Home

About TransImp

TransImp (Translation-based Spatial Transcriptomics Imputation toolkit) is a computational framework and toolbox for reference-based imputation of spatial transcriptomic dataset.

TransImp features in two novel functions:

1. Uncertainty score estimation for each imputed gene, allowing the identification of relible predictions;

2. Spatial regularisation for explicitily control the degree of spatial pattern preservation for imputation modeling.

The toolkit also includes TransDeconv, a translation-based cell-type deconvolution method that infers per-spot cell-type proportions from spatial transcriptomics data using a scRNA-seq reference. As of v0.2, TransDeconv supports an AnnData-based interface with automatic preprocessing, marker gene selection, score-init, cluster-mapping regularization, and JS-divergence loss.

Please refer to our demo notebooks for details:

• Different configurations of TransImp applied to SeqFISH dataset dataset.

• Exploration for unprobed genes with SeqFISH ST dataset.

• Cell type deconvolution with TransDeconv.

References

Manuscript of TransImp with more details is available on bioRxiv now and is currently under review.