Transfer_labels error

[RijndertAriese]

Hi,

Thanks for this great resource, both on the paper and the package! Very useful (and perfectly timed for my needs:) ).

I've downloaded the HNOCA data and the scPoli model from the HNOCA github and loaded them together with my query_adata. Subsequently I've trained the model using map_query and computed wknn. However, now I want to transfer the annotation from the atlas to my data, but this throws an error

TypeError

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[55], line 1
----> 1 mapper.transfer_labels(label_key='snapseed_pca_rss_level_1')

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/hnoca/map/mapper.py:266, in AtlasMapper.transfer_labels(self, label_key)
    263 if self.wknn is None:
    264     raise AttributeError("WKNN needs to be computed first.")
--> 266 scores = transfer_labels(
    267     self.ref_adata,
    268     self.query_adata,
    269     self.wknn,
    270     label_key=label_key,
    271 )
    273 return scores

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/hnoca/map/wknn.py:273, in transfer_labels(ref_adata, query_adata, wknn, label_key)
    267 scores = pd.DataFrame(
    268     wknn @ pd.get_dummies(ref_adata.obs[label_key]),
    269     columns=pd.get_dummies(ref_adata.obs[label_key]).columns,
    270     index=query_adata.obs_names,
    271 )
    272 scores["best_label"] = scores.idxmax(1)
--> 273 scores["best_score"] = scores.max(1)
    274 return scores

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/frame.py:11643, in DataFrame.max(self, axis, skipna, numeric_only, **kwargs)
  11635 @doc(make_doc("max", ndim=2))
  11636 def max(
  11637     self,
   (...)
  11641     **kwargs,
  11642 ):
> 11643     result = super().max(axis, skipna, numeric_only, **kwargs)
  11644     if isinstance(result, Series):
  11645         result = result.__finalize__(self, method="max")

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/generic.py:12401, in NDFrame.max(self, axis, skipna, numeric_only, **kwargs)
  12394 def max(
  12395     self,
  12396     axis: Axis | None = 0,
   (...)
  12399     **kwargs,
  12400 ):
> 12401     return self._stat_function(
  12402         "max",
  12403         nanops.nanmax,
  12404         axis,
  12405         skipna,
  12406         numeric_only,
  12407         **kwargs,
  12408     )

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/generic.py:12374, in NDFrame._stat_function(self, name, func, axis, skipna, numeric_only, **kwargs)
  12370 nv.validate_func(name, (), kwargs)
  12372 validate_bool_kwarg(skipna, "skipna", none_allowed=False)
> 12374 return self._reduce(
  12375     func, name=name, axis=axis, skipna=skipna, numeric_only=numeric_only
  12376 )

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/frame.py:11549, in DataFrame._reduce(self, op, name, axis, skipna, numeric_only, filter_type, **kwds)
  11545     df = df.T
  11547 # After possibly _get_data and transposing, we are now in the
  11548 #  simple case where we can use BlockManager.reduce
> 11549 res = df._mgr.reduce(blk_func)
  11550 out = df._constructor_from_mgr(res, axes=res.axes).iloc[0]
  11551 if out_dtype is not None and out.dtype != "boolean":

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/internals/managers.py:1500, in BlockManager.reduce(self, func)
   1498 res_blocks: list[Block] = []
   1499 for blk in self.blocks:
-> 1500     nbs = blk.reduce(func)
   1501     res_blocks.extend(nbs)
   1503 index = Index([None])  # placeholder

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/internals/blocks.py:404, in Block.reduce(self, func)
    398 @final
    399 def reduce(self, func) -> list[Block]:
    400     # We will apply the function and reshape the result into a single-row
    401     #  Block with the same mgr_locs; squeezing will be done at a higher level
    402     assert self.ndim == 2
--> 404     result = func(self.values)
    406     if self.values.ndim == 1:
    407         res_values = result

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/frame.py:11468, in DataFrame._reduce.<locals>.blk_func(values, axis)
  11466         return np.array([result])
  11467 else:
> 11468     return op(values, axis=axis, skipna=skipna, **kwds)

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/nanops.py:147, in bottleneck_switch.__call__.<locals>.f(values, axis, skipna, **kwds)
    145         result = alt(values, axis=axis, skipna=skipna, **kwds)
    146 else:
--> 147     result = alt(values, axis=axis, skipna=skipna, **kwds)
    149 return result

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/nanops.py:404, in _datetimelike_compat.<locals>.new_func(values, axis, skipna, mask, **kwargs)
    401 if datetimelike and mask is None:
    402     mask = isna(values)
--> 404 result = func(values, axis=axis, skipna=skipna, mask=mask, **kwargs)
    406 if datetimelike:
    407     result = _wrap_results(result, orig_values.dtype, fill_value=iNaT)

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/pandas/core/nanops.py:1098, in _nanminmax.<locals>.reduction(values, axis, skipna, mask)
   1093     return _na_for_min_count(values, axis)
   1095 values, mask = _get_values(
   1096     values, skipna, fill_value_typ=fill_value_typ, mask=mask
   1097 )
-> 1098 result = getattr(values, meth)(axis)
   1099 result = _maybe_null_out(result, axis, mask, values.shape)
   1100 return result

File /hpc/pmc_rios/2.personal/rariese/miniconda/envs/scanpy/lib/python3.10/site-packages/numpy/core/_methods.py:41, in _amax(a, axis, out, keepdims, initial, where)
     39 def _amax(a, axis=None, out=None, keepdims=False,
     40           initial=_NoValue, where=True):
---> 41     return umr_maximum(a, axis, None, out, keepdims, initial, where)

TypeError: '>=' not supported between instances of 'float' and 'str'

A snippet of my code

query_adata_hvg.obs["snapseed_pca_rss_level_1"] = "unknown"
mapper = mapping.AtlasMapper(scpoli_model)
mapper.map_query(
    query_adata_hvg,
    retrain="partial",
    max_epochs=20,
    batch_size=1024,
    unlabeled_prototype_training=False,
    n_epochs=100,
    pretraining_epochs=90,
    early_stopping_kwargs=early_stopping_kwargs,
    eta=10,
    alpha_epoch_anneal=10,
)
mapper.compute_wknn(k=100)
mapper.transfer_labels(label_key='snapseed_pca_rss_level_1')

Am I missing something? Could you maybe give me some pointers on how to properly transfer the labels?

Thanks!

Rijndert

[RijndertAriese]

Hi,

First, congrats on this beautiful publication!
I know I closed this issue, but i ran into it again today. I found out what is causing the issue tho!
Not sure how to structure this, but here is what I found:

def transfer_labels(ref_adata, query_adata, wknn, label_key="celltype"):
    scores = pd.DataFrame(
        wknn @ pd.get_dummies(ref_adata.obs[label_key]),
        columns=pd.get_dummies(ref_adata.obs[label_key]).columns,
        index=query_adata.obs_names,
    )
    scores["best_label"] = scores.idxmax(1)
    scores["best_score"] = scores.max(1)
    return scores

By first running scores["best_label"] = scores.idxmax(1) I introduce a str in my scores df, which cannot be compared against the floats of the other columns. Switched them around in my script and worked for me.

Cheers,
Rijndert

[joschif]

Heyhey, sorry for the late reply! Yes, I see, thanks for reporting, I'll push a fix soon