Allowing to specify 'span' parameter in the Preprocessor

[Szym29]

Hi,

I tried to fine tune the model on my data for integration. However, the sparsity of my data is higher than the common scRNA-seq dataset. And when I used the scgpt.preprocess.Preprocessor to subset the highly variable genes, an error occurred.

ValueError: b'There are other near singularities as well. . 0.090619

I found this relevant issue . I manually set the span as 0.5 in the Preprocessor class, and it worked fine then.

        # step 5: subset hvg
        if self.subset_hvg:
            logger.info("Subsetting highly variable genes ...")
            if batch_key is None:
                logger.warning(
                    "No batch_key is provided, will use all cells for HVG selection."
                )
            sc.pp.highly_variable_genes(
                adata,
                layer=self.hvg_use_key,
                n_top_genes=self.subset_hvg
                if isinstance(self.subset_hvg, int)
                else None,
                batch_key=batch_key,
                flavor=self.hvg_flavor,
                span=0.5,
                subset=True,
            )

So I would suggest allowing users to specify the 'span' when they initialize the Preprocessor and pass it to sc.pp.highly_variable_genes. It would be better if all scanpy parameters could be specified in the preprocessor to improve the robustness of the method.

Thanks,
Yumin

[subercui]

Thank you for the suggestion! I will try to add these features soon, probably by additional kwargs.