single-cell-foundation-model-geneformer

To tokenize raw scRNA-seq data, use the TranscriptomeTokenizer to convert input counts containing ensembl_id and n_counts into tokenized .dataset objects ready for downstream processing.

In silico perturbation simulates state changes in single-cell data to explore gene perturbations. It uses the InSilicoPerturber API to perform these analyses on tokenized datasets across various cell states.

Extract embeddings from scRNA-seq data by applying the EmbExtractor API to your tokenized .dataset files. This process converts the tokenized gene sequences into dense vector representations for downstream tasks.

Yes, single-cell tokenization requires raw counts containing ensembl_id and n_counts fields. The TranscriptomeTokenizer processes these inputs to generate the .dataset format needed for classification and perturbation workflows.

You can fine-tune a classifier on your scRNA-seq dataset by using the Classifier API with your tokenized .dataset objects. This enables cell classification tasks directly on the prepared single-cell data.

Foundation models for single-cell data require tokenized .dataset formats and specific input fields like ensembl_id and n_counts. Workflows depend on proper integration with downstream APIs for classification, embedding extraction, and perturbation.