Research map / Large language models / Language, safety and interpretability
CLM-Access: A Specialized Foundation Model for High-Dimensional Single-Cell ATAC-Seq Analysis
Ziqiang Liu, Bowen Li, Zhenyu Xu, Yantao Li, Junwei Zhang, Chulin Sha, Xiaolin Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| binarize peak signals instead of summing within patches (ablation) | CLM-Access pretraining · scATAC pretraining on 370k cells, NMI (patch 2k) | NMI: 0.0212 -> 0.1617 | Improved |
| use peak-level BCE loss instead of patch-level MSE (ablation) | CLM-Access pretraining · scATAC pretraining on 370k cells, NMI (patch 2k) | NMI: 0.1617 -> 0.5813 | Improved |
| use peak-level BCE loss instead of patch-level MSE at 5k patches (ablation) | CLM-Access pretraining · scATAC pretraining on 370k cells, NMI (patch 5k) | NMI: 0.4781 -> 0.3648 | Hurt |
| mask only peak values instead of patch tokens and values (ablation) | CLM-Access pretraining · scATAC pretraining on 370k cells, NMI (patch 2k) | NMI: 0.0181 -> 0.5813 | Improved |
| increase number of patches from 1k to 2k with peak-level BCE (ablation) | CLM-Access pretraining · scATAC pretraining on 370k cells, NMI | NMI: 0.5078 -> 0.5813 (2k); 0.3648 (5k) | Mixed |
| scale pretraining data from 0.33M to 2.8M cells (ablation) | CLM-Access small model · scATAC zero-shot NMI | NMI: 0.5813 -> 0.6236 | Improved |
| scale model from small to medium (ablation) | CLM-Access · scATAC zero-shot NMI | NMI: 0.6236 -> 0.6618 | Improved |
| scale model from medium to large (ablation) | CLM-Access · scATAC zero-shot NMI | NMI: 0.6618 -> 0.6444 | Hurt |
| CLM-Access vs scATAnno and Cellcano for cell type annotation | CLM-Access fine-tuned · GSE219281 cell type annotation | accuracy: 0.7283 -> 0.7635 | Improved |
| CLM-Access vs BABEL for RNA expression prediction | CLM-Access fine-tuned · RNA expression prediction from scATAC (paired data) | RMSE: 1.5316 -> 1.4185 | Improved |
Closest papers by meaning
- ChromFound: Towards A Universal Foundation Model for Single-Cell Chromatin Accessibiltiy Data (NeurIPS 2025 · similarity 0.90)
- CellVerse: Do Large Language Models Really Understand Cell Biology? (NeurIPS 2025 · similarity 0.89)
- A versatile informative diffusion model for single-cell ATAC-seq data generation and analysis (NeurIPS 2024 · similarity 0.88)
- scPilot: Large Language Model Reasoning Toward Automated Single-Cell Analysis and Discovery (NeurIPS 2025 · similarity 0.88)
- A Simple and Comprehensive Benchmark for Single-Cell Transcriptomics (AAAI 2025 · similarity 0.86)
- Semi-supervised Knowledge Transfer Across Multi-omic Single-cell Data (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
