CHEMBL601719_3.0+3.0': 2383,
 'A549_CHEMBL1213492_3.0': 1527,
 'A549_CHEMBL1421_3.0': 2133,
 'A549_CHEMBL356066+CHEMBL1421_3.0+3.0': 1124,
 'A549_CHEMBL356066+CHEMBL2170177_3.0+3.0': 2975,
 'A549_CHEMBL356066_3.0': 1679,
 'A549_CHEMBL383824+CHEMBL2354444_3.0+3.0': 421,
 'A549_CHEMBL383824_3.0': 690,
 'A549_CHEMBL4297436_3.0': 2491,
 'A549_CHEMBL483254+46245047_3.0+3.0': 1697,
 'A549_CHEMBL483254+CHEMBL116438_3.0+3.0': 2016,
 'A549_CHEMBL483254+CHEMBL1200485_3.0+3.0': 1810,
 'A549_CHEMBL483254+CHEMBL1421_3.0+3.0': 1762,
 'A549_CHEMBL483254+CHEMBL2170177_3.0+3.0': 1629,
 'A549_CHEMBL483254+CHEMBL257991_3.0+3.0': 1631,
 'A549_CHEMBL483254+CHEMBL601719_3.0+3.0': 1472,
 'A549_CHEMBL483254_3.0': 1413,
 'A549_CHEMBL491473+CHEMBL2170177_3.0+3.0': 1950,
 'A549_CHEMBL504_1.0': 1309}

[31]:

drug_adata = cpa_api.get_pert_embeddings()
drug_adata.shape

[31]:

(19, 64)

[32]:

cpa_plots.plot_latent_embeddings(drug_adata.X, kind='perturbations', titlename='Drugs')
../_images/tutorials_combosciplex_Rdkit_embeddings_65_0.png
../_images/tutorials_combosciplex_Rdkit_embeddings_65_1.png
On this page