Abstract:Abstract Background Single-cell RNA (scRNA-seq) sequencing technologies trigger the study of individual cell gene expression and reveal the diversity within cell populations. To measure cell-to-cell similarity based on their transcription and gene expression, many dimension reduction methods are employed to retrieve corresponding low-dimensional embeddings of input scRNA-seq data to conduct clustering. However, the methods lack explainability and may not perform well with scRNA-seq data because they are not customized for high-dimensional sparse scRNA-seq data. Results In this study, we propose an explainable t-SNE: cell-driven t-SNE (c-TSNE) that fuses cell differences reflected from biologically meaningful distance metrics for input data. Our study shows that the proposed method not only enhances the interpretation of the original t-SNE visualization but also demonstrates favorable single cell segregation performance on benchmark datasets compared to state-of-the-art peers. The robustness analysis shows that the proposed cell-driven t-SNE demonstrates robustness to dropout and noise in clustering. It provides a novel and practical way to investigate the interpretability of t-SNE in scRNA-seq data analysis. Unlike the general assumption that the explainability of a machine learning method may need to compromise with learning efficiency, the proposed explainable t-SNE improves both clustering efficiency and explainability. More importantly, our work suggests that widely used t-SNE can be easily misused in existing scRNA-seq analysis, because its default Euclidean distance can bring biases or meaningless results in cell difference evaluation for high-dimensional sparse scRNA-seq data. To the best of our knowledge, it is the first explainable t-SNE proposed in scRNA-seq analysis and will inspire other explainable machine learning method development in the field. Conclusion The proposed explainable t-SNE outperforms classic t-SNE and its peers in meaningful visualization and segregation. The poor performance of the classic t-SNE highlights the importance of developing explainable machine learning methods in scRNA-seq analysis. The explainable t-SNE is a data-centric customized ML enhance efficiency in data analysis through bringing more biological insights and interpretations.

Compound-SNE: Comparative alignment of t-SNEs for multiple single-cell omics data visualisation

Compound-SNE: Comparative alignment of t-SNEs for multiple single-cell omics data visualisation

The art of using t-SNE for single-cell transcriptomics

Online t-SNE for single-cell RNA-seq

Visual cohort comparison for spatial single-cell omics-data

Automatic identification of relevant genes from low-dimensional embeddings of single-cell RNA-seq data.

Automated optimized parameters for T-distributed stochastic neighbor embedding improve visualization and analysis of large datasets

Visualizing and interpreting single-cell gene expression datasets with Similarity Weighted Nonnegative Embedding

Scalable and unbiased sequence-informed embedding of single-cell ATAC-seq data with CellSpace

Statistical Single Cell Multi-Omics Integration

Population-level Integration of Single-Cell Datasets Enables Multi-Scale Analysis Across Samples

A cross entropy test allows quantitative statistical comparison of t-SNE and UMAP representations

Tree-SNE: Hierarchical Clustering and Visualization Using t-SNE

Online Single-Cell Data Integration Through Projecting Heterogeneous Datasets into a Common Cell-Embedding Space

Index $t$-SNE: Tracking Dynamics of High-Dimensional Datasets with Coherent Embeddings

Integrated analysis of multimodal single-cell data with structural similarity

Explainable t-SNE for single-cell RNA-seq data analysis

Cell type matching across species using protein embeddings and transfer learning

ENS-t-SNE: Embedding Neighborhoods Simultaneously t-SNE

scCobra: Contrastive cell embedding learning with domain-adaptation for single-cell data integration and harmonization

Learning Consistency and Specificity of Cells from Single-cell Multi-omic Data