CITE-seq analysis

With totalVI, we can produce a joint latent representation of cells, denoised data for both protein and RNA, integrate datasets, and compute differential expression of RNA and protein. Here we demonstrate this functionality with an integrated analysis of PBMC10k and PBMC5k, datasets of peripheral blood mononuclear cells publicly available from 10X Genomics subset to the 14 shared proteins between them. The same pipeline would generally be used to analyze a single CITE-seq dataset.

If you use totalVI, please consider citing:

  • Gayoso, Adam, et al. “Joint probabilistic modeling of paired transcriptome and proteome measurements in single cells.” bioRxiv (2020).

Open In Colab

[1]:
import sys

# If True, will install via pypi, else will install from source
stable = True
IN_COLAB = "google.colab" in sys.modules

if IN_COLAB and stable:
    !pip install --quiet scvi-tools[tutorials]
elif IN_COLAB and not stable:
    !pip install --quiet --upgrade jsonschema
    !pip install --quiet git+https://github.com/yoseflab/scvi-tools@master#egg=scvi-tools[tutorials]
     |████████████████████████████████| 61kB 3.7MB/s
ERROR: nbclient 0.5.0 has requirement jupyter-client>=6.1.5, but you'll have jupyter-client 5.3.5 which is incompatible.
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
    Preparing wheel metadata ... done
     |████████████████████████████████| 122kB 7.3MB/s
     |████████████████████████████████| 112kB 11.9MB/s
     |████████████████████████████████| 163kB 12.0MB/s
     |████████████████████████████████| 8.7MB 7.6MB/s
     |████████████████████████████████| 2.4MB 61.7MB/s
     |████████████████████████████████| 3.2MB 53.8MB/s
     |████████████████████████████████| 7.7MB 48.8MB/s
     |████████████████████████████████| 51kB 7.5MB/s
     |████████████████████████████████| 51kB 7.6MB/s
     |████████████████████████████████| 112kB 69.6MB/s
     |████████████████████████████████| 51kB 8.1MB/s
     |████████████████████████████████| 61kB 8.8MB/s
  Building wheel for scvi-tools (PEP 517) ... done
  Building wheel for loompy (setup.py) ... done
  Building wheel for sinfo (setup.py) ... done
  Building wheel for numpy-groupies (setup.py) ... done

Imports and data loading

[2]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

import scvi
import scanpy as sc

sc.set_figure_params(figsize=(4, 4))

This dataset was filtered as described in the totalVI manuscript (low quality cells, doublets, lowly expressed genes, etc.)

[3]:
adata = scvi.data.pbmcs_10x_cite_seq(run_setup_anndata=False)
adata.layers["counts"] = adata.X.copy()
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
adata.raw = adata
INFO      Downloading file at data/pbmc_10k_protein_v3.h5ad
Downloading...: 24938it [00:00, 99499.88it/s]
INFO      Downloading file at data/pbmc_5k_protein_v3.h5ad
Downloading...: 100%|██████████| 18295/18295.0 [00:00<00:00, 107363.43it/s]
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1094: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if not is_categorical(df_full[k]):
Observation names are not unique. To make them unique, call `.obs_names_make_unique`.
[4]:
sc.pp.highly_variable_genes(
    adata,
    n_top_genes=4000,
    flavor="seurat_v3",
    batch_key="batch",
    subset=True,
    layer="counts"
)
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1094: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if not is_categorical(df_full[k]):
/usr/local/lib/python3.6/dist-packages/scanpy/preprocessing/_highly_variable_genes.py:144: FutureWarning: Slicing a positional slice with .loc is not supported, and will raise TypeError in a future version.  Use .loc with labels or .iloc with positions instead.
  df.loc[: int(n_top_genes), 'highly_variable'] = True
Observation names are not unique. To make them unique, call `.obs_names_make_unique`.
Observation names are not unique. To make them unique, call `.obs_names_make_unique`.
[5]:
scvi.data.setup_anndata(
    adata,
    layer="counts",
    batch_key="batch",
    protein_expression_obsm_key="protein_expression"
)
INFO      Using batches from adata.obs["batch"]
INFO      No label_key inputted, assuming all cells have same label
INFO      Using data from adata.layers["counts"]
INFO      Computing library size prior per batch
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1094: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if not is_categorical(df_full[k]):
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1094: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if not is_categorical(df_full[k]):
INFO      Using protein expression from adata.obsm['protein_expression']
INFO      Using protein names from columns of adata.obsm['protein_expression']
INFO      Successfully registered anndata object containing 10849 cells, 4000
          vars, 2 batches, 1 labels, and 14 proteins. Also registered 0 extra
          categorical covariates and 0 extra continuous covariates.
INFO      Please do not further modify adata until model is trained.

Prepare and run model

[6]:
vae = scvi.model.TOTALVI(adata, use_cuda=True, latent_distribution="normal")
[7]:
vae.train()
INFO      Training for 400 epochs.
INFO      KL warmup for 8136.75 iterations
Training...: 100%|██████████| 400/400 [06:50<00:00,  1.03s/it]
INFO      Training time:  388 s. / 400 epochs
[8]:
plt.plot(vae.trainer.history["elbo_test_set"], label="test")
plt.title("Negative ELBO over training epochs")
plt.ylim(1200, 1400)
plt.legend()
[8]:
<matplotlib.legend.Legend at 0x7f08b0215630>
../../_images/user_guide_notebooks_totalVI_12_1.png

Analyze outputs

We use Scanpy for clustering and visualization after running totalVI. It’s also possible to save totalVI outputs for an R-based workflow. First, we store the totalVI outputs in the appropriate slots in AnnData.

[9]:
adata.obsm["X_totalVI"] = vae.get_latent_representation()

rna, protein = vae.get_normalized_expression(
    n_samples=25,
    return_mean=True,
    transform_batch=["PBMC10k", "PBMC5k"]
)

adata.layers["denoised_rna"], adata.obsm["denoised_protein"] = rna, protein

adata.obsm["protein_foreground_prob"] = vae.get_protein_foreground_probability(
    n_samples=25,
    return_mean=True,
    transform_batch=["PBMC10k", "PBMC5k"]
)
parsed_protein_names = [p.split("_")[0] for p in adata.obsm["protein_expression"].columns]
adata.obsm["protein_foreground_prob"].columns = parsed_protein_names

Now we can compute clusters and visualize the latent space.

[10]:
sc.pp.neighbors(adata, use_rep="X_totalVI")
sc.tl.umap(adata, min_dist=0.4)
sc.tl.leiden(adata, key_added="leiden_totalVI")
[11]:
sc.pl.umap(
    adata,
    color=["leiden_totalVI", "batch"],
    frameon=False,
    ncols=1,
)
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1192: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if is_string_dtype(df[key]) and not is_categorical(df[key])
... storing 'batch' as categorical
../../_images/user_guide_notebooks_totalVI_17_1.png

To visualize protein values on the umap, we make a temporary protein adata object. We have to copy over the umap from the original adata object.

[12]:
pro_adata = sc.AnnData(adata.obsm["protein_expression"].copy(), obs=adata.obs)
sc.pp.log1p(pro_adata)
# Keep log normalized data in raw
pro_adata.raw = pro_adata
pro_adata.X = adata.obsm["denoised_protein"]
# these are cleaner protein names -- "_TotalSeqB" removed
pro_adata.var["protein_names"] = parsed_protein_names
pro_adata.obsm["X_umap"] = adata.obsm["X_umap"]
pro_adata.obsm["X_totalVI"] = adata.obsm["X_totalVI"]

Observation names are not unique. To make them unique, call `.obs_names_make_unique`.
[13]:
names = adata.obsm["protein_foreground_prob"].columns
for p in names:
    pro_adata.obs["{}_fore_prob".format(p)] = adata.obsm["protein_foreground_prob"].loc[:, p]

Visualize denoised protein values

[14]:
sc.pl.umap(
    pro_adata,
    color=pro_adata.var_names,
    gene_symbols="protein_names",
    ncols=3,
    vmax="p99",
    use_raw=False,
    frameon=False,
    wspace=0.1
)
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1192: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if is_string_dtype(df[key]) and not is_categorical(df[key])
../../_images/user_guide_notebooks_totalVI_22_1.png

Visualize probability of foreground

Here we visualize the probability of foreground for each protein and cell (projected on UMAP). Some proteins are easier to disentangle than others. Some proteins end up being “all background”. For example, CD15 does not appear to be captured well, when looking at the denoised values above we see little localization in the monocytes.

Note

While the foreground probability could theoretically be used to identify cell populations, we recommend using the denoised protein expression, which accounts for the foreground/background probability, but preserves the dynamic range of the protein measurements. Consequently, the denoised values are on the same scale as the raw data and it may be desirable to take a transformation like log or square root.

By viewing the foreground probability, we can get a feel for the types of cells in our dataset. For example, it’s very easy to see a population of monocytes based on the CD14 foregroud probability.

[15]:
sc.pl.umap(
    pro_adata,
    color=["{}_fore_prob".format(p) for p in parsed_protein_names],
    ncols=3,
    color_map="cividis",
    frameon=False,
    wspace=0.1
)
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1192: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if is_string_dtype(df[key]) and not is_categorical(df[key])
../../_images/user_guide_notebooks_totalVI_26_1.png

Differential expression

Here we do a one-vs-all DE test, where each cluster is tested against all cells not in that cluster. The results for each of the one-vs-all tests is concatenated into one DataFrame object. Inividual tests can be sliced using the “comparison” column. Genes and proteins are included in the same DataFrame.

Important

We do not recommend using totalVI denoised values in other differential expression tools, as denoised values are a summary of a random quantity. The totalVI DE test takes into account the full uncertainty of the denoised quantities.

[16]:
de_df = vae.differential_expression(
        groupby="leiden_totalVI",
        delta=0.5,
        batch_correction=True
)
de_df.head(5)
DE...: 100%|██████████| 20/20 [00:37<00:00,  1.90s/it]
[16]:
proba_de proba_not_de bayes_factor scale1 scale2 lfc_mean lfc_median lfc_std lfc_min lfc_max raw_mean1 raw_mean2 non_zeros_proportion1 non_zeros_proportion2 raw_normalized_mean1 raw_normalized_mean2 is_de_fdr_0.05 comparison
AQP9 0.9892 0.0108 4.517349 0.000104 0.000005 8.522517 8.989746 2.848452 -5.698670 16.891201 0.258901 0.011430 0.204835 0.009097 1.062830 0.043297 True 0 vs Rest
NFATC2 0.9882 0.0118 4.427785 0.000002 0.000114 -6.671572 -7.075430 2.494170 -13.148601 8.218616 0.003077 0.127362 0.002637 0.112200 0.009379 1.159511 True 0 vs Rest
IFITM1 0.9872 0.0128 4.345427 0.000005 0.000728 -7.273490 -7.817169 2.652531 -13.630473 4.365607 0.004835 0.685911 0.004835 0.420457 0.023302 7.734947 True 0 vs Rest
C12ORF75 0.9872 0.0128 4.345427 0.000004 0.000543 -6.676181 -6.864652 2.767672 -14.146434 4.747614 0.009231 0.595638 0.009231 0.316772 0.032727 5.364836 True 0 vs Rest
MARC1 0.9870 0.0130 4.329720 0.000137 0.000005 8.172179 8.604403 2.845083 -6.426585 15.854793 0.321319 0.011197 0.247912 0.009214 1.248495 0.040035 True 0 vs Rest

Now we filter the results such that we retain features above a certain Bayes factor (which here is on the natural log scale) and genes with greater than 10% non-zero entries in the cluster of interest.

[17]:
filtered_pro = {}
filtered_rna = {}
cats = adata.obs.leiden_totalVI.cat.categories
for i, c in enumerate(cats):
    cid = "{} vs Rest".format(c)
    cell_type_df = de_df.loc[de_df.comparison == cid]
    cell_type_df = cell_type_df.sort_values("lfc_median", ascending=False)

    cell_type_df = cell_type_df[cell_type_df.lfc_median > 0]

    pro_rows = cell_type_df.index.str.contains('TotalSeqB')
    data_pro = cell_type_df.iloc[pro_rows]
    data_pro = data_pro[data_pro["bayes_factor"] > 0.7]

    data_rna = cell_type_df.iloc[~pro_rows]
    data_rna = data_rna[data_rna["bayes_factor"] > 3]
    data_rna = data_rna[data_rna["non_zeros_proportion1"] > 0.1]

    filtered_pro[c] = data_pro.index.tolist()[:3]
    filtered_rna[c] = data_rna.index.tolist()[:2]

We can also use general scanpy visualization functions

[18]:
sc.tl.dendrogram(adata, groupby="leiden_totalVI", use_rep="X_totalVI")
sc.tl.dendrogram(pro_adata, groupby="leiden_totalVI", use_rep="X_totalVI")
[19]:
sc.pl.dotplot(
    adata,
    filtered_rna,
    groupby="leiden_totalVI",
    dendrogram=True,
    standard_scale="var",
    swap_axes=True
)
../../_images/user_guide_notebooks_totalVI_35_0.png

Matrix plot displays totalVI denoised protein expression per leiden cluster.

[20]:
sc.pl.matrixplot(
    pro_adata,
    pro_adata.var["protein_names"],
    groupby="leiden_totalVI",
    gene_symbols="protein_names",
    dendrogram=True,
    swap_axes=True,
    use_raw=False, # use totalVI denoised
    cmap="Greens",
    standard_scale="var"
)
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1192: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if is_string_dtype(df[key]) and not is_categorical(df[key])
/usr/local/lib/python3.6/dist-packages/anndata/_core/anndata.py:1094: FutureWarning: is_categorical is deprecated and will be removed in a future version.  Use is_categorical_dtype instead
  if not is_categorical(df_full[k]):
../../_images/user_guide_notebooks_totalVI_37_1.png

This is a selection of some of the markers that turned up in the RNA DE test.

[21]:
sc.pl.umap(
    adata,
    color=[
           "leiden_totalVI",
           "IGHD",
           "FCER1A",
           "SCT",
           "GZMH",
           "NOG",
           "FOXP3",
           "CD8B",
           "C1QA",
           "SIGLEC1",
           "XCL2",
           "GZMK",
           ],
    legend_loc="on data",
    frameon=False,
    ncols=3,
    layer="denoised_rna",
    wspace=0.1
)
../../_images/user_guide_notebooks_totalVI_39_0.png