Similarities between Ciona Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?

Code ↔ Paper

1 match between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 1 match

[1] § Materials and Methods › Single-cell RNA sequencing analysis ↔ src/scvi/module/_vae.py, lines 750–893 · score 0.69 · linearly decoded variational, auto encoder, LDVAE, gene expression, space, mapped

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

Python · 893 lines · 34 KB · BSD-3-Clause · 1 match

from __future__ import annotations
import logging
import warnings
from typing import TYPE_CHECKING
import numpy as np
import torch
from torch.nn.functional import one_hot
from scvi import REGISTRY_KEYS, settings
from scvi.data._constants import ADATA_MINIFY_TYPE
from scvi.distributions._utils import _needs_cpu_detour
from scvi.module._constants import MODULE_KEYS
from scvi.module.base import (
BaseMinifiedModeModuleClass,
EmbeddingModuleMixin,
LossOutput,
auto_move_data,
)
from scvi.utils import unsupported_if_adata_minified
if TYPE_CHECKING:
from collections.abc import Callable
from typing import Literal
from torch.distributions import Distribution
logger = logging.getLogger(__name__)
class VAE(EmbeddingModuleMixin, BaseMinifiedModeModuleClass):
"""Variational auto-encoder :cite:p:`Lopez18`.
Parameters
----------
n_input
Number of input features.
n_batch
Number of batches. If ``0``, no batch correction is performed.
n_labels
Number of labels.
n_hidden
Number of nodes per hidden layer. Passed into :class:`~scvi.nn.Encoder` and
:class:`~scvi.nn.DecoderSCVI`.
n_latent
Dimensionality of the latent space.
n_layers
Number of hidden layers. Passed into :class:`~scvi.nn.Encoder` and
:class:`~scvi.nn.DecoderSCVI`.
n_continuous_cov
Number of continuous covariates.
n_cats_per_cov
A list of integers containing the number of categories for each categorical covariate.
dropout_rate
Dropout rate. Passed into :class:`~scvi.nn.Encoder` but not :class:`~scvi.nn.DecoderSCVI`.
dispersion
Flexibility of the dispersion parameter when ``gene_likelihood`` is either ``"nb"`` or
``"zinb"``. One of the following:
* ``"gene"``: parameter is constant per gene across cells.
* ``"gene-batch"``: parameter is constant per gene per batch.
* ``"gene-label"``: parameter is constant per gene per label.
* ``"gene-cell"``: parameter is constant per gene per cell.
log_variational
If ``True``, use :func:`~torch.log1p` on input data before encoding for numerical stability
(not normalization).
gene_likelihood
Distribution to use for reconstruction in the generative process. One of the following:
* ``"nb"``: :class:`~scvi.distributions.NegativeBinomial`.
* ``"zinb"``: :class:`~scvi.distributions.ZeroInflatedNegativeBinomial`.
* ``"poisson"``: :class:`~scvi.distributions.Poisson`.
* ``"normal"``: :class:`~torch.distributions.normal.Normal`.
latent_distribution
Distribution to use for the latent space. One of the following:
* ``"normal"``: isotropic normal.
* ``"ln"``: logistic normal with normal params N(0, 1).
encode_covariates
If ``True``, covariates are concatenated to gene expression prior to passing through
the encoder(s). Else, only the gene expression is used.
deeply_inject_covariates
If ``True`` and ``n_layers > 1``, covariates are concatenated to the outputs of hidden
layers in the encoder(s) (if ``encoder_covariates`` is ``True``) and the decoder prior to
passing through the next layer.
batch_representation
Method for encoding batch information. One of the following:
* ``"one-hot"``: represent batches with one-hot encodings.
* ``"embedding"``: represent batches with continuously-valued embeddings using
:class:`~scvi.nn.Embedding`.
Note that batch representations are only passed into the encoder(s) if
``encode_covariates`` is ``True``.
use_batch_norm
Specifies where to use :class:`~torch.nn.BatchNorm1d` in the model. One of the following:
* ``"none"``: don't use batch norm in either encoder(s) or decoder.
* ``"encoder"``: use batch norm only in the encoder(s).
* ``"decoder"``: use batch norm only in the decoder.
* ``"both"``: use batch norm in both encoder(s) and decoder.
Note: if ``use_layer_norm`` is also specified, both will be applied (first
:class:`~torch.nn.BatchNorm1d`, then :class:`~torch.nn.LayerNorm`).
use_layer_norm
Specifies where to use :class:`~torch.nn.LayerNorm` in the model. One of the following:
* ``"none"``: don't use layer norm in either encoder(s) or decoder.
* ``"encoder"``: use layer norm only in the encoder(s).
* ``"decoder"``: use layer norm only in the decoder.
* ``"both"``: use layer norm in both encoder(s) and decoder.
Note: if ``use_batch_norm`` is also specified, both will be applied (first
:class:`~torch.nn.BatchNorm1d`, then :class:`~torch.nn.LayerNorm`).
use_size_factor_key
If ``True``, use the :attr:`~anndata.AnnData.obs` column as defined by the
``size_factor_key`` parameter in the model's ``setup_anndata`` method as the scaling
factor in the mean of the conditional distribution. Takes priority over
``use_observed_lib_size``.
use_observed_lib_size
If ``True``, use the observed library size for RNA as the scaling factor in the mean of the
conditional distribution.
extra_payload_autotune
If ``True``, will return extra matrices in the loss output to be used during autotune
library_log_means
:class:`~numpy.ndarray` of shape ``(1, n_batch)`` of means of the log library sizes that
parameterize the prior on library size if ``use_size_factor_key`` is ``False`` and
``use_observed_lib_size`` is ``False``.
library_log_vars
:class:`~numpy.ndarray` of shape ``(1, n_batch)`` of variances of the log library sizes
that parameterize the prior on library size if ``use_size_factor_key`` is ``False`` and
``use_observed_lib_size`` is ``False``.
var_activation
Callable used to ensure positivity of the variance of the variational distribution. Passed
into :class:`~scvi.nn.Encoder`. Defaults to :func:`~torch.exp`.
extra_encoder_kwargs
Additional keyword arguments passed into :class:`~scvi.nn.Encoder`.
extra_decoder_kwargs
Additional keyword arguments passed into :class:`~scvi.nn.DecoderSCVI`.
batch_embedding_kwargs
Keyword arguments passed into :class:`~scvi.nn.Embedding` if ``batch_representation`` is
set to ``"embedding"``.
"""
def __init__(
self,
n_input: int,
n_batch: int = 0,
n_labels: int = 0,
n_hidden: int = 128,
n_latent: int = 10,
n_layers: int = 1,
n_continuous_cov: int = 0,
n_cats_per_cov: list[int] | None = None,
dropout_rate: float = 0.1,
dispersion: Literal["gene", "gene-batch", "gene-label", "gene-cell"] = "gene",
log_variational: bool = True,
gene_likelihood: Literal["zinb", "nb", "poisson", "normal"] = "zinb",
latent_distribution: Literal["normal", "ln"] = "normal",
encode_covariates: bool = False,
deeply_inject_covariates: bool = True,
batch_representation: Literal["one-hot", "embedding"] = "one-hot",
use_batch_norm: Literal["encoder", "decoder", "none", "both"] = "both",
use_layer_norm: Literal["encoder", "decoder", "none", "both"] = "none",
use_size_factor_key: bool = False,
use_observed_lib_size: bool = True,
extra_payload_autotune: bool = False,
library_log_means: np.ndarray | None = None,
library_log_vars: np.ndarray | None = None,
var_activation: Callable[[torch.Tensor], torch.Tensor] = None,
extra_encoder_kwargs: dict | None = None,
extra_decoder_kwargs: dict | None = None,
batch_embedding_kwargs: dict | None = None,
):
from scvi.nn import DecoderSCVI, Encoder
super().__init__()
self.dispersion = dispersion
self.n_latent = n_latent
self.log_variational = log_variational
self.gene_likelihood = gene_likelihood
self.n_batch = n_batch
self.n_input = n_input
self.n_labels = n_labels
self.n_hidden = n_hidden
self.n_layers = n_layers
self.latent_distribution = latent_distribution
self.encode_covariates = encode_covariates
self.use_size_factor_key = use_size_factor_key
self.use_observed_lib_size = use_size_factor_key or use_observed_lib_size
self.extra_payload_autotune = extra_payload_autotune
if not self.use_observed_lib_size:
if library_log_means is None or library_log_vars is None:
raise ValueError(
"If not using observed_lib_size, "
"must provide library_log_means and library_log_vars."
)
self.register_buffer("library_log_means", torch.from_numpy(library_log_means).float())
self.register_buffer("library_log_vars", torch.from_numpy(library_log_vars).float())
if self.dispersion == "gene":
self.px_r = torch.nn.Parameter(torch.randn(n_input))
elif self.dispersion == "gene-batch":
self.px_r = torch.nn.Parameter(torch.randn(n_input, n_batch))
elif self.dispersion == "gene-label":
self.px_r = torch.nn.Parameter(torch.randn(n_input, n_labels))
elif self.dispersion == "gene-cell":
pass
else:
raise ValueError(
"`dispersion` must be one of 'gene', 'gene-batch', 'gene-label', 'gene-cell'."
)
self.batch_representation = batch_representation
if self.batch_representation == "embedding":
self.init_embedding(REGISTRY_KEYS.BATCH_KEY, n_batch, **(batch_embedding_kwargs or {}))
batch_dim = self.get_embedding(REGISTRY_KEYS.BATCH_KEY).embedding_dim
elif self.batch_representation != "one-hot":
raise ValueError("`batch_representation` must be one of 'one-hot', 'embedding'.")
use_batch_norm_encoder = use_batch_norm == "encoder" or use_batch_norm == "both"
use_batch_norm_decoder = use_batch_norm == "decoder" or use_batch_norm == "both"
use_layer_norm_encoder = use_layer_norm == "encoder" or use_layer_norm == "both"
use_layer_norm_decoder = use_layer_norm == "decoder" or use_layer_norm == "both"
n_input_encoder = n_input + n_continuous_cov * encode_covariates
if self.batch_representation == "embedding":
n_input_encoder += batch_dim * encode_covariates
cat_list = list([] if n_cats_per_cov is None else n_cats_per_cov)
else:
cat_list = [n_batch] + list([] if n_cats_per_cov is None else n_cats_per_cov)
encoder_cat_list = cat_list if encode_covariates else None
_extra_encoder_kwargs = extra_encoder_kwargs or {}
self.z_encoder = Encoder(
n_input_encoder,
n_latent,
n_cat_list=encoder_cat_list,
n_layers=n_layers,
n_hidden=n_hidden,
dropout_rate=dropout_rate,
distribution=latent_distribution,
inject_covariates=deeply_inject_covariates,
use_batch_norm=use_batch_norm_encoder,
use_layer_norm=use_layer_norm_encoder,
var_activation=var_activation,
return_dist=True,
**_extra_encoder_kwargs,
)
# l encoder goes from n_input-dimensional data to 1-d library size
self.l_encoder = Encoder(
n_input_encoder,
1,
n_layers=1,
n_cat_list=encoder_cat_list,
n_hidden=n_hidden,
dropout_rate=dropout_rate,
inject_covariates=deeply_inject_covariates,
use_batch_norm=use_batch_norm_encoder,
use_layer_norm=use_layer_norm_encoder,
var_activation=var_activation,
return_dist=True,
**_extra_encoder_kwargs,
)
n_input_decoder = n_latent + n_continuous_cov
if self.batch_representation == "embedding":
n_input_decoder += batch_dim
_extra_decoder_kwargs = extra_decoder_kwargs or {}
self.decoder = DecoderSCVI(
n_input_decoder,
n_input,
n_cat_list=cat_list,
n_layers=n_layers,
n_hidden=n_hidden,
inject_covariates=deeply_inject_covariates,
use_batch_norm=use_batch_norm_decoder,
use_layer_norm=use_layer_norm_decoder,
scale_activation="softplus" if use_size_factor_key else "softmax",
**_extra_decoder_kwargs,
)
def _get_inference_input(
self,
tensors: dict[str, torch.Tensor | None],
full_forward_pass: bool = False,
) -> dict[str, torch.Tensor | None]:
"""Get input tensors for the inference process."""
if full_forward_pass or self.minified_data_type is None:
loader = "full_data"
elif self.minified_data_type in [
ADATA_MINIFY_TYPE.LATENT_POSTERIOR,
ADATA_MINIFY_TYPE.LATENT_POSTERIOR_WITH_COUNTS,
]:
loader = "minified_data"
else:
raise NotImplementedError(f"Unknown minified-data type: {self.minified_data_type}")
if loader == "full_data":
return {
MODULE_KEYS.X_KEY: tensors[REGISTRY_KEYS.X_KEY],
MODULE_KEYS.BATCH_INDEX_KEY: tensors[REGISTRY_KEYS.BATCH_KEY],
MODULE_KEYS.CONT_COVS_KEY: tensors.get(REGISTRY_KEYS.CONT_COVS_KEY, None),
MODULE_KEYS.CAT_COVS_KEY: tensors.get(REGISTRY_KEYS.CAT_COVS_KEY, None),
}
else:
return {
MODULE_KEYS.QZM_KEY: tensors[REGISTRY_KEYS.LATENT_QZM_KEY],
MODULE_KEYS.QZV_KEY: tensors[REGISTRY_KEYS.LATENT_QZV_KEY],
REGISTRY_KEYS.OBSERVED_LIB_SIZE: tensors[REGISTRY_KEYS.OBSERVED_LIB_SIZE],
}
def _get_generative_input(
self,
tensors: dict[str, torch.Tensor],
inference_outputs: dict[str, torch.Tensor | Distribution | None],
) -> dict[str, torch.Tensor | None]:
"""Get input tensors for the generative process."""
size_factor = tensors.get(REGISTRY_KEYS.SIZE_FACTOR_KEY, None)
if size_factor is not None:
size_factor = torch.log(size_factor)
return {
MODULE_KEYS.Z_KEY: inference_outputs[MODULE_KEYS.Z_KEY],
MODULE_KEYS.LIBRARY_KEY: inference_outputs[MODULE_KEYS.LIBRARY_KEY],
MODULE_KEYS.BATCH_INDEX_KEY: tensors[REGISTRY_KEYS.BATCH_KEY],
MODULE_KEYS.Y_KEY: tensors[REGISTRY_KEYS.LABELS_KEY],
MODULE_KEYS.CONT_COVS_KEY: tensors.get(REGISTRY_KEYS.CONT_COVS_KEY, None),
MODULE_KEYS.CAT_COVS_KEY: tensors.get(REGISTRY_KEYS.CAT_COVS_KEY, None),
MODULE_KEYS.SIZE_FACTOR_KEY: size_factor,
}
def _compute_local_library_params(
self,
batch_index: torch.Tensor,
) -> tuple[torch.Tensor, torch.Tensor]:
"""Computes local library parameters.
Compute two tensors of shape (batch_index.shape[0], 1) where each
element corresponds to the mean and variances, respectively, of the
log library sizes in the batch the cell corresponds to.
"""
from torch.nn.functional import linear
n_batch = self.library_log_means.shape[1]
local_library_log_means = linear(
one_hot(batch_index.squeeze(-1), n_batch).float(), self.library_log_means
)
local_library_log_vars = linear(
one_hot(batch_index.squeeze(-1), n_batch).float(), self.library_log_vars
)
return local_library_log_means, local_library_log_vars
@auto_move_data
def _regular_inference(
self,
x: torch.Tensor,
batch_index: torch.Tensor,
cont_covs: torch.Tensor | None = None,
cat_covs: torch.Tensor | None = None,
n_samples: int = 1,
) -> dict[str, torch.Tensor | Distribution | None]:
"""Run the regular inference process."""
x_ = x
if self.use_observed_lib_size:
library = torch.log(x.sum(1)).unsqueeze(1)
if self.log_variational:
x_ = torch.log1p(x_)
if cont_covs is not None and self.encode_covariates:
encoder_input = torch.cat((x_, cont_covs), dim=-1)
else:
encoder_input = x_
if cat_covs is not None and self.encode_covariates:
categorical_input = torch.split(cat_covs, 1, dim=1)
else:
categorical_input = ()
if self.batch_representation == "embedding" and self.encode_covariates:
batch_rep = self.compute_embedding(REGISTRY_KEYS.BATCH_KEY, batch_index)
encoder_input = torch.cat([encoder_input, batch_rep], dim=-1)
qz, z = self.z_encoder(encoder_input, *categorical_input)
else:
qz, z = self.z_encoder(encoder_input, batch_index, *categorical_input)
ql = None
if not self.use_observed_lib_size:
if self.batch_representation == "embedding":
ql, library_encoded = self.l_encoder(encoder_input, *categorical_input)
else:
ql, library_encoded = self.l_encoder(
encoder_input, batch_index, *categorical_input
)
library = library_encoded
if n_samples > 1:
untran_z = qz.sample((n_samples,))
z = self.z_encoder.z_transformation(untran_z)
if self.use_observed_lib_size:
library = library.unsqueeze(0).expand(
(n_samples, library.size(0), library.size(1))
)
else:
library = ql.sample((n_samples,))
return {
MODULE_KEYS.Z_KEY: z,
MODULE_KEYS.QZ_KEY: qz,
MODULE_KEYS.QL_KEY: ql,
MODULE_KEYS.LIBRARY_KEY: library,
}
@auto_move_data
def _cached_inference(
self,
qzm: torch.Tensor,
qzv: torch.Tensor,
observed_lib_size: torch.Tensor,
n_samples: int = 1,
) -> dict[str, torch.Tensor | None]:
"""Run the cached inference process."""
from torch.distributions import Normal
qz = Normal(qzm, qzv.sqrt())
# use dist.sample() rather than rsample because we aren't optimizing the z here
untran_z = qz.sample() if n_samples == 1 else qz.sample((n_samples,))
z = self.z_encoder.z_transformation(untran_z)
library = torch.log(observed_lib_size)
if n_samples > 1:
library = library.unsqueeze(0).expand((n_samples, library.size(0), library.size(1)))
return {
MODULE_KEYS.Z_KEY: z,
MODULE_KEYS.QZ_KEY: qz,
MODULE_KEYS.QL_KEY: None,
MODULE_KEYS.LIBRARY_KEY: library,
}
@auto_move_data
def generative(
self,
z: torch.Tensor,
library: torch.Tensor,
batch_index: torch.Tensor,
cont_covs: torch.Tensor | None = None,
cat_covs: torch.Tensor | None = None,
size_factor: torch.Tensor | None = None,
y: torch.Tensor | None = None,
transform_batch: torch.Tensor | None = None,
) -> dict[str, Distribution | None]:
"""Run the generative process."""
from torch.nn.functional import linear
from scvi.distributions import (
NegativeBinomial,
Normal,
Poisson,
ZeroInflatedNegativeBinomial,
)
# TODO: refactor forward function to not rely on y
# Likelihood distribution
if cont_covs is None:
decoder_input = z
elif z.dim() != cont_covs.dim():
decoder_input = torch.cat(
[z, cont_covs.unsqueeze(0).expand(z.size(0), -1, -1)], dim=-1
)
else:
decoder_input = torch.cat([z, cont_covs], dim=-1)
if cat_covs is not None:
categorical_input = torch.split(cat_covs, 1, dim=1)
else:
categorical_input = ()
if transform_batch is not None:
batch_index = torch.ones_like(batch_index) * transform_batch
if not self.use_size_factor_key:
size_factor = library
if self.batch_representation == "embedding":
batch_rep = self.compute_embedding(REGISTRY_KEYS.BATCH_KEY, batch_index)
decoder_input = torch.cat([decoder_input, batch_rep], dim=-1)
px_scale, px_r, px_rate, px_dropout = self.decoder(
self.dispersion,
decoder_input,
size_factor,
*categorical_input,
y,
)
else:
px_scale, px_r, px_rate, px_dropout = self.decoder(
self.dispersion,
decoder_input,
size_factor,
batch_index,
*categorical_input,
y,
)
if self.dispersion == "gene-label":
px_r = linear(
one_hot(y.squeeze(-1), self.n_labels).float(), self.px_r
) # px_r gets transposed - last dimension is nb genes
elif self.dispersion == "gene-batch":
px_r = linear(one_hot(batch_index.squeeze(-1), self.n_batch).float(), self.px_r)
elif self.dispersion == "gene":
px_r = self.px_r
px_r = torch.exp(px_r)
if self.gene_likelihood == "zinb":
px = ZeroInflatedNegativeBinomial(
mu=px_rate,
theta=px_r,
zi_logits=px_dropout,
scale=px_scale,
)
elif self.gene_likelihood == "nb":
px = NegativeBinomial(mu=px_rate, theta=px_r, scale=px_scale)
elif self.gene_likelihood == "poisson":
px = Poisson(rate=px_rate, scale=px_scale)
elif self.gene_likelihood == "normal":
px = Normal(px_rate, px_r, normal_mu=px_scale)
# Priors
if self.use_observed_lib_size:
pl = None
else:
(
local_library_log_means,
local_library_log_vars,
) = self._compute_local_library_params(batch_index)
pl = Normal(local_library_log_means, local_library_log_vars.sqrt())
pz = Normal(torch.zeros_like(z), torch.ones_like(z))
return {
MODULE_KEYS.PX_KEY: px,
MODULE_KEYS.PL_KEY: pl,
MODULE_KEYS.PZ_KEY: pz,
}
@unsupported_if_adata_minified
def loss(
self,
tensors: dict[str, torch.Tensor],
inference_outputs: dict[str, torch.Tensor | Distribution | None],
generative_outputs: dict[str, Distribution | None],
kl_weight: torch.Tensor | float = 1.0,
) -> LossOutput:
"""Compute the loss."""
from torch.distributions import kl_divergence
x = tensors[REGISTRY_KEYS.X_KEY]
kl_divergence_z = kl_divergence(
inference_outputs[MODULE_KEYS.QZ_KEY], generative_outputs[MODULE_KEYS.PZ_KEY]
).sum(dim=-1)
if not self.use_observed_lib_size:
kl_divergence_l = kl_divergence(
inference_outputs[MODULE_KEYS.QL_KEY], generative_outputs[MODULE_KEYS.PL_KEY]
).sum(dim=1)
else:
kl_divergence_l = torch.zeros_like(kl_divergence_z)
reconst_loss = -generative_outputs[MODULE_KEYS.PX_KEY].log_prob(x).sum(-1)
kl_local_for_warmup = kl_divergence_z
kl_local_no_warmup = kl_divergence_l
weighted_kl_local = kl_weight * kl_local_for_warmup + kl_local_no_warmup
loss = torch.mean(reconst_loss + weighted_kl_local)
# a payload to be used during autotune
if self.extra_payload_autotune:
extra_metrics_payload = {
"z": inference_outputs["z"],
"batch": tensors[REGISTRY_KEYS.BATCH_KEY],
"labels": tensors[REGISTRY_KEYS.LABELS_KEY],
}
else:
extra_metrics_payload = {}
return LossOutput(
loss=loss,
reconstruction_loss=reconst_loss,
kl_local={
MODULE_KEYS.KL_L_KEY: kl_divergence_l,
MODULE_KEYS.KL_Z_KEY: kl_divergence_z,
},
extra_metrics=extra_metrics_payload,
)
@torch.inference_mode()
def sample(
self,
tensors: dict[str, torch.Tensor],
n_samples: int = 1,
max_poisson_rate: float = 1e8,
generative_kwargs: dict | None = None,
) -> torch.Tensor:
r"""Generate predictive samples from the posterior predictive distribution.
The posterior predictive distribution is denoted as :math:`p(\hat{x} \mid x)`, where
:math:`x` is the input data and :math:`\hat{x}` is the sampled data.
We sample from this distribution by first sampling ``n_samples`` times from the posterior
distribution :math:`q(z \mid x)` for a given observation, and then sampling from the
likelihood :math:`p(\hat{x} \mid z)` for each of these.
Parameters
----------
tensors
Dictionary of tensors passed into ``VAE.forward``.
n_samples
Number of Monte Carlo samples to draw from the distribution for each observation.
max_poisson_rate
The maximum value to which to clip the ``rate`` parameter of
:class:`~scvi.distributions.Poisson`. Avoids numerical sampling issues when the
parameter is very large due to the variance of the distribution.
generative_kwargs
Keyword args for ``generative()`` in fwd pass
Returns
-------
Tensor on CPU with shape ``(n_obs, n_vars)`` if ``n_samples == 1``, else
``(n_obs, n_vars,)``.
"""
from scvi.distributions import Poisson
inference_kwargs = {"n_samples": n_samples}
_, generative_outputs = self.forward(
tensors,
inference_kwargs=inference_kwargs,
generative_kwargs=generative_kwargs,
compute_loss=False,
)
dist = generative_outputs[MODULE_KEYS.PX_KEY]
if self.gene_likelihood == "poisson":
on_mps = self.device.type == "mps"
rate = torch.clamp(dist.rate, max=max_poisson_rate)
if _needs_cpu_detour(on_mps, torch.poisson):
rate = rate.to("cpu")
dist = Poisson(rate)
# (n_obs, n_vars) if n_samples == 1, else (n_samples, n_obs, n_vars)
samples = dist.sample()
# (n_samples, n_obs, n_vars) -> (n_obs, n_vars, n_samples)
samples = torch.permute(samples, (1, 2, 0)) if n_samples > 1 else samples
return samples.cpu()
@torch.inference_mode()
@auto_move_data
def marginal_ll(
self,
tensors: dict[str, torch.Tensor],
n_mc_samples: int,
return_mean: bool = False,
n_mc_samples_per_pass: int = 1,
):
"""Compute the marginal log-likelihood of the data under the model.
Parameters
----------
tensors
Dictionary of tensors passed into ``VAE.forward``.
n_mc_samples
Number of Monte Carlo samples to use for the estimation of the marginal log-likelihood.
return_mean
Whether to return the mean of marginal likelihoods over cells.
n_mc_samples_per_pass
Number of Monte Carlo samples to use per pass. This is useful to avoid memory issues.
"""
from torch import logsumexp
from torch.distributions import Normal
batch_index = tensors[REGISTRY_KEYS.BATCH_KEY]
to_sum = []
if n_mc_samples_per_pass > n_mc_samples:
warnings.warn(
"Number of chunks is larger than the total number of samples, setting it to the "
"number of samples",
RuntimeWarning,
stacklevel=settings.warnings_stacklevel,
)
n_mc_samples_per_pass = n_mc_samples
n_passes = int(np.ceil(n_mc_samples / n_mc_samples_per_pass))
for _ in range(n_passes):
# Distribution parameters and sampled variables
inference_outputs, _, losses = self.forward(
tensors,
inference_kwargs={"n_samples": n_mc_samples_per_pass},
get_inference_input_kwargs={"full_forward_pass": True},
)
qz = inference_outputs[MODULE_KEYS.QZ_KEY]
ql = inference_outputs[MODULE_KEYS.QL_KEY]
z = inference_outputs[MODULE_KEYS.Z_KEY]
library = inference_outputs[MODULE_KEYS.LIBRARY_KEY]
# Reconstruction Loss
reconst_loss = losses.dict_sum(losses.reconstruction_loss)
# Log-probabilities
p_z = (
Normal(torch.zeros_like(qz.loc), torch.ones_like(qz.scale)).log_prob(z).sum(dim=-1)
)
p_x_zl = -reconst_loss
q_z_x = qz.log_prob(z).sum(dim=-1)
log_prob_sum = p_z + p_x_zl - q_z_x
if not self.use_observed_lib_size:
(
local_library_log_means,
local_library_log_vars,
) = self._compute_local_library_params(batch_index)
p_l = (
Normal(local_library_log_means, local_library_log_vars.sqrt())
.log_prob(library)
.sum(dim=-1)
)
q_l_x = ql.log_prob(library).sum(dim=-1)
log_prob_sum += p_l - q_l_x
if n_mc_samples_per_pass == 1:
log_prob_sum = log_prob_sum.unsqueeze(0)
to_sum.append(log_prob_sum)
to_sum = torch.cat(to_sum, dim=0)
batch_log_lkl = logsumexp(to_sum, dim=0) - np.log(n_mc_samples)
if return_mean:
batch_log_lkl = torch.mean(batch_log_lkl).item()
else:
batch_log_lkl = batch_log_lkl.cpu()
return batch_log_lkl
class LDVAE(VAE):
"""Linear-decoded Variational auto-encoder model.
Implementation of :cite:p:`Svensson20`.
This model uses a linear decoder, directly mapping the latent representation
to gene expression levels. It still uses a deep neural network to encode
the latent representation.
Compared to standard VAE, this model is less powerful but can be used to
inspect which genes contribute to variation in the dataset. It may also be used
for all scVI tasks, like differential expression, batch correction, imputation, etc.
However, batch correction may be less powerful as it assumes a linear model.
Parameters
----------
n_input
Number of input genes
n_batch
Number of batches
n_labels
Number of labels
n_hidden
Number of nodes per hidden layer (for encoder)
n_latent
Dimensionality of the latent space
n_layers_encoder
Number of hidden layers used for encoder NNs
dropout_rate
Dropout rate for neural networks
dispersion
One of the following
* ``'gene'`` - dispersion parameter of NB is constant per gene across cells
* ``'gene-batch'`` - dispersion can differ between different batches
* ``'gene-label'`` - dispersion can differ between different labels
* ``'gene-cell'`` - dispersion can differ for every gene in every cell
log_variational
Log(data+1) prior to encoding for numerical stability. Not normalization.
gene_likelihood
One of
* ``'nb'`` - Negative binomial distribution
* ``'zinb'`` - Zero-inflated negative binomial distribution
* ``'poisson'`` - Poisson distribution
use_batch_norm
Bool whether to use batch norm in decoder
bias
Bool whether to have bias term in linear decoder
latent_distribution
One of
* ``'normal'`` - Isotropic normal
* ``'ln'`` - Logistic normal with normal params N(0, 1)
use_observed_lib_size
Use observed library size for RNA as scaling factor in mean of conditional distribution.
**kwargs
"""
def __init__(
self,
n_input: int,
n_batch: int = 0,
n_labels: int = 0,
n_hidden: int = 128,
n_latent: int = 10,
n_layers_encoder: int = 1,
dropout_rate: float = 0.1,
dispersion: str = "gene",
log_variational: bool = True,
gene_likelihood: str = "nb",
use_batch_norm: bool = True,
bias: bool = False,
latent_distribution: str = "normal",
use_observed_lib_size: bool = False,
**kwargs,
):
from scvi.nn import Encoder, LinearDecoderSCVI
super().__init__(
n_input=n_input,
n_batch=n_batch,
n_labels=n_labels,
n_hidden=n_hidden,
n_latent=n_latent,
n_layers=n_layers_encoder,
dropout_rate=dropout_rate,
dispersion=dispersion,
log_variational=log_variational,
gene_likelihood=gene_likelihood,
latent_distribution=latent_distribution,
use_observed_lib_size=use_observed_lib_size,
**kwargs,
)
self.use_batch_norm = use_batch_norm
self.z_encoder = Encoder(
n_input,
n_latent,
n_layers=n_layers_encoder,
n_hidden=n_hidden,
dropout_rate=dropout_rate,
distribution=latent_distribution,
use_batch_norm=True,
use_layer_norm=False,
return_dist=True,
)
self.l_encoder = Encoder(
n_input,
1,
n_layers=1,
n_hidden=n_hidden,
dropout_rate=dropout_rate,
use_batch_norm=True,
use_layer_norm=False,
return_dist=True,
)
self.decoder = LinearDecoderSCVI(
n_latent,
n_input,
n_cat_list=[n_batch],
use_batch_norm=use_batch_norm,
use_layer_norm=False,
bias=bias,
)
@torch.inference_mode()
def get_loadings(self) -> np.ndarray:
"""Extract per-gene weights in the linear decoder."""
# This is BW, where B is diag(b) batch norm, W is the weight matrix
if self.use_batch_norm is True:
w = self.decoder.factor_regressor.fc_layers[0][0].weight
bn = self.decoder.factor_regressor.fc_layers[0][1]
sigma = torch.sqrt(bn.running_var + bn.eps)
gamma = bn.weight
b = gamma / sigma
b_identity = torch.diag(b)
loadings = torch.matmul(b_identity, w)
else:
loadings = self.decoder.factor_regressor.fc_layers[0][0].weight
loadings = loadings.detach().cpu().numpy()
if self.n_batch > 1:
loadings = loadings[:, : -self.n_batch]
return loadings

_vae.py at commit fa9470f, under BSD-3-Clause · at the source

Overview

Authors: Matthew J. Kourakis¹, Yishen Miao², Erin D. Newman-Smith¹, Kerrianne Ryan³, William C. Smith^1,2

ORCID iDs: Yishen Miao

Institutions: University of California, Santa Barbara (United States); Dalhousie University (Canada)

Journal: eNeuro, volume 13, issue 8, pages ENEURO.0362-25.2026

Dates: received 30 September 2025; accepted 18 July 2026; published online 19 August 2026; in print August 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.1523/eneuro.0362-25.2026 · PMID 42637546 · PMCID PMC13509005 · OpenAlex W7204162582

Open access: gold, a free copy (OpenAlex)

Status: code verified

Methods: Evoked potentials, Statistics, Single-unit activity, calcium imaging

Keywords: cerebellum, chordate, Ciona, evolution, hindbrain

MeSH: Biological Evolution*, Cerebellum*, Ciona*, Ganglia, Spinal*, Rhombencephalon*, Animals, Ciona intestinalis (* major topic)

Journal subjects: Theory/New Concepts, Development

Topic: Developmental Biology and Gene Regulation (Molecular Biology, Biochemistry, Genetics and Molecular Biology), according to OpenAlex

Funding: National Instutes of Health (NS127106); U.S. Department of Energy (DOE) (DE-SC0021978)

Citations: not cited yet (Europe PMC); 107 references in the paper

Abstract

The cerebellum, a major structure of the vertebrate central nervous system, forms the dorsal region of the hindbrain. While evidence from connectomics, neuron classification, and gene expression analyses suggests that the major CNS divisions of forebrain, midbrain, hindbrain, and spinal cord predate the vertebrate/tunicate divergence, it is unknown if the subdivision of the hindbrain into a cerebellum, or cerebellum precursor structure, does as well. In fact, conflicting views exist over whether even the most basal vertebrates, the cyclostomes, have a cerebellum homolog. We report here that the motor ganglion (MG), a structure that has been homologized to the vertebrate hindbrain in the chordate Ciona (a tunicate), has functionally distinct dorsal and ventral domains, similar to the generalized hindbrain of vertebrates. In Ciona, the dorsal MG functions as an ascending relay and processing center for peripheral sensory neurons, while the ventral domain processes and relays descending sensorimotor commands. These differences are reflected in distinct neuron classes present in the two domains, as well as in gene expression. While the dorsal MG of Ciona and HB of vertebrates differ in function and complexity, they share a number of similarities. Thus, while it is likely that the cerebellum proper evolved within vertebrates, our findings support the hypothesis that a simpler, dorsally positioned antecedent was present before the tunicate–vertebrate divergence.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 1 match between paragraphs and lines of code.

scverse/scvi-tools

License: BSD-3-Clause

State: the link answers, verified on 27 September 2026

Evidence: files inventoried

Commit: fa9470f9b59f47a1df1677eda44161ac14845db3, 24 September 2026

Languages: Python (342), JavaScript (1)

Size: 559 files, 343 scripts

Software Heritage: archived

Found in: the Zenodo archive record

Holds: README, license file, environment (Dockerfile, pyproject.toml), tests, continuous integration, documentation

Not found: CITATION.cff

Tools: PyTorch (165 files), NumPy (158 files), anndata (107 files), pandas (72 files), SciPy (38 files), PyTorch Lightning (22 files), scikit-learn (21 files), Pyro (18 files), Scanpy (8 files), xarray (6 files), PyTorch Geometric (5 files), h5py (4 files), Matplotlib (4 files), Biopython (1 file), Numba (1 file), seaborn (1 file), SHAP (1 file), statsmodels (1 file)

Availability: 1 check, the latest on 27 September 2026: the link answers

27 September 2026: the link answers

345 files

conftest.py, Python, 62 lines
docs/_static/js/custom.js, JavaScript, 164 lines
docs/conf.py, Python, 343 lines
docs/extensions/edit_colab_url.py, Python, 39 lines
docs/extensions/filterlist.py, Python, 319 lines
docs/extensions/typed_returns.py, Python, 35 lines
src/scvi/__init__.py, Python, 31 lines
src/scvi/_constants.py, Python, 21 lines
src/scvi/_settings.py, Python, 231 lines
src/scvi/_types.py, Python, 13 lines
src/scvi/autotune/__init__.py, Python, 34 lines
src/scvi/autotune/_experiment.py, Python, 896 lines
src/scvi/autotune/_tune.py, Python, 181 lines
src/scvi/criticism/__init__.py, Python, 4 lines
src/scvi/criticism/_constants.py, Python, 14 lines
src/scvi/criticism/_create_criticism_report.py, Python, 226 lines
src/scvi/criticism/_ppc.py, Python, 481 lines
src/scvi/data/__init__.py, Python, 68 lines
src/scvi/data/_anntorchdataset.py, Python, 169 lines
src/scvi/data/_built_in_data/__init__.py, Python, 1 line
src/scvi/data/_built_in_data/_brain_large.py, Python, 112 lines
src/scvi/data/_built_in_data/_cellxgene.py, Python, 62 lines
src/scvi/data/_built_in_data/_cite_seq.py, Python, 187 lines
src/scvi/data/_built_in_data/_cortex.py, Python, 58 lines
src/scvi/data/_built_in_data/_csv.py, Python, 49 lines
src/scvi/data/_built_in_data/_dataset_10x.py, Python, 171 lines
src/scvi/data/_built_in_data/_heartcellatlas.py, Python, 47 lines
src/scvi/data/_built_in_data/_loom.py, Python, 103 lines
src/scvi/data/_built_in_data/_pbmc.py, Python, 121 lines
src/scvi/data/_built_in_data/_smfish.py, Python, 81 lines
src/scvi/data/_built_in_data/_synthetic.py, Python, 113 lines
src/scvi/data/_compat.py, Python, 173 lines
src/scvi/data/_constants.py, Python, 60 lines
src/scvi/data/_datasets.py, Python, 698 lines
src/scvi/data/_download.py, Python, 116 lines
src/scvi/data/_manager.py, Python, 570 lines
src/scvi/data/_preprocessing.py, Python, 492 lines
src/scvi/data/_read.py, Python, 71 lines
src/scvi/data/_utils.py, Python, 383 lines
src/scvi/data/fields/__init__.py, Python, 64 lines
src/scvi/data/fields/_arraylike_field.py, Python, 522 lines
src/scvi/data/fields/_base_field.py, Python, 151 lines
src/scvi/data/fields/_dataframe_field.py, Python, 293 lines
src/scvi/data/fields/_layer_field.py, Python, 147 lines
src/scvi/data/fields/_mudata.py, Python, 143 lines
src/scvi/data/fields/_protein.py, Python, 165 lines
src/scvi/data/fields/_scanvi.py, Python, 113 lines
src/scvi/data/fields/_uns_field.py, Python, 84 lines
src/scvi/dataloaders/__init__.py, Python, 33 lines
src/scvi/dataloaders/_ann_dataloader.py, Python, 138 lines
src/scvi/dataloaders/_anncollection.py, Python, 120 lines
src/scvi/dataloaders/_concat_dataloader.py, Python, 95 lines
src/scvi/dataloaders/_custom_dataloaders.py, Python, 1,521 lines
src/scvi/dataloaders/_data_splitting.py, Python, 771 lines
src/scvi/dataloaders/_samplers.py, Python, 89 lines
src/scvi/dataloaders/_semi_dataloader.py, Python, 114 lines
src/scvi/distributions/__init__.py, Python, 24 lines
src/scvi/distributions/_beta_binomial.py, Python, 153 lines
src/scvi/distributions/_constraints.py, Python, 41 lines
src/scvi/distributions/_gamma.py, Python, 201 lines
src/scvi/distributions/_lognormal.py, Python, 316 lines
src/scvi/distributions/_negative_binomial.py, Python, 724 lines
src/scvi/distributions/_normal.py, Python, 57 lines
src/scvi/distributions/_utils.py, Python, 94 lines
src/scvi/external/__init__.py, Python, 50 lines
src/scvi/external/cellassign/__init__.py, Python, 4 lines
src/scvi/external/cellassign/_model.py, Python, 294 lines
src/scvi/external/cellassign/_module.py, Python, 265 lines
src/scvi/external/contrastivevi/__init__.py, Python, 11 lines
src/scvi/external/contrastivevi/_contrastive_data_splitting.py, Python, 218 lines
src/scvi/external/contrastivevi/_contrastive_dataloader.py, Python, 123 lines
src/scvi/external/contrastivevi/_model.py, Python, 860 lines
src/scvi/external/contrastivevi/_module.py, Python, 610 lines
src/scvi/external/cytovi/__init__.py, Python, 15 lines
src/scvi/external/cytovi/_constants.py, Python, 22 lines
src/scvi/external/cytovi/_model.py, Python, 1,241 lines
src/scvi/external/cytovi/_module.py, Python, 576 lines
src/scvi/external/cytovi/_plotting.py, Python, 268 lines
src/scvi/external/cytovi/_preprocessing.py, Python, 408 lines
src/scvi/external/cytovi/_read_write.py, Python, 270 lines
src/scvi/external/cytovi/_utils.py, Python, 229 lines
src/scvi/external/decipher/__init__.py, Python, 4 lines
src/scvi/external/decipher/_components.py, Python, 99 lines
src/scvi/external/decipher/_model.py, Python, 360 lines
src/scvi/external/decipher/_module.py, Python, 183 lines
src/scvi/external/decipher/_trainingplan.py, Python, 143 lines
src/scvi/external/decipher/utils/__init__.py, Python, 4 lines
src/scvi/external/decipher/utils/_rotate.py, Python, 109 lines
src/scvi/external/decipher/utils/_trajectory.py, Python, 138 lines
src/scvi/external/diagvi/__init__.py, Python, 7 lines
src/scvi/external/diagvi/_base_components.py, Python, 270 lines
src/scvi/external/diagvi/_model.py, Python, 1,245 lines
src/scvi/external/diagvi/_module.py, Python, 597 lines
src/scvi/external/diagvi/_task.py, Python, 525 lines
src/scvi/external/diagvi/_utils.py, Python, 384 lines
src/scvi/external/drvi/__init__.py, Python, 14 lines
src/scvi/external/drvi/_base_components.py, Python, 313 lines
src/scvi/external/drvi/_constants.py, Python, 10 lines
src/scvi/external/drvi/_distributions.py, Python, 132 lines
src/scvi/external/drvi/_generative_mixin.py, Python, 156 lines
src/scvi/external/drvi/_interpretability_mixin.py, Python, 722 lines
src/scvi/external/drvi/_model.py, Python, 109 lines
src/scvi/external/drvi/_module.py, Python, 348 lines
src/scvi/external/drvi/_trainingplan.py, Python, 48 lines
src/scvi/external/drvi/_utils.py, Python, 244 lines
src/scvi/external/gimvi/__init__.py, Python, 4 lines
src/scvi/external/gimvi/_model.py, Python, 690 lines
src/scvi/external/gimvi/_module.py, Python, 529 lines
src/scvi/external/gimvi/_task.py, Python, 195 lines
src/scvi/external/gimvi/_utils.py, Python, 125 lines
src/scvi/external/joint_embedding_scvi/__init__.py, Python, 4 lines
src/scvi/external/joint_embedding_scvi/_model.py, Python, 157 lines
src/scvi/external/joint_embedding_scvi/_module.py, Python, 222 lines
src/scvi/external/joint_embedding_scvi/_utils.py, Python, 159 lines
src/scvi/external/methylvi/__init__.py, Python, 17 lines
src/scvi/external/methylvi/_base_components.py, Python, 525 lines
src/scvi/external/methylvi/_constants.py, Python, 9 lines
src/scvi/external/methylvi/_methylanvi_model.py, Python, 275 lines
src/scvi/external/methylvi/_methylanvi_module.py, Python, 270 lines
src/scvi/external/methylvi/_methylvi_model.py, Python, 267 lines
src/scvi/external/methylvi/_methylvi_module.py, Python, 283 lines
src/scvi/external/methylvi/_utils.py, Python, 68 lines
src/scvi/external/mrvi/__init__.py, Python, 5 lines
src/scvi/external/mrvi/_components.py, Python, 432 lines
src/scvi/external/mrvi/_model.py, Python, 2,083 lines
src/scvi/external/mrvi/_module.py, Python, 714 lines
src/scvi/external/mrvi/_types.py, Python, 103 lines
src/scvi/external/mrvi/_utils.py, Python, 25 lines
src/scvi/external/poissonvi/__init__.py, Python, 3 lines
src/scvi/external/poissonvi/_model.py, Python, 439 lines
src/scvi/external/resolvi/__init__.py, Python, 4 lines
src/scvi/external/resolvi/_model.py, Python, 748 lines
src/scvi/external/resolvi/_module.py, Python, 1,329 lines
src/scvi/external/resolvi/_utils.py, Python, 566 lines
src/scvi/external/scar/__init__.py, Python, 4 lines
src/scvi/external/scar/_model.py, Python, 385 lines
src/scvi/external/scar/_module.py, Python, 384 lines
src/scvi/external/scbasset/__init__.py, Python, 4 lines
src/scvi/external/scbasset/_model.py, Python, 446 lines
src/scvi/external/scbasset/_module.py, Python, 398 lines
src/scvi/external/scviva/__init__.py, Python, 13 lines
src/scvi/external/scviva/_components.py, Python, 262 lines
src/scvi/external/scviva/_constants.py, Python, 30 lines
src/scvi/external/scviva/_log_likelihood.py, Python, 109 lines
src/scvi/external/scviva/_model.py, Python, 1,173 lines
src/scvi/external/scviva/_module.py, Python, 638 lines
src/scvi/external/scviva/differential_expression/__init__.py, Python, 11 lines
src/scvi/external/scviva/differential_expression/_de_utils.py, Python, 113 lines
src/scvi/external/scviva/differential_expression/_marker_classifier.py, Python, 105 lines
src/scvi/external/scviva/differential_expression/_niche_de_core.py, Python, 233 lines
src/scvi/external/scviva/differential_expression/_results_dataclass.py, Python, 259 lines
src/scvi/external/solo/__init__.py, Python, 3 lines
src/scvi/external/solo/_model.py, Python, 484 lines
src/scvi/external/stereoscope/__init__.py, Python, 4 lines
src/scvi/external/stereoscope/_model.py, Python, 394 lines
src/scvi/external/stereoscope/_module.py, Python, 265 lines
src/scvi/external/sysvi/__init__.py, Python, 4 lines
src/scvi/external/sysvi/_base_components.py, Python, 221 lines
src/scvi/external/sysvi/_model.py, Python, 259 lines
src/scvi/external/sysvi/_module.py, Python, 527 lines
src/scvi/external/sysvi/_priors.py, Python, 206 lines
src/scvi/external/tangram/__init__.py, Python, 4 lines
src/scvi/external/tangram/_model.py, Python, 355 lines
src/scvi/external/tangram/_module.py, Python, 164 lines
src/scvi/external/totalanvi/__init__.py, Python, 4 lines
src/scvi/external/totalanvi/_model.py, Python, 540 lines
src/scvi/external/totalanvi/_module.py, Python, 452 lines
src/scvi/external/velovi/__init__.py, Python, 4 lines
src/scvi/external/velovi/_constants.py, Python, 9 lines
src/scvi/external/velovi/_model.py, Python, 1,153 lines
src/scvi/external/velovi/_module.py, Python, 656 lines
src/scvi/external/vivs/__init__.py, Python, 15 lines
src/scvi/external/vivs/_constants.py, Python, 10 lines
src/scvi/external/vivs/_model.py, Python, 708 lines
src/scvi/external/vivs/_module.py, Python, 211 lines
src/scvi/external/vivs/_plotting.py, Python, 137 lines
src/scvi/external/vivs/_utils.py, Python, 94 lines
src/scvi/hub/__init__.py, Python, 8 lines
src/scvi/hub/_constants.py, Python, 26 lines
src/scvi/hub/_metadata.py, Python, 364 lines
src/scvi/hub/_model.py, Python, 621 lines
src/scvi/hub/_template.py, Python, 210 lines
src/scvi/hub/_url.py, Python, 57 lines
src/scvi/model/__init__.py, Python, 51 lines
src/scvi/model/_amortizedlda.py, Python, 272 lines
src/scvi/model/_autozi.py, Python, 295 lines
src/scvi/model/_condscvi.py, Python, 471 lines
src/scvi/model/_destvi.py, Python, 700 lines
src/scvi/model/_linear_scvi.py, Python, 192 lines
src/scvi/model/_mlxscvi.py, Python, 216 lines
src/scvi/model/_multivi.py, Python, 1,282 lines
src/scvi/model/_peakvi.py, Python, 665 lines
src/scvi/model/_scanvi.py, Python, 345 lines
src/scvi/model/_scvi.py, Python, 258 lines
src/scvi/model/_totalvi.py, Python, 1,557 lines
src/scvi/model/_utils.py, Python, 427 lines
src/scvi/model/base/__init__.py, Python, 58 lines
src/scvi/model/base/_archesmixin.py, Python, 540 lines
src/scvi/model/base/_base_model.py, Python, 2,016 lines
src/scvi/model/base/_constants.py, Python, 16 lines
src/scvi/model/base/_de_core.py, Python, 219 lines
src/scvi/model/base/_differential.py, Python, 743 lines
src/scvi/model/base/_embedding_mixin.py, Python, 38 lines
src/scvi/model/base/_log_likelihood.py, Python, 136 lines
src/scvi/model/base/_mlxmixin.py, Python, 283 lines
src/scvi/model/base/_pyromixin.py, Python, 718 lines
src/scvi/model/base/_rnamixin.py, Python, 963 lines
src/scvi/model/base/_save_load.py, Python, 246 lines
src/scvi/model/base/_training_mixin.py, Python, 714 lines
src/scvi/model/base/_vaemixin.py, Python, 540 lines
src/scvi/model/utils/__init__.py, Python, 17 lines
src/scvi/model/utils/_annbatch_de_core.py, Python, 471 lines
src/scvi/model/utils/_minification.py, Python, 67 lines
src/scvi/module/__init__.py, Python, 48 lines
src/scvi/module/_amortizedlda.py, Python, 361 lines
src/scvi/module/_autozivae.py, Python, 413 lines
src/scvi/module/_classifier.py, Python, 79 lines
src/scvi/module/_constants.py, Python, 27 lines
src/scvi/module/_mlxvae.py, Python, 392 lines
src/scvi/module/_mrdeconv.py, Python, 676 lines
src/scvi/module/_multivae.py, Python, 1,060 lines
src/scvi/module/_peakvae.py, Python, 348 lines
src/scvi/module/_scanvae.py, Python, 278 lines
src/scvi/module/_totalvae.py, Python, 866 lines
src/scvi/module/_utils.py, Python, 32 lines
src/scvi/module/_vae.py, Python, 893 lines, 1 match
src/scvi/module/_vaec.py, Python, 382 lines
src/scvi/module/base/__init__.py, Python, 25 lines
src/scvi/module/base/_base_module.py, Python, 595 lines
src/scvi/module/base/_decorators.py, Python, 116 lines
src/scvi/module/base/_embedding_mixin.py, Python, 50 lines
src/scvi/module/base/_priors.py, Python, 301 lines
src/scvi/module/base/_pyro.py, Python, 11 lines
src/scvi/nn/__init__.py, Python, 28 lines
src/scvi/nn/_base_components.py, Python, 1,099 lines
src/scvi/nn/_embedding.py, Python, 98 lines
src/scvi/nn/_utils.py, Python, 10 lines
src/scvi/train/__init__.py, Python, 70 lines
src/scvi/train/_callbacks.py, Python, 285 lines
src/scvi/train/_config.py, Python, 388 lines
src/scvi/train/_constants.py, Python, 17 lines
src/scvi/train/_logger.py, Python, 124 lines
src/scvi/train/_metrics.py, Python, 91 lines
src/scvi/train/_progress.py, Python, 91 lines
src/scvi/train/_trainer.py, Python, 228 lines
src/scvi/train/_trainingplans.py, Python, 1,927 lines
src/scvi/train/_trainrunner.py, Python, 269 lines
src/scvi/utils/__init__.py, Python, 21 lines
src/scvi/utils/_attrdict.py, Python, 12 lines
src/scvi/utils/_decorators.py, Python, 18 lines
src/scvi/utils/_dependencies.py, Python, 44 lines
src/scvi/utils/_docstrings.py, Python, 258 lines
src/scvi/utils/_mlflow.py, Python, 144 lines
src/scvi/utils/_track.py, Python, 59 lines
tests/__init__.py, Python, 1 line
tests/autotune/test_experiment.py, Python, 1,181 lines
tests/autotune/test_tune.py, Python, 336 lines
tests/conftest.py, Python, 168 lines
tests/criticism/test_criticism.py, Python, 122 lines
tests/data/__init__.py, Python, 1 line
tests/data/conftest.py, Python, 20 lines
tests/data/test_anndata.py, Python, 548 lines
tests/data/test_anntorchdataset.py, Python, 176 lines
tests/data/test_built_in_data.py, Python, 112 lines
tests/data/test_data_utils.py, Python, 32 lines
tests/data/test_dataset10X.py, Python, 47 lines
tests/data/test_mudata.py, Python, 394 lines
tests/data/test_preprocessing.py, Python, 104 lines
tests/data/test_synthetic_iid.py, Python, 17 lines
tests/data/utils.py, Python, 142 lines
tests/dataloaders/__init__.py, Python, 1 line
tests/dataloaders/sparse_utils.py, Python, 118 lines
tests/dataloaders/test_annbatch.py, Python, 1,014 lines
tests/dataloaders/test_custom_dataloader.py, Python, 1,205 lines
tests/dataloaders/test_dataloaders.py, Python, 301 lines
tests/dataloaders/test_datasplitter.py, Python, 272 lines
tests/dataloaders/test_samplers.py, Python, 155 lines
tests/distributions/test_beta_binomial.py, Python, 77 lines
tests/distributions/test_constraints.py, Python, 21 lines
tests/distributions/test_gamma.py, Python, 154 lines
tests/distributions/test_lognormal.py, Python, 146 lines
tests/distributions/test_negative_binomial.py, Python, 230 lines
tests/external/cellassign/test_model_cellassign.py, Python, 61 lines
tests/external/contrastivevi/test_contrastive_dataloaders.py, Python, 54 lines
tests/external/contrastivevi/test_contrastive_datasplitter.py, Python, 158 lines
tests/external/contrastivevi/test_contrastivevae.py, Python, 458 lines
tests/external/contrastivevi/test_contrastivevi.py, Python, 165 lines
tests/external/cytovi/test_cytovi.py, Python, 301 lines
tests/external/decipher/test_decipher.py, Python, 135 lines
tests/external/diagvi/test_diagvi.py, Python, 1,680 lines
tests/external/drvi/test_drvi.py, Python, 641 lines
tests/external/gimvi/test_gimvi.py, Python, 232 lines
tests/external/joint_embedding_scvi/test_joint_embedding_scvi.py, Python, 151 lines
tests/external/joint_embedding_scvi/test_joint_embedding_utils.py, Python, 61 lines
tests/external/methylvi/test_methylanvi.py, Python, 45 lines
tests/external/methylvi/test_methylvi.py, Python, 62 lines
tests/external/mrvi/test_mrvi_components.py, Python, 82 lines
tests/external/mrvi/test_mrvi_model.py, Python, 375 lines
tests/external/poissonvi/test_poissonvi.py, Python, 85 lines
tests/external/resolvi/test_resolvi.py, Python, 227 lines
tests/external/scar/test_scar.py, Python, 27 lines
tests/external/scbasset/test_scbasset.py, Python, 40 lines
tests/external/scviva/test_scviva.py, Python, 478 lines
tests/external/solo/test_solo.py, Python, 113 lines
tests/external/stereoscope/test_stereoscope.py, Python, 42 lines
tests/external/sysvi/test_sysvi.py, Python, 343 lines
tests/external/tangram/test_tangram.py, Python, 95 lines
tests/external/totalanvi/test_totalanvi.py, Python, 517 lines
tests/external/velovi/test_velovi.py, Python, 65 lines
tests/external/vivs/test_vivs.py, Python, 467 lines
tests/hub/test_hub_metadata.py, Python, 196 lines
tests/hub/test_hub_model.py, Python, 399 lines
tests/hub/test_url.py, Python, 49 lines
tests/model/base/test_base_model.py, Python, 90 lines
tests/model/base/test_rnamixin.py, Python, 33 lines
tests/model/test_amortizedlda.py, Python, 124 lines
tests/model/test_autozi.py, Python, 167 lines
tests/model/test_condscvi.py, Python, 81 lines
tests/model/test_destvi.py, Python, 194 lines
tests/model/test_differential.py, Python, 163 lines
tests/model/test_differential_abundance.py, Python, 174 lines
tests/model/test_docstrings.py, Python, 38 lines
tests/model/test_linear_scvi.py, Python, 157 lines
tests/model/test_mlxscvi.py, Python, 224 lines
tests/model/test_models_with_minified_data.py, Python, 546 lines
tests/model/test_models_with_mudata_minified_data.py, Python, 512 lines
tests/model/test_multigpu.py, Python, 338 lines
tests/model/test_multivi.py, Python, 609 lines
tests/model/test_peakvi.py, Python, 238 lines
tests/model/test_pyro.py, Python, 562 lines
tests/model/test_scanvi.py, Python, 741 lines
tests/model/test_scvi.py, Python, 1,716 lines
tests/model/test_totalvi.py, Python, 831 lines
tests/model/test_utils.py, Python, 33 lines
tests/module/test_scanvae.py, Python, 41 lines
tests/module/test_vae.py, Python, 71 lines
tests/nn/test_embedding.py, Python, 76 lines
tests/nn/test_fclayers.py, Python, 245 lines
tests/test_warnings.py, Python, 46 lines
tests/train/__init__.py, Python, 1 line
tests/train/test_callbacks.py, Python, 267 lines
tests/train/test_config.py, Python, 44 lines
tests/train/test_trainingplans.py, Python, 212 lines
LICENSE, License, 30 lines
README.md, Text, 147 lines

At the source: github.com/scverse/scvi-tools

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
343 scripts, each with its path and the digest of its content;
1 match between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

Datasets cited

zenodo:15320250, at Zenodo; found in “Data and materials availability”

Data and materials availability

The integrated C. robusta larva-stage single-cell RNA sequencing dataset and the complete differential expression data for all clusters shown in Figure 4C are deposited at Zenodo (https://doi.org/10.5281/zenodo.15320250).

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 5 authors, 5 keywords, 7 MeSH terms, 2 funders, 104 references.

Cite

This paper

Kourakis, M. J., Miao, Y., Newman-Smith, E. D., Ryan, K., & Smith, W. C. (2026). Similarities between Ciona Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor? eNeuro, 13(8), ENEURO.0362-25.2026. https://doi.org/10.1523/eneuro.0362-25.2026

BibTeX

@article{kourakis2026similarities, author = {Kourakis, Matthew J. and Miao, Yishen and Newman-Smith, Erin D. and Ryan, Kerrianne and Smith, William C.}, title = {{Similarities between \<i\>Ciona\</i\> Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?}}, journal = {eNeuro}, year = {2026}, month = aug, volume = {13}, number = {8}, pages = {ENEURO.0362--25.2026}, publisher = {Society for Neuroscience}, issn = {2373-2822}, doi = {10.1523/eneuro.0362-25.2026}, url = {https://doi.org/10.1523/eneuro.0362-25.2026}, pmid = {42637546}, pmcid = {PMC13509005} }

RIS

TY - JOUR AU - Kourakis, Matthew J. AU - Miao, Yishen AU - Newman-Smith, Erin D. AU - Ryan, Kerrianne AU - Smith, William C. TI - Similarities between Ciona Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor? T2 - eNeuro J2 - eNeuro PY - 2026 DA - 2026/08/24 VL - 13 IS - 8 SP - ENEURO.0362 EP - 25.2026 SN - 2373-2822 PB - Society for Neuroscience DO - 10.1523/eneuro.0362-25.2026 UR - https://doi.org/10.1523/eneuro.0362-25.2026 LA - en ER -

CSL-JSON

{ "id": "10.1523/eneuro.0362-25.2026", "type": "article-journal", "title": "Similarities between Ciona Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?", "container-title": "eNeuro", "author": [ { "family": "Kourakis", "given": "Matthew J." }, { "family": "Miao", "given": "Yishen" }, { "family": "Newman-Smith", "given": "Erin D." }, { "family": "Ryan", "given": "Kerrianne" }, { "family": "Smith", "given": "William C." } ], "container-title-short": "eNeuro", "volume": "13", "issue": "8", "page": "ENEURO.0362-25.2026", "DOI": "10.1523/eneuro.0362-25.2026", "PMID": "42637546", "PMCID": "PMC13509005", "ISSN": "2373-2822", "publisher": "Society for Neuroscience", "URL": "https://doi.org/10.1523/eneuro.0362-25.2026", "language": "en", "issued": { "date-parts": [ [ 2026, 8, 24 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1093/nar/gkag706 [code]: scDifformer: diffusion-based post-training for virtual cell modeling across large-scale single-cell data.
Journal: Nucleic acids research
In common: PyTorch Geometric, anndata, Numba, 10 other tools, 1 reference
[2] doi:10.1038/s41592-026-03057-2 [code]: CREsted: modeling genomic and synthetic cell-type-specific enhancers across tissues and species.
Journal: Nature methods
In common: Biopython, anndata, Numba, 10 other tools, 1 reference
[3] doi:10.1016/j.isci.2026.116055 [code]: Mapping the transcriptional diversity of calcium signaling in the mouse and human brain.
Journal: iScience
In common: PyTorch Geometric, anndata, Numba, 10 other tools, 1 reference
[4] doi:10.1093/bioinformatics/btag652 [code]: mmVelo: a deep generative model for estimating cell state-dependent dynamics across multiple modalities.
Journal: Bioinformatics (Oxford, England)
In common: Pyro, PyTorch Lightning, anndata, 9 other tools, 1 reference
[5] doi:10.1038/s41586-026-10629-x [code]: Whole-genome duplication shaped cell-type evolution in the vertebrate brain.
Journal: Nature
In common: anndata, Numba, Scanpy, 7 other tools, 3 references
[6] doi:10.1002/advs.77003 [code]: SemanticST: A Scalable Multi-Contextual Graph Learning Framework for Uncovering Spatial Niches and Robust Multi-Sample Integration in Spatial Transcriptomics.
Journal: Advanced science (Weinheim, Baden-Wurttemberg, Germany)
In common: PyTorch Geometric, anndata, Numba, 9 other tools, 1 reference
[7] doi:10.1093/bioinformatics/btag540 [code]: Deciphering spatial heterogeneity by multimodal spatial transcriptomics modelling with SpatialModal.
Journal: Bioinformatics (Oxford, England)
In common: PyTorch Geometric, anndata, Numba, 9 other tools, 1 reference
[8] doi:10.1038/s42003-026-10957-8 [code]: Brain defence by the extracellular matrix protein Cochlin.
Journal: Communications biology
In common: PyTorch Lightning, Biopython, SHAP, 9 other tools
[9] doi:10.64898/2026.03.30.714220 [code]: An integrated single cell and spatial omics atlas of human prenatal development
Journal: bioRxiv (preprint)
In common: Biopython, anndata, Scanpy, 9 other tools, 1 reference
[10] doi:10.1038/s41467-026-74694-6 [code]: Semi-supervised Omics Factor Analysis (SOFA) disentangles known and latent sources of variation in multi-omic data.
Journal: Nature communications
In common: Pyro, anndata, Scanpy, 8 other tools, 1 reference

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.

Similarities between <i>Ciona</i> Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?

Code ↔ Paper

Paper

The authors' code

Overview

Abstract

Repository

scverse/scvi-tools

Tracing map

Data

Datasets cited

Data and materials availability

Versions

Version 1, 27 September 2026: the first record

Cite

This paper

Similar papers

Contribute

Claim this paper

Correct its record

Validate its tracing map

Add the badge to its README

Request its removal

Discussion, reproductions, activity

Similarities between &lt;i&gt;Ciona&lt;/i&gt; Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?

Code ↔ Paper

Overview

Abstract

Repository

scverse/scvi-tools

Tracing map

Data

Datasets cited

Data and materials availability

Versions

Version 1, 27 September 2026: the first record

Cite

This paper

Similar papers

Contribute

Claim this paper

Correct its record

Validate its tracing map

Add the badge to its README

Request its removal

Discussion, reproductions, activity

Similarities between <i>Ciona</i> Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?