Winnow-KAN: single-cell RNA-seq location recovery with small-gene-set spatial transcriptomics
Bibliographic record
Abstract
Keywords: Cell mapping, Deep Learning, Kolmogorov-Arnold network, Single-cell RNA-seq, Spatial transcriptomics.Single-cell RNA sequencing (scRNA-seq) has revolutionized our understanding of cellular heterogeneity. However, its collection process prevents the investigation of tissue organization due to the lack of spatial origins for the cells. Recent advances in computational methods have addressed this gap by leveraging spatial transcriptomics, which simultaneously profiles gene expression and spatial coordinates. While these state-of-the-art methods demonstrate excellent performance in cell location recovery, their effectiveness is often specific to the particular pair of scRNA-seq and spatial transcriptomics datasets used, limiting their scalability to larger datasets and generalizability to external query scRNA-seq data. In this study, we demonstrate the feasibility of leveraging a novel model architecture to address the redundancy in scRNA-seq datasets and facilitate prediction with a much smaller set of genes. We present Winnow-KAN, a method designed to reduce the number of required gene variables in cell-mapping tasks. Built on a modified structure of the Kolmogorov-Arnold Network, Winnow-KAN leverages the Kolmogorov-Arnold Representation Theorem to facilitate location predictions using fewer features than those required by multi-layer perceptron-based methods. Winnow-KAN includes a selector layer that reduces the size of the gene set used for prediction, enabling the model to recover query scRNA-seq data with performance comparable to MLP-based approaches that rely on the full gene set. We benchmarked Winnow-KAN using multiple datasets generated from brain and cancer tissues, derived from platforms such as Visium and MERFISH.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".